A LEXICAL PHONOLOGY OF BASQUE

by
José Ignacio Hualde

A Dissertation Presented to the

FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Linguistics)

August 1988

Copyright 1988 José Ignacio Hualde

 UMI Number: DP29068

All rights reserved

INFORMATION TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

OissedBtion PWMIsMng

UMI DP29068
Published by ProQuest LLC (2014). Copyright in the Dissertation held by the Author.
Microform Edition © ProQuest LLC.
All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code

ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106- 1346
 UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CAUFORNIA 90089

This dissertation, written by

......................J o s e ^ I g n a c i o ^^Hu a i d e ....................

under the direction of h A .? .. . . . . Dissertation

Committee, and approved by all its members,
has been presented to and accepted by The
Graduate School, in partial fulfillment of re­
quirements for the degree of
DO CTO R OF PHILOSOPHY

Dean o f Graduate Studies

June 8 , 1988
Date ...

DISSERTATION COMMITTEE

Chairperson
 Acknowledgments

A number of people must be thanked for whatever valuable

or interesting there is in this piece of writing.
A first subgroup of people to whom thanks must be given by
me are those who taught me Linguistics. I consider myself
extremely fortunate to have spent these past few/many years at
u s e where I have been able to learn from a group of truly
excellent people, both as scholars and as human beings. My
gratitude goes to all my teachers. Especially thankable are my
dissertation committee members: Larry Hyman, who has been
extraordinarily supportive and accessible since my first day in
the Linguistics Department and from whom I first learnt about
phonology and linguistic field work; Bernard Comrie, who has
taught me many a thing; most notably to have clarity in the
argumentation (I hope I have assimilated some of this); Mario
Saltarelli, with whom I share an interest in Basque; and lastly
but mostly, Doug Pulleyblank, my debt to whom is too great and
varied to even try to specify. I, therefore, will not try to specify
it here. Let it only be said that there is little that I saw that he
didn't first tell me to look at.
In a separate paragraph, I want to thank those who
graciously gave me their words: Amaia Iparragirre, Xabier
Arbizu, Izaskun Intxausti, Arantza Apraiz, José Antonio
 lii

Sarasola, Karmele Ikazuriaga, Edurne Arregi, Txomin Irungaray,

Inaki Alberdi, Lourdes Telletxea, Goizane Zabala, Jose Mari
Carrere, Lourdes Aranguren and others from the four corners of
the Basque Country whose names I may not remember now.
In a third subgroup of people to be thanked, I must mention
Jon Ortiz de Urbina and Pello Salaburu, who read parts of the
typescript and answered written questions about specific points
of the data with commendable promptness. Heather Goad
patiently listened to several hours of recorded data and checked
tone patterns in the transcriptions.
I will give Joyce Tolliver her own paragraph. Joyce, in
addition to unthankable emotional support, actually helped in the
actual process of writting this thesis by lending me her musical
ear for the transcription of tone in my recordings.
The data on tone/stress as well as some of the data that
appear in other chapters, were mostly collected in the summer
of 1987. I must thank the Del Amo Foundation for giving me a
grant that allowed me to spend the summer collecting data in
the Basque Country.
Writing a dissertation is not an easy task; but reading one
is not easy either. That is why this dissertation is dedicated to
YOU.
 IV

TABLE OF CONTENTS
Chapter 1: Introduction
0. Aims and structure................................................................ 1
1. Theoretical Assumptions..................................................... 3
1.1. Lexical Phonology................................................ 3
1.2. Hierarchical representations ............................. 5
1.2.1. The representation of segments ... 5
1.2.2. Underspecification.............................. 9
1.2.3. Locality ........................................... 12
2. Background information on Basque.................................... 14
2.1. Geolinguistics........................................................ 14
2.2. The sound inventory.............................................. 17
2.3. Basic phonotactic constraints ......................... 20
2.3.1. Word-initial restrictions .................. 20
2.3.2. Syllable and word-final
restrictions............................................. 22
2.3.3. Syllable-internal restrictions ....... 24
2.4. The structure of words and phrases ............... 26
2.4.1. Word formation.................................... 27
2.4.2. Structure of the noun p h ra s e 28
2.4.3. The verbal complex............................. 35
2.4.4. Blocking of periphrastic
constructions........................................ 38
3. Summary................................................................................ 41
 V

Chapter 2 : The Structure of the Lexicon

0. introduction............................................................................ 42
1. The rule of Vowel Assimilation............................. 43
2. Vowel Assimilation in Baztan.............................................. 46
2.1. Interaction with other ru le s ............................... 46
2.1.1. Mid Vowel Raising................................ 46
2.1.2. Stem-Final Low Vowel Deletion .... 49
2.1.3. Palatalization....................... ................ 50
2.2. Domain of Vowel Assimilation in Baztan .... 53
2.2.1. Morpheme-internal.............................. 53
2.2.2. Inflectional morphology .................... 55
2.2.3. Derivational m orphology................. 57
2.2.4. Compounds........................................... 59
2.2.5. Vowel Assimilation across word-
boundaries: cliticization .................. 60
2.2.6. Exceptions............................................ 63
2.3. Analysis................................................................. 65
3. Arbizu Lexical Phonology...................................................... 71
3.1. Vowel Assimilation............................................... 71
3.1.1. Exceptions to Vowel Assimilation 78
3.2. Palatalization........................................................ 81
3.3. Consonant Insertion.............................................. 86
4. Ondarroa................................................................................ 93
4.1. Vowel Assimilation.............................................. 93
 VI

4.1.1. Domain of application of Vowel

Assimilation in O ndarroa................. 98
4.2. Palatalization......................................................... 1 14
4.3. Flapping................................................................. 1 19
4.4. Summary............................................................... 121
5. Is stratum II the postlexical stratum? ............................ 121

Chapter 3: Stem Alternations

0. Introduction.......................................................................... 125
1. Phonological rules governing stem alternations ........ 126
3. Alternating and non-alternating stems ........................... 1 37
4. Lexical domain of the Derivational Stem Formation
rules........................................................................................ 1 40
5. Alternating adjectives............................................................ 1 49

Chapter 4: Rule Interaction in a Hierarchical Model of

Phonological Representations and in Lexical Phonology
0. Introduction............................................................................ 1 53
1. Place Assimilation of Nasals and Laterals M 54 T
2. Continuancy Assignment....................................................... 162
3. Palatalization.......................................................................... 176
3.1. Palatalization in O ndarroa................................... 1 78
3.1.1. Lexical and postlexical rule
application............................................. 184
3.2. A restrictive dialect: Baztan .............................. 189
 VII

3.3. An intermediate dialect: Donostia-

San Sébastian.......................................................... 200
3.4. Affective Palatalization........................................ 202
4. Conclusion................................ 206

Chapter 5: The Structure and Behavior of Affricates

0. Introduction............................................................................. 207
1. Affricates in Generative Phonology.................................... 208
2. Affricates are single segm ents............................................ 211
3. Stop Deletion.......................................................................... 213
4. Excursus: where is the feature[continuant]? .................. 218
5. Sibilant harmony..................................................................... 222
6. Fricativization.......................................................................... 227
7. Voice Assimilation.................................................................. 230
8. Palatalization ......................................................................... 235

Chapter 6: Suprasegmentals
0. Introduction.............................................................................. 237
1. Stress-accentsystems........................................................... 238
1.1. Soule (Zuberoa)...................................................... 238
1.2. Baztan..................................................................... 242
1.3. Bortzerrleta (Cinco V illa s )................................... 243
2. Pitch-accent systems.............................................................. 244
2.1. Gernika....................... 245
2.1.1. Raising and lowering........................... 260
 VIII

2.1.2. A note on historical change .............. 272

2.2. Ondarroa.................................................................. 275
2.3. Tone rules and lexical phonology....................... 286
2.4. Altzaga (Goiherri).................................................... 287
2.5. Arbizu....................................... ............................... 291
3. Discussion............................................................................... 292

References................................................................................... 299

Appendix: Map of the Basque-speaking a re a ......................... 31 0

 IX

Abstract

Basque is a morphologically rich and fairly regular language

with a high degree of dialectal fragmentation. These aspects
make Basque a good testing ground to examine the ways in which
phonological rules may interact with each other and with
morphological processes. Adopting the framework known as
Lexical Phonology, it is argued that a number of phonological
rules of Basque require a bi-stratal conception of the phonology.
It is also argued that clitic groups, formed in the syntax, are
sent back to stratum I and undergo the phonological rules of
this stratum. There is no evidence for cyclic application of rules
but, at least in one dialect, there is evidence that a
dem onstrably noncyclic rule is blocked from applying in
underived environments. It is thus shown that cyclic application
and restriction to derived environments are not properties of
rules that need always go hand-in-hand.
The study explores the kinds of morphological information
to which phonological rules may need to have access, and it is
argued that certain rules are sensitive to a head-of-word / non-
head distinction in the morphological structure of words.
Different ways in which the underlying or derived structure
of segments may affect the application of phonological rules are
also examined, adopting a hierarchical model of phonological
 X

representations. It is argued that assim ilation rules create

branching structures which can then be utilized in the
application of latter rules.
Special attention is given to the structure and behavior of
affricates. It is shown that, in Basque, whether affricates will
pattern together with stops or with fricatives cannot be
predicted from the position of the conditioning environment of
the rule.
Finally, the different prosodic systems found in Basque
dialects are exam ined, concentrating on the pitch-accent
system s found in western varieties. In several of these
varieties, the processes of tone assignm ent and their
interaction with other processes is studied in some detail.
 XI

There is still so much work to do

in the galaxy !
- Captain Kirk
 Chapter 1
Introduction

0. Aims and Structure

In this dissertation, I explore the interaction of phono­
logical rules in Basque whose domain of application is morpho­
logically conditioned, arguing for a particular conception of the
way phonological rules are organized and for a particular way of
representing phonological rules and segments.
With respect to the organization of the phonology, I adopt
the model known as Lexical Phonology (cf. Kiparsky (1982a;
1982b; 1984; 1985), Mohanan (1986), Archangeli (1984), Kaisse
and Shaw (1985), Halle and Mohanan (1985), Pulleyblank (1986),
Booij and Rubach (1987), among others). As for the represent­
ation of segments and processes, I argue for a Hierarchical
Model of phonological representations (cf. Clements (1985),
Archangeli and Pulleyblank (1986), Sagey (1986)) and adopt a
p aram etrical expression of phonological rules (follow ing
Archangeli and Pulleyblank (1986), Archangeli (1987)).
This dissertation is organized as follows:
In this chapter I provide some necessary background
information, both about the theoretical models applied in the
dissertation and about the sound system and the morphology of
Basque.
 In chapter 2, I show that a number of rules of Basque are
restricted to a dom ain of application com prising some
suffixation (mainly inflectional) and excluding compounding. In
this domain (stratum I), rules apply in a non-cyclic fashion (i.e.
after all m orphological operations of this stratum ). C litic
groups are shown to revert to stratum I.
In chapter 3, it is argued that the morphological notion of
head of a word can play a role in determining the domain of
application of phonological operations.
In chapter 4, I examine a group of interrelated lexical and
postlexical rules and argue for the hierarchical representation
of phonological features, giving evidence for the representation
of assimilatory processes as creating branching structures.
In chapter 5, the behavior of affricates is examined with
respect to a number of rules of Basque. It is argued that
affricates contain two contradictory specifications for the
feature [continuant] which underlyingly are unordered.
Finally, in chapter 6, the prosody of Basque dialects is
examined, concentrating on the analysis of the tonal rules of
pitch-accent varieties, and showing how tonal rules interact
with other phonological and morphological rules.
1. Theoretical assumptions
1.1. Lexical Phonology
Most of the phonological rules that are studied in this
dissertation do not apply every tim e that their structural
description seems to be met; but, rather, have a morphologically
limited domain of application. A sizable number of these rules
apply both morpheme-internally and across the boundaries of
inflectional suffixes.
In the framework of Lexical Phonology and Morphology,
which I adopt, morphological operations are paired with sets of
phonological rules. The morphophonology of a language is thus
assumed to contain several components or strata. Each stratum
w ill consist of a number of m orphological operations and
associated phonological rules. In Basque, one of these
components, which I will refer to as stratum I, will Include the
morphological operation by means of which inflectional suffixes
are attached and a set of phonological rules which apply in this
domain. Stratum I phonological rules will, in principle, also
apply to monomorphemic forms.
Stratum I phonological rules are systematically excluded
from applying across the members of a compound in Basque. This
fact will be captured by assigning Compound Formation to a
different component of the morphophonology. Stratum II. This
stratum has a set of associated phonological rules which is
different from the set that applies at Stratum I.
Derivational suffixes present a less consistent behavior
across Basque dialects. W hereas in some dialects most
derivational affixes are treated in the same way as inflectional
affixes in terms of the phonological rules that apply to them, in
other dialects, derived forms show the same behavior as
compounds. Different derivational affixes may also be treated
differently by the Phonology. In terms of Lexical Phonology, the
attachm ent of derivational suffixes is a Stratum I in some
dialects, where Stratum I will contain the bulk of suffixation. In
other dialects, derivational suffixes are attached at Stratum II.
Derivational suffixation may also be divided between Stratum I,
for some suffixes, and Stratum II, for others.
It is assumed that phonological rules can be cyclic, if they
apply after every morphological operation in the stratum, or
noncyclic, if they apply only once in the stratum, after all
morphology has been created. Perhaps being cyclic or noncyclic
is a property of strata and not of rules (Halle and Mohanan
(1985)). Basque does not present any evidence for cyclical rule
application. There is clear evidence, on the other hand, that some
stratum I rules apply noncyclically, after all stratum I
morphology has been added.
There are phonological rules whose domain of application
is not morphologically limited, applying across-the-board. These
rules apply in the Postlexical Component of the Phonology. Rules
that apply in the postlexical component may also have a previous
lexical application (Mohanan (1986)). Evidence for this repeated
application of rules is given in chapter 4.
In Basque, clitic groups undergo stratum I phonological
rules. The assumption will be that when clitic groups are formed
by word concatenation in the syntax, these groups are sent back
to Stratum I of the lexical phonology (of. Pranka (1983),
Pulleyblank and Akinlabi (1988)). The picture that emerges is
thus as in (1) (cf. Kiparsky (1982b; 1984)):

(1) Morphology Phonology

Stratum I: Underived forms
Suffixation — > rules 1, 2, 3 <

Stratum II: Compounding — > rules 3, 4

Postlexical Syntax
Component: c litic iz a tio n ........................... .
rules 5, 6

1.2. Hierarchical representations

1.2.1. The representation of segments
I will assume that segments are not bundles of unordered
phonological features, as in the SPE model (Chomsky and Halle
(1968)); but, rather, that features are grouped in subsets and
hierarchically organized (following Clements (1985), Archangeli
and Pulleyblank (1986; to appear), Pulleyblank (1988), Sagey,
(1986), among others. Cf., also Hualde (to appear)). The general
segmental structure assumed in this thesis is shown in (2) and
(3). In (2) the subsets of features that may act as a unit for the
purposes of phonological rules are defined. The dependencies
that obtain among these sets of features are shown in (3):

(2) R = root node. Groups all features in segment.

T = tonal node. Tonal features.
L = laryngeal node. Laryngeal features.
SL = supralaryngeal node. Groups all features correspond­
ing to articulations above the larynx.
P= place node. Features that define the point of
articulation of the segment.

I assume with Sagey (1986) that every segm ent is

characterized as involving (at least) one of three major
articulators: the lips (labial), the tip/blade of the tongue
(coronal) or the body of the tongue (dorsal). These are
monovalent features. Each of these major articulator features/
nodes will dominate other place features. Features such as
[back], [high] and [low] are situated under the [dorsal] node.The
feature [round] depends on [labial] and other features may hang
on [coronal].
Segments are associated to skeletal slots (X in diagram
below) that correspond to time units.

(3)
R
T
L (tonal features)
[voice]
(& other laryngeal
features)
SL [sonorant]
& other oral cavity manner
features )

[coronal]

[^ rs a l]
[high]
[back]

Whereas the general adequacy of the Hierarchical Model is

assumed, and evidence for it is given, many particular details of
organization will be indifferent for the processes to be studied
in this dissertation.

1.2.2. The representation of rules

P h o no log ical rules are o p e ra tio n s on h ie ra rch ica l
representations. Following Archangeli and Pulleyblank (1986),
 8

Archangeli (1987), I adopt a formulation of rules in terms of

conditions and parameters. The form ulation of a rule must
contain the following information:
a/ Type of operation. Four operations are recognized:
spread, delink, insert and delete. A rule may affect one feature
or more than one feature; but in this second case, the group of
features affected will correspond to a node in the hierarchy.
That is, a spreading rule R may produce the effect of spreading
features F and G only if there is a node N that dominates F and G
and only these features. The generalization is thus that a
phonological rule will affect only one feature or node.
b/Argument of the rule. This is the element (feature or
node) that is spread, delinked, inserted or deleted.
c/ Direction (in spreading rules). We must indicate if
spreading takes place from right to left or from left to right.
d/ Target conditions. These define the set of segments that
undergo the rule.
e/ Trigger conditions. These define the environment of the
rule. Rules may also contain conditions that must be satisfied by
both trigger and target; for, instance that they share structure
or features.
The parametric expression of rules has the advantage over
their "graphic" expression that only those elements that are
crucial for the rule are mentioned. Graphic representations, on
the other hand, will often include hierarchical structure that is
not made use of by the rule in question but that, for instance,
intervenes betw een relevant nodes (cf. A rchangeli and
Pulleyblank (1986)).

1.2.2. Underspecification
In this dissertation, I adopt a restrictive view of under­
specification. I will assume that every segment is in principle
specified for all features, except for redundant or meaningless
features (Steriade 1987). For instance, the feature [distributed]
only plays a role in Basque to distinguish different sibilants
from each other. I will assume that nonsibilants do not bear a
value for this feature.
When a feature value is always determ ined by the
environment in such a way that it would be arbitrary to assign a
value or another to the underlying segment, I will assume that
that value is left unspecified in the underlying representations.
For instance, the voiced obstruents /b/, /d/, /g/ present both
continuant and non-continuant allophones, depending on what
precedes them (see chapter 4). There are no convincing
arguments for specifying these segments either as underlyingly
[+cont] or underlyingly [-cent]. The value for the feature
co n tin u a n t w ill be le ft u n sp e cifie d in the u nd erlying
representations of these segments. To give another example, the
place features of morpheme-internal nasals in a coda will be
 10

left unspecified underlyingly, since the surface specification

w ill necessarily be obtained by spreading from the following
consonant (a nasal always agrees in point of articulation with a
following consonant); e.g.: /kaNpo/ [kampo] 'outside,' /kaNta/
[kajita] 'song,' /aNka/ [agka] 'leg.'
In particular, I will not adopt a radical or maximalistic
view of underspecification as expressed in Archangeli (1984),
Archangeli and Pulleyblank (1986), Pulleyblank (1988) among
others. Pulleyblank (1988) argues that in a vocalic system there
could be an unmarked vowel which w ill com pletely lack
underlying feature specifications. This vowel would by [i] in
Yoruba, according to Pulleyblank and could be a different vowel
in other languages. In the Basque dialect of Arbizu, and perhaps
in all Basque dialects, the unmarked vowel is clearly [e]. This is,
for instance, the vowel that productively appears in epenthesis
before otherwise word-initial [r], which is generally disallowed
in Basque. Thus, Spanish rueda 'wheel,' ram a 'branch,' rosario
'rosary,' etc. are adapted in Arbizu as [e rw e d a ], [ e r a m a ],
[ero^aayjo].
Consider now the Arbizu genitive plural and genitive
indefinite forms in (4) (see chapter 2, sections 3.2 and 3.3):

(4) stem gen pi gen indef

/m e n d i/ [m endijen] [m endiin] 'm ountain'
 11

/eéku/ [eékuben] [edkuun] 'hand'

/aéto/ [aëtuen] [a^toon] 'donkey'
/a l a b a / [alaben] [alabaan] 'daughter'
/tS a k u r / [tSakuren] [tSakuren] 'dog'

My Interpretation of the indefinite genitive forms is that

they are produced by means of a /-Vn/ suffix, whose vowel is
totally unspecified underlyingly. The feature values of this
vowel will be filled in by spreading from a vowel to its left. The
result of the spreading rule will be a long vowel. If no vowel is
available to the left of the empty V, default values (those of [e])
will be inserted, as in the last of the examples in (1).
The genitive plural, on the other hand, is /-en/. Now, if,
given the fact that [e] is the unmarked vowel, we were to leave
all surface instances of this vowel featureless in underlying
s p e cific a tio n , fo llo w in g the p ostu lates of radical under­
specification, the representation of the genitive plural would
also be /-Vn/. The prediction would be that genitive indefinite
and plural forms would be identical in all cases; which, as the
examples in (1) show, is not the case.
Contrary to the assumptions of radical underspecification,
I will assume that none of the five vowels of Basque is
completely unspecified in general. This is compatible with the
fact that one of the five vowels, [e], behaves as the unmarked
 12

vowel. Some particular morphemes may contain vowels which

are empty of vocalic feature specifications. This will be in
cases, like the one just examined, where surface values are
provided by rule. If the empty vowel cannot receive its features
by rule in some specific case (as in /t^ a k u r-V n /), unm arked
values will be inserted.

1.2.3. Locality
I assume that all phonological processes are strictly local;
that is to say, that the trigger and the target of a rule must be
adjacent to each other (cf. Archangeli and Pulleyblank (1986)).
Now, by adjacent I mean that no element that is relevant for a
particula r process intervenes between trigg er and target.
Irrelevant elements in a rule of feature spreading are segments
that cannot bear the feature in question contrastively. A rule
spreading a feature w ill be local if it does not skip any
segments that could bear the feature that is spread by the rule
(cf. Hayes (1988)).
Several processes studied in this dissertation seem to
violate locality. I will claim that In every case locality is
respected once we look at the process more closely.
Many Basque dialects possess a rule that changes /a/ into
[e]after a high vowel or glide (see chapter 2). Consonants (but
not vowels) may intervening between trigger and target. This is
still a local process, under the assumption that segments that
 13

cannot bear a specification for a given feature cannot block the

spreading of this feature across them. This rule of Vowel
Assimilation spreads the feature [-low]. Consonants cannot bear
a value for the feature [low] in Basque. The prediction is that
they will be skipped in searching for an adjacent target in the
spreading of this feature. No vowels can intervene between
trigger and target, on the other hand, since vowels can bear the
feature [low].
Let us consider another case. In chapter 5, it is shown that
all sibilants in a morpheme must agree in point of articulation
in Basque. This means that the value for the feature [distrib]
of all sibilants in a morpheme, except for one, can be left
unspecified. Uspecified sibilants will receive their feature value
by spreading from the sibilant bearing the specification for the
morpheme. Non-sibilants will not block the spreading beacuse
they are not feature bearing units for [distrib].
The issue of locality is also raised with respect to the
behavior of affricates (cf. chapter 5). A cluster sim plification
process that suppresses the first occlusion in a sequence such
as /tp/, giving [p], also applies to an affricate-stop sequence
such as /tsp/, giving [sp]. The rule might seem to have a non­
local application in this second case, since it operates across
the continuant branch of the affricate. This problem disappears
under the assumption that the articulations of affricates are
 14

underlyingly unordered. The ordering of articulations, which is

predictable: first occlusion, then fricative release, can be
considered a surface phenomenon.

2. Background information on Basque

In this section, after a brief description of the geo­
linguistics of Basque, I present the consonantal and vocalic
inventories of this language as well as a basic description of its
morphological and syntactic structures. The purpose is to offer
some necessary background for a better understanding of the
operations of the phonological processes studied in this
d is s e rta tio n .

2.1. Geolinguistics
Basque is a language isolate spoken in a small area of
northern Spain and southern France. In Spain, it is spoken in
parts of the Autonomous Basque Community, which comprises
the provinces of Biscay (Sp. Vizcaya, B. Bizkaia), Guipuscoa (Sp.
Guipuzcoa, B. Gipuzkoa) and Alava (B. Araba), as well as in parts
of Navarre (Sp. Navarra, B. Nafarroa). In France, the Basque
speaking area corresponds to about half of the Department des
Pyrenees A tlantiques, com prising the historical regions of
Labourd (B. Lapurdi), Basse Navarre (part of the former Kingdom
of Navarre) and Soule (B. Zuberoa). The other half of the
 15

Departm ent is occupied by the traditionally Gascon-speaking

region of Beam, All in all, approximately 700,000 people speak
Basque. The vast majority of Basque speakers are bilingual in
Spanish or French.
Since Basque is a language of considerable dialectal
d ive rs ity , the d ivision of the B asque-speaking area in
geographical dialects and subdialects has been a traditional
concern of researchers on the Basque language (of. Bonaparte
(1864; 1873), Azkue (1923), Yrizar (1981), Etxebarria (1983),
T xillardegi (1983), among many others). The fact is that
virtually every town or village speaks its own variety. It is
equally true, however, that there are phonological processes
that are found in all Basque dialects or throughout large areas of
the Basque speaking domain. The comparative study of Basque
d ia lects thus lends its e lf extrem ely w ell to testing the
different ways in which a given phonological rule may interact
with other rules of the phonology and with the morphology of
d ifferent linguistic systems.
At different historical points, written koines based on the
dialects of some particular area have emerged. Some of them,
such as Literary Navarro-Labourdin (on which Lafitte’s (1979)
Grammaire Basque is based), and Literary Biscayan, still enjoy a
certain popularity and may be occasionally used as oral
languages in preaching, etc. A much more powerful influence on
dialectal diversity is being exerted by the recently encoded
 16

Standard Basque or Euskara Batua. This variety is not only used

in the vast majority of all written production in Basque, but also
is used virtually exclusively in all education in Basque language
and in the media. There is thus nowadays a sizeable group of
fluent speakers of Standard Basque. Whereas some Basque-
educated speakers w ill keep the standard variety and their
native code as two different languages used on separate
occasions, and will allow little mixing of the two, some other
speakers, especially of varieties morphologically very close to
the standard, will show different degrees of mixing.
In this dissertation, Standard Basque will be used in the
description and analysis of those phonological processes that
are reflected in the orthography (such as the processes involved
in Stem Alternation; see chapter 3) or that are otherwise
general to the language (such as Nasal and Lateral Assimilation;
see chapter 5). There are other rules that are instantiated in
different ways in different dialects. Some processes appear to
have different phonological conditions in different dialects; in
some cases, the order of phonological rules varies across
dialects; the domain of application of a given rule can also vary
among dialects. For the study of rules presenting such
differe nce s in application across-dialects, such as Vowel
Assim ilation (see chapter 2) or Palatalization (see chapters 2
and 4), a number of what I believe to be representative local
 17

varieties have been selected. In each case, I discuss the

theoretical interest of dialectal differences in rule application.
The map in the addendum shows the location of every
variety mentioned in this dissertation.

2.2. The sound inventory

A common consonant inventory in Basque is as in (5) (for
detailed inventories of a large number of Basque dialects, see
Moutard (1975a; 1975b; 1976)):

(5) Consonant chart

labial dental/alveolar palatal velar
p t f k

b d j g

f s é Ï X
ts ié
m n n
1 /C

r
r

The sibilant fricative /s/ and affricate /ts / differ from

the /é/, /té/ fricative-affricate pair in that /s/ and /ts/ are
 18

d o rs o -a lv e o la r (la m in ai) w hereas H J and J\iJ are apico-

alveolar. The distinction between these two articulations is not
made in any of the dialects spoken in Biscay and has also been
lost in parts of Guipuscoa. In transcribing forms from dialects
that do not distinguish these two series, the symbols /s/ and
/ts/ will be used. Phonetically the realizations are generally
apico-alveolar [é] and dorso-alveolar [ts] in these dialects. In
the orthography, §_ = apico-alveolar fricative [é]\ t ^ = apico-
alveolar affricate [td]; z. = dorso-alveolar fricative [s]; tz =
dorso-alveolar affricate [ts]; x_ = palatal fricative [§]; Ix. =
palatal affricate [tS].
The obstruents /t/ and /d / are dental, w hereas the
sonorants /I, n, r, r / are alveolar.
The voiced obstruents present both stop [b, d, g] and
fricative allophones [b, é, ÿ]. The allophonic distribution will be
studied in chapter 4.
The sound that I note as /j/ is a voiced palatal, which
often has both stop and fricative realizations, in the same
contexts as /d/. The status and precise phonetic realizations of
this sound vary a great deal across dialects. Its allophones
generally merge with the palatal realizations of /d/. In some
dialects it may also be an allophonic variant of /I/ syllable-
initia lly. O rthographic j. represents very different sounds in
d iffe re n t dialects. A word such as ja n 'to eat' will be
 19

pronounced [jan], [yan], [2an], [xan], among other possible

pronunciations, depending on the speaker's dialect (cf. Hualde
(1987) for the historical origin of this diversity).
The two rhotic segm ents, flap /r/ and trill /r/, only
contrast intervocalically, as in Spanish: e re [ere] 'also,' e rre
[e re ] 'to burn.' Elsewhere the distinction is neutralized. Unlike
Spanish, in pre- or postconsonantal position the realization is a
trilled [r] in many Basque dialects arto [arto] 'corn,' andre [andre]
'woman.' In the spelling, the double grapheme rr indicates the
trill [r] between vowels; r is a flap [r] intervocalically.
Some of the dialects spoken in France possess a phoneme
/h/ and also aspirated stops. No dialect has both /h/ and /x/. The
grapheme | i , which is used in the standard language following
the usage in the northern dialects, is a silent character for most
speakers.
Some Biscayan dialects have a voiced coronal affricate
/dz/ in a reduced number of words.
Souletin has a voiced fricative phoneme /z/ which
contrasts with /s/ in intervocalic position. In all other dialects,
the fricatives /s/ and H I present voiced allophones before
voiced consonants and are voiceless elsewhere.
In some central and western dialects, there is no /f/. In
borrowings, this sound is replaced by /p/: [p ra n t^ e ^ ] (Sp.
fra n cé s ) 'French,' [pernando] 'Fernando.' In other dialects, the
voiceless labiodental fricative is found mainly in borrowings,
 20

but also in some native items, where it may alternate with [p],
such as a ta ri, apari 'dinner' a lte r, a lper 'lazy.' [f] may also occur
for French [v]: fitg (Fr. v ite ) 'quickly.'
Most Basque dialects have a system of five vowels /i, e, a,
o, u/. The dialect of Soule also possesses /ü/, as well as
contrastively nasalized vowels. There are no underlying glides.
The high vowels /u/ and /i/ are realized as the glides [w] and [y]
in postvocalic position. E.g. : haundi /aundi/[awndi] 'big' b a im e n
/baim en/ [baymen] 'perm it.' In eastern dialects, underlying or
derived high vowels also glide before another vowel; e.g.: negua
/negu-a/ [negwa] 'the winter,' etxea /et^e-a/ [etSya] 'the house.'
In western dialects, high vowels do not lose their syllabicity
before another vowel: [nefu.a], [etSi.a].
In the orthography, the double characters li, tt, and d d
indicate palatal versions of 1,1, and d. respectively; that is, [x ],
[f] and [j]. E.g.: T x illa rd e g i [tS ix a r é e g i] 'a name,' ttik i [t'iki]
'sm all,' M a d d a le n [majalen] 'a name' (with affective palatal­
ization of /d/, of. chapter 4, section 3.4), o n d d o [onjo] 'mush­
room.'

2.3. Basic phonotactic constraints

2.3.1. W ord-initial restrictions
W ord-initially, the two rhotics, [r], [ r ] do not occur in
native words. Assimilated borrowings from Spanish with initial
 21

[r] (orthographie r-) in the lender language undergo epenthesis,

generally of [e-], although there is some variation, e.g.:
errepublika (Sp. republican 'republic,' errespuesta (Sp. respuesta)
'answer,' e rro m e ria (Sp. ro m e ria ) 'pilgrimage,' e rre s p e to (Sp.
respeton 'respect,' a rra to i (Sp. raton 'mouse') 'rat,' a rro p a (Sp.
ropa) 'clothes,' arraza (Sp. raza) 'race.'
The a ffricates is and lz_ are rare w ord-initially in all
dialects. The apico-alveolar [té] is not found word-initially in
any native items. The dorso-alveolar affricate Iz [ts] occurs in
the adjective tz a r 'huge, evil, ' and derivatives and very few
o th e r ite m s; g e n e ra lly w ith a u g m e n ta tiv e / d e s p e c tiv e
connotations, e.g. tzakur ' big dog,' tx a k u r. zakur 'dog.' The other
affricate, the palatal Ix, is also rare in this position in many
eastern dialects, but not in Biscayan and Guipuscoan, where
historically word initial [S] x, was strengthened to [tS] tx. E.g.
Labourdin x o r i. G uipuscoan/B iscayan tx o ri 'bird.' In these
w estern dialects, w ord-initial x_ has been preserved only in
dim inutives (e.g. x e x e n . diminutive of z e z e n 'bull') and in
personal names. Only in Souletin do both x and b i occur freely in
w o rd -in itia l position.
W ord-initially, the palatal lateral [x] and nasal [h] are also
rare or not found depending on the dialect, if we leave aside
diminutive and affective words (of. chapter 4, section 3.4).
 22

2.3.2. Syllable and word-final restrictions

Only coronal segments do freely occur word (and syllable)
finally. Dialects differ on whether the palatals [X], [h], [f] do or
do not occur word-finally. When they occur, they are the result
of a palatalization rule (see chapter 4), e.g.: /m util/ [m utix]
'boy,' /egin/ [eyh] 'to do, make,* /d-a-ki-t/ [dakif] 'I know.' The
palatal stop [f] may also occur in affective or onomatopoetic
words; e.g.: [pot'] 'kaput' (in Arbizu). The coronals [t, é, s, t^, ts,
I, n, r ] are frequent word-finally: b a t [bat] 'one,' ikas [ika^] 'to
learn,' m aiz [mays] 'often,' hots [ot^] 'sound,' latz [lats] 'rough,'
azal [asal] 'skin,' Ian [Ian] 'work,' enbor [embor] 'trunk.'
As mentioned before, there is syllable-final neutralization
of the two rhotics. The stem-final position is, nonetheless, one
of potential contrast for the rothics, since the vowel of a suffix
may fo llo w , thus placing the rhotic in the co ntra stive
intervocalic environm ent. The vast m ajority of items with a
stem -final rhotic have a phonological trill, which shows up
before a vowel, e.g.: /u r / u r /u r r a 'h a z e ln u t (u n in fle c te d /
absolutive singular),' /e d e r/ e d e r/e d e rra 'beautiful.' There are
probably less than ten items with an underlyingly final flap in
the standard dialect, and fewer in most dialects. The exhaustive
list of flap-final items that I am aware of includes only three
native nouns (/ur/ u r/u ra 'water,' /or/ o r/o ra 'dog' and /sur/
zu r/zu ra 'wood'), a few borrowings such as p la te r/p la te ra 'plate'
and p u ta n e r/p u ta n e ra 'that consorts with prostitutes,' and a
 23

handful of short-list forms. These last forms include the two

questlon-words /nor/ n or/n oren 'who (absolutive/ genitive)' and
/serf z e r/z e re n 'what (absolutive/ genitive),' an allomorph of
the distal dem onstrative, /(h)ar/ h a rk /h a re n 'that over there
(ergative/genitive),' and two numerals which in the standard and
in western dialects do not present a final rhotic in their
uninflected form, but whose underlying rhotic appears when an
inflectional or derivational suffix follows: /(h)irur/ hiru 'three,'
h iru ra k 'the three (absolutive);' /laur/ la u r 'four,' la u ra k 'the
four (absolutive),' laurden 'quart.'
The palatal fricative x_ [ Ï ] and affricate tx [t$] are rare
word-finally, but are found in a few examples: q Ix 'piss,' a ra tx
'there (it is).'
The voiced obstruent /d/ [d], [b] does not occur word-
finally. In this position there is no contrast between voiced and
voiceless segments.
Velar and labial consonants are not found w ord-finally
with one exception. The exception is the velar voiceless stop fkf
(often realized as [§] in this position), which is found in three
frequent affixes: ergative /-k/, e.g.: k o ld o - k 'Koldo (erg),'
absolutive plural /-ak/, e.g.: m u til-a k 'the boys,' and second
person fam iliar masculine /-k/, e.g.: d -a -u k a -k 'you (man) have
it.' Stem-finally /k/ does not occur. The other velar and labial
consonants are totally excluded from the word-final position.
 24

2.3.3. Syllable-internal restrictions

The only consonant clusters allowed in an onset are those
formed by a stop or /f/ and a liquid. E.g.: prakak 'trousers,' andre
'woman,' g rin a 'worry, desire,' fruitu 'fruit,' z a p la d a 'blow.' Even
these groups seem to have been disallowed in earlier stages of
the language and are still relatively rare. Old borrowings from
Latin and Romance show consistent sim plification of these
clusters, either by epenthesis as in liburu < libru- 'book,' k e le ta
< d a ta 'gate,' a u ru tze < c ru c e - 'cross,' etc. or by deletion of the
obstruent as in lum a < plum a 'feather,' lore < flo re 'flower,' loria
< gloria 'glory,' laket < placet 'pleasure,' etc.
In a coda, the clusters in (6) are found:

(6) f r i c - / t / p re s i tx o rro x l
'ready' 'an interjection'
s o n -/t/ b e ra n i b a it LaxaJl
'late' 'last night' 'a name'
s o n - fr ic mendiranzko hereto beiztu
'bound for 'clo su re ' 'blacken
the mountain'
son-affr antz b oriz b e ilz
'resemblance' 'five' 'bla ck'
 25

B ilintx b e ltx
'a name' 'b la c k is h '
sonVk/ nork
'who (erg)'

These are all groups of two coronals, with the exception of

the group /-rk/, which arises from the affixation of /-k/ to
rhotic-final stems. This group is disallowed in some dialects.
Where the group would arise by morpheme concatenation, it is
sim p lifie d either by deletion of the rhotic or by vowel
epenthesis: /nor-k/ [nok] 'who (erg),' /ser-k/ [sek] 'what (erg),'
/d -a -k a r-k / [dakarek] 'thou (man) bringest.'
In many dialects, including the standard, the distinction
between fric a tiv e s and a ffric a te s is n eu tralized a fte r a
sonorant consonant. In this position, the realization is an
affricate, unless another consonant follows, in which case it
will be a fricative. This is why the examples of sonorant plus
fric a tiv e clu ste rs in (6) are not w o rd -fin a l; a follow ing
consonant is necessary for the group sonorant-fricative to
appear in a coda. The alternation between affricates and
fricative s after a sonorant depending on w hether another
consonant follows can be observed in examples such as /mendi-
ranZ/ m e n d ira n tz 'towards the mountain' vs. /m endi-ranZ-ko/
m e n d ir a n z k o 'bound for the mountain' (lit. 'of towards the
 26

mountain'). The process of fricativization of affricates before a

stop will be studied in chapter 5, section 1, The neutralization
of fricatives and affricates after a sonorant is independent of
syllabification. It is also found in heterosyllabic groups, as in
the follow ing exam ples, which show affrication of fricatives
after a sonorant in borrowings, where the syllable boundary
occurs between the sonorant and the affricate: b e rts o 'verse,'
tra n ts itib o 'transitive,' z ie n tz ia 'science.'

2.4. The structure of words and phrases

A large number of phonological rules are morphologically
conditioned in Basque. Quite a few processes apply within a
morpheme and also between stem and inflectional suffix, but not
between the stem and a derivational suffix or across the
members of a compound. Different types of compounding also
trigger the application of different phonological rules. Before
studying the phonological processes of Basque we must thus give
some notion of its morphology.
In subsection 2.4.1. I summarize the different types of
d e riva tio n and co m p ositio n in Basque. The in fle c tio n a l
m orphology, which is phrasal in its scope, is presented in
subsection 2.4.2. where the basic structure of noun phrases is
also introduced. Finally, the section on morphology is completed
with an overview of the morphology of verbal forms.
 27

2.4.1. Word formation

The two main means of word form ation in Basque are
derivation by suffixation and compounding (of. Lafitte (1979),
Villasante (1976), Mujika (1978)).
The Basque language possesses a large number of
derivational suffixes, many of them borrowed from Romance. A
few of them will appear in examples in the next sections.
Prefixes, on the other hand, are rare. Perhaps the only prefixes
that have some productivity are b e r-/b ir- 're-,' which attaches
to non-conjugated verbal forms, e.g. berr-eqin 'to do again,' and
the negative e z -/d e s - 'un-, dis-.' e.g. e z - b e r d in /d e s - b e r d in
'unequal, uneven,' des-egin 'undo.'
S everal typ es of co m p ositio n are ve ry p ro du ctive
processes. The difference between composition and derivation is
that in composition, both elements in the word can also be used
as free-standing words. Three important types of composition
are cocompounding (dvandva) (I.e. if the two members are X and
Y, then X-Y = X and Y), e.g. : sem e-alabak (sem e= 'son,' alaba =
'd a u g h te r,' ak. = p lu ral a rticle ) 'sons and d a u g h te rs,'
subcom pounding (tatpurusa), where the head is the second
member of the compound (i.e. X-Y = type of Y), e.g. : g a u -tx o ri
(g a u = 'nig ht,' tx o r i= 'b ird ') 'n ig h t-b ird ' and e x o c e n tric
composition (i.e. X-Y = having an X which is Y), e.g. : b u ru -x u rj
(b uru = 'head,' xuri = 'white') 'white-headed.'
 28

As indicated above, the most intimate case of morpheme-

concatenation, with respect to the phonology, takes place in the
inflectional morphology. In Basque, inflection is phrasal in its
scope: only the leftm ost constituent in the noun phrase is
inflected. In the next subsection the inflectional morphology of
Basque is presented in connection with the structure of the noun
phrase.

2.4.2. Structure of the Noun Phrase^

Within the Noun Phrase, the order of elements is noun-
adjective-determ iner. The determ iner position can be occupied
by demonstratives, as in (7a), the definite article, (7b), the
numeral b a t 'one' which also functions as an indefinite article
(plural b a tz u fk l 'some'), (7c) and some quantifiers. In Biscayan
dialects the numeral b i 'two' also occupies the determ iner
position. All other numerals (7d), some quantifiers (7e), usually
possessives (7f), b e ste 'other' and the adjectival specifier o s o
'very' (7g) appear in pronominal position. With some quantifiers
there is dialectal variation in their placement (7h). In Biscayan
dialects demonstratives appear both before the noun and in the
determ iner position (7i):

1 For a more detailed description than what we can offer here, see Goenaga
(1978), Euskaltzaindia (1985). Basque syntax is also studied in detail in Ortiz
de Urbina (1986), Eguzkitza (1986) and Saltarelli (1988).
 29

(7) a. etxe hauek

house these
'these houses'
b. etxe zuri-a
house white-the
'the white house'
0. etxe bat
house one
'one/a house'
d. hiru etxe zuri
three house white
'three white houses'
e. edozein etxe
any house
'any house'
f. gure hiru etxe zuri hauek
our three house white these
'these three white houses of ours'
g. oso etxe zuri-a
very house white-the
'a very white house'
h. anitz etxe / etxe anitz
many house house many
'many houses'
 30

i. Biscayan; au etxe zuri au (Standard: etxe zuri hau )

this house white this
'this white house'

Notice that the definite article is the only determ iner

that is orthographically attached to a host. Other determiners
such as the num eral/article bat 'one' or the dem onstratives
appear as independent words. There are syntactic reasons for
this convention. The definite article behaves like a suffix, in
m ost dia lects, in that it cannot appear by itself. O ther
determiners, on the other hand, can be the only members of their
NPs. This is shown in (8):

(8) hau dator 'this (one) is coming'

bat dator 'one is coming'
*a dator^ 'the (one) is coming'

From a p ho no log ical point of view , how ever, the

num eral/indefinite article and the demonstratives behave as if
they were suffixed to a word to their left within the noun
phrase, undergoing all sorts of rules that are otherwise blocked
across word-boundaries. Consider the examples in (9), which
show the parallel behavior of the definite article and the

2 This is im possible, except for Biscayan dialects w here a. is also a

demonstrative, as it was originally in all dialects.
 31

dem onstratives with respect to a rule that raises /a/ to [e]

after a high vowel in the Baztan dialect (9a) (cf. chapter 2) and
a rule that inserts [§] between /i/ and another vowel in the
Ondarroa dialect (9b) (cf. Chapter 2 ). In both cases it is shown
that the rule is blocked between noun and adjective:

(9)a. /buru-a/ [burue] 'the head'

/buru au/ [buruew] 'this head'
/buru aundi/ [buruawndi] 'big head' *[buruewndi]
b./mendi-a/ [mendiée] 'the mountain'
/au mendi au/ [awmendiSaw] 'this mountain'^
/mendi aundi/ [mendiawndi] 'big mountain' *[mendiSawndi]

Since dem onstratives and the indefinite article can also

function as independent words, as shown in (8), we will speak of
cliticization in this case.
All examples so far have been given in the absolutive case,
which is pho no log ically unm arked. Case m arkers occupy
otherwise the rightmost position within the Noun Phrase. Unlike
languages such as Latin or German, Case is marked only once in
the Noun Phrase, as shown in (10):

3 For the a p p a re n tly d issim ilar b eh avio r of d e fin ite a rtic le and
demonstrative with respect to the rule raising /a / to [e] in the Ondarroa
examples, see section 2.3.
 32

(10) lib u r u - a - r i
book-sg-dat 'to the book'
liburu berri-a-ri 'to the new book'
liburu berri hor-ri 'to that new book

Also unlike those languages, nouns do not belong to

different declensions. There is only one set of case suffixes and
surface differences and the phonological shape that a stem will
take when marked for a given shape is always predictable from
purely phonological considerations, with the exception of the
demonstratives, where there is a certain amount of suppletion.
The complete set of declensional forms is shown in (11) for the
noun m endi 'mountain' in the standard language (adapted from
Kintana et al. (1980), see also Txillardegi (1978)). I separate the
stem from the inflectional suffixes by a hyphen:

(11) in d e fin ite s in g u la r p lu ra l

a b s o lu tiv e mendi m endi-a m endi-ak
e rg a tiv e m e n d i-k m endi-ak m endi-ek
d a tiv e m e n d i-ri m e n d i-a ri m e n d i-e i
g e n itiv e m endi-ren m endi-aren m endi-en
c o m ita tiv e m e n d i-re kin m en di-are kin m en di-ekin
b e n e fa ctive m e n d i-re n tz a t m e n d i-a re n tz a t m e n d i-e n tz a t
m o tiv a tiv e m e n d i-re n g a tik m endi-arengatik m endi-engatik
 33

in s tru m e n ta l m endi-z m endi-az m endi-ez

p a r titiv e m e n d i-rik
p ro la tiv e m e n d i-tz a t
Locative Cases
in e sive m endi-tan m endi-an m endi-etan
a b la tiv e m e n d i-ta tik m e n d i-tik m e n d i-e ta tik
a d ia tiv e m e n d i-ta ra m e n d i-ra m e n d i-e ta ra
d ire c t-a d i m e n d i-ta ra n tz m e n d i-ra n tz m e n d i-e ta ra n tz
g o a l-a d i. m e n d i-ta ra in o m en di-ra ino m e n d i-e ta ra in o
g e n itiv e m en di-ta ko m endi-ko m endi-etako

There is a clear m orphological distinction between non­

locative and locative cases because of the way number and case
are encoded. In non-locative cases, comparison of indefinite and
singular forms shows that in the singular an a. (the singular
definite article) intervenes between the stem and case markers
which are identical to those that appear in the indefinite. The
plural case morphology is more complex to segment between
number and case markers. In the absolutive, which bears no mark
for case, a plural -ak appears. In all other cases the plural
marker is In a number of cases (dative, genitive, comitative,
and cases based on the genitive), an -r- which is present in the
indefinite and singular is absent in the plural. In locative cases,
 34

absence of a number mark signals the singular, whereas the

indefinite bears -la - and the plural -e ta - .
Both non-locative and locative genitive suffixes can be
followed by other case markers, e.g. : aizon-en-ari (man-gen pl-
dat sg) 'to the one of the men;' m e n d i-k o -a re k in (mountain-gen
loc sg- com sg) 'with the one of the mountain.' The genitive-
locative ko. can also be added to other locative markers, e.g.
m e n d i-e ta ra -k o -e k (mountain-adi pl-gen loc-erg pi) 'the ones
bound for the mountains.'
In many dialects, especially Guipuscoan and Biscayan,
number and case markers are fused to a greater degree than in
Standard Basque. Also singular/plural form s are segm entally
identical for a number of cases (ergative, genitive, dative). In
these dialects, however, singular and plural forms often show
different suprasegm ental (tonal) patterns, as in the Ondarroa
examples in (12) (of. Chapter 6):

(12) alkatiai 'to the mayor' vs. a ik a tia i 'to the mayors'

gixonak 'the man (erg)' vs. gixonak 'the men

(abs/erg)'

txakurran 'of the dog' vs. tx a k u rran 'of the dogs'

 35

2.4.3. The verbal Complex

Verbal form s can be eithe r syn the tic or a n a lytica l.
Nowadays, only a handful of verbs possess synthetic forms, and
then only for a few tenses.^ These form s contain, besides
tense, agreement markers with ergative, absolutive and dative
arguments. An example is given in (13):

(13) d-a-kar-gu 'we are bringing it'

d- = 3 sg abs
-a- = [-past]
-kar- = verbal root 'bring'
-gu = 1 pi erg

Analytical constructions are formed with one of four non­

co n ju g a te d form s of the main verb m arked for aspect
(perfective, imperfective, future, or root form) and a transitive
or intransitive auxiliary which carries tense and agreem ent
markers. These auxiliaries can also be used as main verbs with
the meanings of 'to be' (intransitive) or 'to have' (transitive).
Most verbs have a suffix /-tu/ ([du] after nasal or lateral)
in their perfective form. Smaller groups of verbs have an d or -n

4 Synthetic forms w ere far more numerous in past centuries, in the

sixteenth century, over sixty verbs possessed synthetic forms (cf. Lafon
(1980)). Nowadays most dialects possess only from five to ten verbs that can
be synthetically conjugated.
 36

ending and a handful of other verbs do not have a morphologically

identifiable ending (e.g: io_ ’to hit,’ h ji ’to die/kill'). In the
imperfective aspect the perfective ending is replaced by -tze n
(su b je ct to c e rta in p ho no log ical re s tric tio n s )^ except for
verbs ending in /-n/ in the perfective aspect, which take -te n
in s te a d .® The future suffix is -ko (identical to the genitive
locative) for all verbs, in Guipuscoan and Biscayan dialects. In
eastern dialects (Souletin) a suffix - (r)e n which bears formal
resem blance with the non-locative genitive m arker is used
instead. In some Navarrese dialects -ko is used with most
verbs, but those ending in -a take -en . The Standard dialect
allows both p ossib ilities for - a verbs. The future suffix is
always added to perfective forms. Finally there is a root form
derived from the perfective form minus the suffix -la or -i and
identical to the perfective for verbs ending in -a and others.
Examples of the four aspectual forms of some verbs are given in
(14) for Standard Basque:

^ The ending /-i/ is not lost in the imperfective, if the resulting consonant
cluster would violate phonological constraints, e.o. ireki (p e rf.)/ irekitzen
(imperf.) 'to open,' not *ire k tze n . If the last consonant of the root is a sibilant,
/-ts e n / is simplified to /-ten /; e.g .Jkasi (p erf.)/ ikasten (imperf.) 'to learn,'
not * ikastzen: poztu (perf.)/ pozten 'to be happy' not * poztzen. See chapter
5, section 4.
® In some Biscayan dialects, the imperfective suffix is /-te n / for all verbs,
and this ending can be added to forms carrying the perfective m arker (i.e.
perfective markers are taken as part of the stem) e.g. a rtu te n vs. Standard
ha r,tze.Q.
 37

(14) perfective imperfective future ro o t

hartu hartzen hartuko har 'to take'
ibili ibiltzen ibiliko ibil 'to walk'

jo jotzen joko jo 'to hit'

eman ematen emango/ eman 'to give'
emanen

These forms combine with the auxiliaries to give the different

tenses, as the examples in (15) show;

(15) hartu dut ' 1 have taken it' hartu nuen '1 took it'
hartzen dut '1 take it' hartzen nuen ' 1 used to take it'
hartuko dut ' 1 will take it' hartuko nuen ' 1 would take it'
har dezadan ' that I take it (subjunctive)'

Only a restricted set of elements may intervene between

main verb and auxiliary in affirm ative clauses (cf. Hualde and
Ortiz de Urbina, to appear), but inflected form s in negative
clauses are preposed together with the negative particle &z (e.g
ez dut liburua hartu 'I have not taken the book.' Most
Com plementizers appear attached to the inflected verbal form
as Suffixes (e.g. -lako 'because,' hartu dud ala ko 'because I have
taken it'). A few complementizers (ba 'if,' bait 'since'), however,
 38

appear right before the inflected verbal form, and their status
as prefixes or clitics is not uncontroversial.

2.4.4. Blocking of periphrastic constructions

An interesting aspect of synthetic verbal form s is that
they block the use with these verbs of an otherwise totally
productive verbal periphrasis. The interest of this fact is that,
in Lexical Phonology (Kiparsky (1982; 1985)), there is an
expectation that the blocking of one form by the existence of
another form with the same meaning will be limited to word
form ation within the lexicon. The existence of blocking effects
within the lexicon is readily observable in the im possibility of
applying general morphological and phonological rules to create
a regular form when an irregular form exists. Thus, the
existence of an irregular past ta u o h t for the verb to teach
precludes the application of the general rule of past tense
formation that would produce *te a c h e d . The blocking of regular
forms by the existence of irregular forms is taken by Kiparsky
to be a consequence of the same principle that dictates that a
more specific rule has preference in its application over a more
general rule (the Elsewhere C ondition). W ithin the Lexical
Phonology model, the types of blocking effects that are possible
follow from the organization of rules in the lexicon. The more
sp ecific processes that generate irregular form s w ill be
 39

ordered before more general processes and w ill block their

application in particular cases. Phrasal constructions, which are
generated outside the lexicon, are not expected to be subject to
blocking effects. The case to be discussed immediately below
would constitute a case of exceptional blocking outside the
lexicon.^
In Basque, the progressive aspect is expressed by means of
a periphrasis with a r i. which in Hualde and Ortiz de Urbina
(1986) is argued to be a defective verb. This verbal element a ii
imposes the use of the intransitive auxiliary, even when the
main verb would otherwise require the use of the transitive
auxiliary. Consequently, in progressive constructions the subject
of the clause receives absolutive Case in every instance.
Compare the present habitual constructions in (16a) where
exam ples are given for both a transitive and an intransitive
verb, with the present progressive a ri- constructions in (16b),
where the choice of auxiliary and Case marking is determined by
the presence of a ii :

(16) a. i. Jon abiatzen da

leave-imp AUX-intr
'Jon leaves'

7 Two other exam ples of blocking outside the lexicon, in Japanese and
English, are discussed in Poser (1986).
 40

ii. Jon-ek kantatzen du

erg sing-imp AUX-trans
'Jon sings'
b. i. Jon abiatzen ari da
leave-imp AUX-intr
'Jon is leaving'
ii. Jonkantantzen ari da
sing-imp AUX-intr
'Jon is singing'

The progressive construction with â l i can in principle be

used with all verbs, except for those few verbs that possess
synthetic present (and past) forms. Verbs such as ja k in 'to
know ,' jo a n 'to go,' e k a rri 'to bring,' ib ili 'to walk,' e to r r i 'to
come' or e d u k i 'to possess, to contain,'which have synthetic
forms, cannot be used in a ri-co n stru ctio n s:

(17) daki 'he knows' *jakiten ari da

doa 'he goes' *joaten ari da
dakar 'he brings' *ekartzen ari da
d a b il 'he walks' *ibiltzen ari da
d a to r 'he comes' *etortzen ari da
dauka 'he possesses' *edukitzen ari da
 41

To make these facts com patible w ith a theory that

restricts blocking effects to word formation within the lexicon,
we may try to argue that âlL forms a compound verb with the
imperfective form to its left or that it is some sort of affix,
i.e. k a n ta tz e n -a rl would be a single word, the progressive form
of the verb 'to sing.' An argument against this position is that
a ri can be separated from the main verb in negative
co n s tru c tio n s :

(18) Jon ez da ari madrigalak kantatzen

neg AUX madrigals sing-imp
'Jon is not singing madrigals'

I believe this example to constitute solid evidence that

blocking effects may obtain between lexical forms and phrasal
co n s tru c tio n s .

3. Summary
In this chapter, I have outlined the theoretical framework
that is assumed in this dissertation and I have offered some
necessary background information on Basque. We are now ready
to study the ways in which the phonological system of Basque
interacts with the morphology of the language, which will be
done in the next chapters.
 42

Chapter 2
The Structure of the Lexicon

0. Introduction
In this chapter, I explore the interaction between the
phonological rules of Basque and d iffe re n t m orphological
o pe ra tio ns, w ith the purpose of determ ining the role of
morphology in defining the domain of application of phonological
processes.
The ch ap te r is organized around a rule of Vowel
Assim ilation whose interaction with the m orphology and with
other phonological rules is particularly clear. Three dialects are
studied in some detail: those of Baztan, Arbizu and Ondarroa.
First, I show that in Baztan the domain of application of
Vowel Assim ilation includes the m orphem e-internal context as
w ell as m orphological dom ains created by su ffixa tio n of
in fle c tio n a l and som e d e riv a tio n a l a ffix e s (s tra tu m I
morphology). Clitic groups, which are assumed to revert to the
stratum I of the lexicon, also undergo the rule. The rule of Vowel
Assimilation does not apply between the members of a compound
or with certain derivational suffixes (stratum II morphology).
It is shown that Vowel Assimilation cannot apply to the
first m orphological domain where its structural description is
 43

m et, the stem ; ra th e r the firs t a p p lic a tio n of Vow el

Assim ilation must take place after inflectional affixation.
A distinction between two m orpho-phonological strata is
also warranted by the Arbizu facts. In this dialect, however, all
derivational suffixes constitute stratum II morphology. The bi-
stratal conception of the lexicon is confirmed by other rules of
the dialect which are also restricted to stratum I: Palatalization
and Consonant Insertion.
Finally, in the dialect of Ondarroa, Vowel Assim ilation
applies to both inflection and derivation (in a non-cyclic
manner) but not to bare stems. Compounding is excluded from the
dom ain of the rule as in the other dialects. The distinction
between suffixation, on the one hand, and compounding, on the
other, is ratified by Palatalization facts. This d ia lect also
presents evidence that clitics are added at a later stage than
suffixation and clitic groups are sent back to stratum I, giving
stratum I rules a second chance to apply.

1. The Rule of Vowel Assimilation

In a number of Basque dialects, extending from Navarre to
Biscay, but quite irregularly distributed from a geographic point
of view, there is a rule that raises a low vowel to /e/ after a
high vowel, with or without any intervening consonants, e.g.:
/m util-a/ [m utile] 'the boy,’ /m endi-a/ [mendie] 'the mountain,'
/e gu n -a / [egune] 'the day.' In its m ost general form , this
 44

assim ilatory rule can be given a prelim inary linear formulation

as in (1):i

(1) Vowel Assim ilation^

V > [-low] / [+high] C q ___

This is an assim ilatory process and, hence, its correct

formulation must arguably be as a spreading rule. A low vowel
rises to mid under the influence of a preceding high vowel. In
this way the distance between both vowels is reduced. A
formulation of the rule in terms of parameters and conditions is
given in (2):

(2) Vowel Assim ilation

Operation: Spread
D irection: Rightwards
Argum ent: [-low]
Trigger Conditions: [+high], V
Target: V

A rather similar process is found in the Galician dialect of Ancares. In this

Romance dialect, stressed /a / is fronted and raised to [a] or [e] when
preceded by a high vowel or glide, as in the following examples (from Carballo
C alero (1 9 7 9 , 120) who refers the reader to Alonso and G arcia Yebra
(1 9 6 1 )): axuntâr 'to bring together,' aco uhâr 'to calm down,' c in c a r 'to stick,'
fre itad a 'land-slide.'
2 I take the vowel /a / to be [ back] in Basque, for which there is also
phonetic evidence (cf. Navarro Tom âs (1922; 192 5), Txillardegi (1 9 8 0 )).
Txillardegi (1980, 20) states that Basque /a / is normally more closed and
-palataLthan_that-Of_Gaslilian_Spanish.__________________________________________
 45

The process is formulated as the rightward spreading of

the feature [-low] from a high vowel and to another vowel. In the
concept of vowel we must here include glides, since glides will
also trigger the process. Glides are never underlyingly distinct
from vowels and can be assumed to be created by a later rule. A
graphic representation of rule (2) is given in (3):^

(3) V
R 0

SL
P
1
[c o r j^
t[cor]
[+high] [-low]

In following sections the domain of application of the rule

of Vowel Assimilation will be examined, since the interaction of
this rule with different m orphological operations allows us to
see into the organization of the Basque lexicon. Three dialects
will be studied in detail. First, the application of the rule in the
Baztan dialect, in Navarre, will be examined. This is a rather
typical dialect with respect to the m orphological dom ain of

^ The formulation of the rule In the text requires the previous assignment of
the redundant [-low] feature to [+high] vowels. In an alternative formulation,
the feature [+high] would be spread to a low vowel, the resulting [+high],
[+low] feature combination being interpreted as a mid vowel. In Hualde (in
prep, b) it is shown that such an interpretation of the incompatible features
[+high], [ low] is required in Asturian Spanish.
 46

application of Vowel A ssim ilation. Then two other dialects

which, in addition to the rule of Vowel Assimilation, have other
processes which refer to the organization of the lexicon in a
sim ilar m anner will be studied. The varieties studied will be
that of Arbizu in Navarre and that of Ondarroa in Biscay. In both
varieties, the rule of Vowel Assim ilation presents interesting
restrictions with respect to Baztan and the general pattern of
application which bear directly on the ordering of phonological
and morphological operations.

2. Vowel Assimilation in Baztan

2.1. Interaction with other rules
Before studying the distribution of Vowel Assim ilation we
must consider other rules that create or remove inputs for its
a p p lica tio n .
In Baztan, the Vowel Assim ilation rule interacts in a
crucial way with three other rules: Mid Vowel Raising, Stem-
Final Low Vowel Deletion and Palatalization.

2.1.1. Mid Vowel Raising

A firs t rule that interacts with Vowel Harm ony is Mid
Vowel Raising. This is a rule that raises mid vowels to high
when imm ediately to the left of another vowel; e.g.: /a d to -a /
[aétua] 'the donkey.' This rule is formulated in a linear fashion
in (4a) and in terms of parameters and conditions in (4b):
 47

(4) Mid Vowel Raising

a. [-low] —> [+high] / V

b. Operation; Insert
Argument: [+high]
Target Conditions: [-low]
Trigger Conditions: ____ V

This is not an assimilatory process. The feature [+high] is

then inserted rather than being spread from a neighboring
segment.
Basque dialects differ on w hether Mid Vowel Raising is
ordered before Vowel Assimilation, creating inputs for this rule,
or after Vowel Assimilation. In Baztan, where Mid Vowel Raising
is an optional rule, although preferred in rapid speech, Vowel
Harm ony precedes this rule in its application; high vowels
created by Mid Vowel Raising do not trigger Vowel Harmony. In
Arbizu and also in Biscayan dialects, such as the Markina and
Ondarroa dialects, where Mid Vowel Raising is obligatory, on the
other hand, the ordering of the rules is the opposite: Mid Vowel
Raising feeds Vowel Harmony. Compare the Baztan and Markina
forms in (5):
 48

(5) Baztan Markina

/e g u n -a / [egune] [egune] 'the day'
/g is o n - a / [gisona] [giëona] 'the man'
/b u r u - a / [burue] [burue] 'the head'
/a ë t o - a / [aëtua] [aëtue] 'the donkey'
/m e n d i- a / [mendie] [m endiée] 'the mountain'
/etS e - a / [e t^ ia ] [etSie] 'the house'

The two different surface forms in Baztan and Markina for

stems ending in mid vowels are obtained by two different
orderings of the rules of Mid Vowel Raising and Vowel
Assimilation, as shown in (6):

(6) a. Baztan
/b u r u - a / /a ë to -a /
Vowel Assimilation e
Mid Vowel Raising u
[burue] [aëtua]

b. Markina
/buru-a/ /a ë to -a /
Mid Vowel Raising u
Vowel Assimilation e e
[burue] [aëtue]
 49

2.1.2. Stem-Final Low Vowel Deletion

A second process that interacts with Vowel Assim ilation
is the reduction of vowel sequences which arise when vowel-
initial inflectional markers are added to stems ending in a low
vowel. In Baztan, and in most eastern and central dialects (and
also in Standard Basque), the final low vowel of the stem is lost
in these cases:

(7) /alaba/ [alaba] 'daughter'

/alaba-a/ [alaba] 'the daughter'
/alaba-ak/ [alabak] 'the daughters'
/alaba-ekin/ [alabekin] 'with the daughters'

(8) Stem-Final Low Vowel Deletion

a -> 0 / ]V

This is a rule whose application is limited to inflectional

morphology. It does not affect either m orphem e-internal vowel
sequences or sequences created in derivation or compounding.
The rule of Stem -Final Low Vowel D eletion crucia lly
applies before Vowel Assim ilation. Only this ordering gives us
the attested [m uge] 'the lim it' from /m uga-a/. The correct
derivation is thus as in (9a). Ordering Vowel Assimilation before
Vowel C oalescence, on the other hand, would give us the
incorrect form in (9b) (of. Salaburu (1983,139-40)). In (9c) it is
 50

shown that the uninflected form undergoes Vowel Assim ilation,

producing a surface form identical to that created by the
attachm ent of the singular determiner:

(9) a. /muga-a/ > muga - > muge [muge]

b. /muga-a/ - > mugea --> mugia *[m ugia]
c. /muga/ - > muge [muge]

W hat this implies is that Vowel Assim ilation cannot apply

in the firs t m o rp h o lo g ica l dom ain w here its s tru c tu ra l
description is met. The firs t dom ain w here the structure
d e s c rip tio n of Vowel A s s im ila tio n is m et is the stem ;
nevertheless, the rule does not apply at this point. It is not
possible, on the other hand, to maintain that this mode of
application is a consequence of the postlexical application of
Vowel A ssim ilation. As we w ill see in section 2.2, Vowel
A ssim ilation applies only lexically, com pounding and certain
types of derivation being excluded from the domain of the rule.

2.1.3. Palatalization
A final process that we must consider in connection with
Vowel Assimilation is Palatalization. In Baztan, only a glide can
trigger the palatalization of a consonant, and the operation of
palatalization invariably triggers the absorption of the glide (cf.
chapter 4, section 3.2). E.g. : /seyn-en/ [sehen] 'of which.

4 Glides a re actually not_ underlying but created by rule (see chapter 1,

 51

Since the glide that triggers Palatalization is also a potential

trigger of Vowel Assimilation, the absorption of the glide bleeds
the application of this rule. The fact is that both assim ilated
and unassimilated forms are found in these cases, e.g. /a ra y n -a /
[a ra h e ], [araha] 'the fish.' It thus seems that either ordering of
the two rules is allowed (of. Salaburu (1983, 169-70 and 185):

(10) a. /arayn-a/
Palat./Glide Absorption h
Vowel Assimilation ..........
[araha]

b. /a ra y n -a /
Vowel Assimilation e
Palat./Glide Absorption h
[arahe]

A lte rn a tiv e ly , P a latalizatio n could be alw ays ordered

before Vowel Assim ilation. In that case, palatal consonants
could constitute an optional trigger of Vowel Assim ilation.5

section 2.2).
^ In the dialect of Arbizu, there is quite clear evidence that palatal
consonants trigger Vowel Assimilation after a m orphem e-internal (i.e. non­
alternating) palatal consonant /a / is raised. This can be observed even in the
adaptation of new borrowings, e.g. [de^aro/Cetu] 'to develop' from Sp.
d e s a rro lla d o . Other examples: [texetu] 'ro o f pronounced [te/catu] in other
dialects, ultimately from Old Navarrese Romance te lla d o : [o x e r ] rooster,'
 52

To summarize the results obtained, the rules examined are

ordered as follows in the Baztan dialect;

(11) Stem-Final Low Vowel Deletion

P a la ta liz a tio n
Vowel A ssim ilation
Mid Vowel Raising

The derivations in (9) show that Vowel Assim ilation must

follow Stem-Final Low Vowel Deletion. The examples in (5)-(6)
show that Vowel A ssim ilation is ordered before Mid Vowel
Raising in Baztan. Finally, we concluded that the results in (10)
co u ld be o b ta in e d o rd e rin g P a la ta liz a tio n a fte r V ow el
A s s im ila tio n .
A fter determ ining the way Vowel A ssim ilation interacts
w ith o the r rules th a t create or rem ove contexts for its
application, we can now exam ine the interaction of this rule
with d iffere nt m orphological operations. In this respect, the
Baztan dialect can be taken as representative of the situation in
many other Basque dialects.

[o/Car] in other dialects. But these palatals can also be viewed as glide-plain
consonant sequences at a deeper level (cf. chapter 4 , section 3.2).
 53

2.2. Domain of Vowel Assimilation in Baztan

2.2.1. Morpheme-internal
In the dialect of Baztan, as in many other Basque dialects,
the Vowel Assim ilation rule form ulated in (2)-(3) causes the
sequences I\0 q 3J - and /u C o a /-/u C o e / to be neutralized
in favor of the second member of each pair. This can be observed
by com paring Baztan items with the corresponding words in
other Basque dialects that have not undergone this process. To
show the effects of this phonological process, Baztan examples
are given in (12) together with their cognates in Standard
Basque. The Baztan data are mainly from N'Diaye (1970) and
S a la b u ru (1983).® All exam ples are given in conventional
Basque orthography. As the examples show, both [a] and [e] occur
in Standard Basque after a high vowel in a lexical item
(examples in (12a) and (12b) respectively), but the distinction
between [a] and [e] has been neutralized in this position in favor
of [e] in Baztan:

(12) Standard Baztan

a.fabrika fa b rik e 'fa c to ry '
ik a ra tu ik e ra tu 'frig h te n '
ira b a z i idebazi 'to earn, win'

G I want to thank Pello Salaburu for reading an earlier version of this section
and discussing the data. I am also grateful to Txomin Irungaray for comfirming
some of the data presented here.
 54

bealunikatu belauniketu 'to kneel'

ik a tz ik e tz 'co a l'
k ir a ts k ire ts 'stench'
ik a s i ik e s i 'to learn'
izan izen 'to be'
b ih a r b ie r 'to m o rro w '
usain usein 's m e ll'
hura ure 'th a t'
u r t a r r il u r t e r r il 'January'
b a ta ia tu b a ta ie tu 'to baptize'

u rte urte 'year'

izen izen 'name'
iz e rd i iz e rd i 's w e a t'

The rule of Vowel Assim ilation whose effects in Baztan

Basque are illustrated in (12) constitutes more than a purely
diachronical process which has brought about a neutralization of
a contrast in a few lexical entries. In fact, the sound change
illustrated in (12) gave rise to a very active rule in the
phonology of the Baztan dialect manifested in a great number of
alternations. This is because this process of vowel assimilation
applies beyond the boundaries of the lexical item in a number of
m orphological contexts.
 55

2.2.2. Inflectional morphology

A first m orphological context where the rule of Vowel
A ssim ila tion applies is in in fle ctio na l m orphology. In (13)
examples are given of inflectional markers whose first vowel is
/a/, which surfaces as [e] when attached to a stem whose last
vowel is high, such as buru 'head,' but as [a] otherwise, as in
the corresponding inflectional forms of the noun g o g o 'spirit,
w ill' offered for com parison:^

(13) Inflectional suffixes:

C ase/num ber /g o g o / /b u r u /
abs. sg. gogoa burue
abs. pi. gogoak buruek
dat. sg. gogoai buruei
adlat.sg. gogora burure
ablat. indef. g og otaik b u ru te ik
ines. indef. gogotan buruten
adlat. indef. gogotara b u ru te ra
gen.-loo. indef. gogotako buruteko

The genitive singular m arker /-aren/ is often reduced to

[ayn], in which case the front glide thus created can also trigger
Vowel Assim ilation in a follow ing suffix (of. Salaburu (1983,
2 0 0 -3 )):

7 F u ll inflectional paradigm s for the Baztan dialect are given in N'Diaye

(1970, 117-26). Our examples are from N'Diaye.
 56

(14) /andre-a-ren-a/ [andriayne]

/andre-a-ren-tako/ [andriayndeko]

Notice, incidentally, that the examples in (14) present two

types of surface opacity with respect to rules operating in the
dialect. First of all, the second low vowel in the examples in
(14) is not raised to [e] by Vowel Assimilation, even though it is
to the right of a possible trigger in the surface form. This is
because the rule that creates this high vowel. Mid Vowel
Raising, is o blig a to rily ordered after Vowel A ssim ilation in
Baztan (of. section 2.1). That is, the derivation of the forms in
(14) is as in (15), after the application of the rules that reduce
/aren/ to [ayn] and other irrelevant rules:

(15) /a n d re a y n a / /a n d re a yn d a ko /
Vowel Assimilation e e
Mid Vowel Raising i i
[andriayne] [andriayndeko]

Secondly, a form such as [andriayne] presents a possible

context for palatalization. Salaburu points out that, along with
[andriayne], forms such as [andriahe] and [andriaha] are indeed
also found in Baztan. These last two form s show th at
Palatalization-cum -G lide Absorption can apply either before or
after Vowel Assim ilation (or that Palatalization being ordered
 57

before Vowel A ssim ilation, palatal consonants are optional

triggers of Vowel Assim ilation; of. Section 2.1). If it applies
before, the context for Vowel Assimilation will be removed and
the re su lt w ill be [a n d ria h a ]. S im ila rly, the o rd ering of
P a latalization before the reduction of /a re n / to [ayn] w ill
produce opaque forms where the context for Palatalization is
met but the rule does not apply (i.e. [andriayne]).

2.2.3. Derivational morphology

Not all der i vat i onal suffixes present t he same behavior

w i t h r e s p e c t t o V o w e l A s s i m i l a t i o n . A f i r s t , l a r g e r , gr oup of

d e r i v a t i o n a l s u f f i x e s w i t h a l o w v o w e l in t h e i r f i r s t s y l l a b l e do

un d e r g o a s s i m i l a t i o n w h e n a t t a c h e d t o a r o o t w i t h a hi gh v o w e l

on i t s l a s t s y l l a b l e . E x a m p l e s are g i v e n in ( 1 6 ) . For e a c h s u f f i x ,

b o t h e x a m p l e s s h o w i n g a s s i m i l a t i o n and c o m p a r a t i v e e x a m p l e s

s h o w i n g t h e u n d e r l y i n g v o w e l o f t h e s u f f i x are gi ven:

(16) /-S ka / d im inu tive

[afariS ke] 'little dinner' of. [la ra S k a ] 'little prairie'
[m endiS ke] 'little mountain'

/ - lari/ profession

[tratuleri] 'dealer' of. [eskelari] 'beggar'

 58

/-ka/ adverbial
[aldiske] 'by turns' cf. [lasterka] 'quickly'

/-ts a / a c tiv ity

[laguntse] 'help' cf. [arantsa] 'fishing'

A second, more reduced, group of derivational suffixes does

not undergo the assimilation rule. Examples showing absence of
assimilation in contexts where we would expect it to take place
are given in (17):

(17) /-g a re n / ordinal

[b iga re n ] 'second'; [irugaren] 'third'; [law garen] 'fourth'
/-ta r/ o rig in
[menditar] 'mountain-dweller'; [ayskundar] 'inhabitant of
A riz k u n '
/-ag o / more
[a rg ia g o ] 'clearer'; [tragkilago] 'quieter'

F inally, w ith the su ffixes /- g a r i/ 'able' and /- ta S u n /

'quality,' the usage varies. In N'Diaye's texts, along with the
a ssim ila te d iz io e rrl 'frightening' we find the non-assim ilated
a r r io a r r i 'su rp risin g .' The suffix /-ta S u n / g e n e ra lly show s
assimilation. We thus find in N'Diaye's corpus examples such as
ja k in te s u n 'knowledge' and b u ru te s u n 'th o u g h tfu ln e ss.' But
 59

s a m u rta s u n 'softness' does not show assim ilation, and both

arintasun and arintesun 'speed' are found.
Verbal suffixes generally undergo assim ilation. Examples
are given in (18) with the com plem entizer /-la/, which attaches
to in fle c te d form s, the s u ffix /-ta / w hich a tta c h e s to
perfective verbs to form a participle and article /-a/, which
may also attach to p erfective form s, also creating past
p a rtic ip le s :

(18) /-la / 'that'

d u a u -le 'that we have' cf. dire-la 'that they are'

/-ta/ past participle

itz u li-te 'turned' cf. e tz a n -ta 'lied'

/-a/ singular article

itz u li-e 'turned' cf. e rra n -a 'said'

2.2.4. Compounds
Vowel Assim ilation does not apply across an intervening
morpheme boundary in compounds. This is shown in (19) with
exocentric compounds (bahuvrlhl) (a), subcompounds (tatpurusa)
(b) and cocompounds (dvandva) (c):
 60

(19) a. exocentric compounds

buru-aundi *buru-eundi 'big-head'
begi-makur *begi-mekur 'cross-eyed'

b. subcompounds
su-arri *su-erri 'fire-stone'
begi-tarte *begi-terte 'face' (lit. 'between-eyes')
ardi-alde *ardi-elde 'flock' (lit. 'sheep-group')

c. cocompounds
jaun-andreak *jaun-endreak 'ladies and gentlemen' (lit.
's ir - la d ie s ')
iri-basuetan *iri-besuetan 'in town and forests'®

2.2.5. Vowel Assim ilation across w ord-boundaries: cllticization

The a pp lica tio n of Vowel A ssim ila tio n across w ord-
boundaries is limited to some very specific environments. Vowel
Assim ilation does affect determiners with a low vowel in their
firs t syllable such as the dem onstrative g j i 'this,' and the
n u m e ra l/in d e fin ite a rtic le b a t 'one/a' and its plural b a tz u k
'some.' Examples are given in (20):

® N 'D ia y e (1 9 7 0 , 1 15 ) offers two e x a m p le s of com po unds that,

exceptionally, show the effects of vowel assimilation. These are s u k e ld e
'kitchen' from 'fire' and aide 'side,' with an intrusive k ; and suker 'fever,'
from &Ü 'fire' and g a i 'flame.' Very likely these two exam ples are not felt as
com pounds by the speakers. Notice in particular the non-com positional,
unexpected meaning of suker .
 61

(20) buru eu 'this head' cf. base au 'this forest'

esku aundi eu 'this big hand'
mendi bet 'one mountain' of. baso bat 'one forest'
egun betzuk 'some days' of. baso batzuk 'some forests'
erri betean 'in one village' of.baso batean 'in one
forest'

As argued in c h a p te rl, these determiners are not suffixes,

but independent words, which can appear alone in an NP. When
they appear to the right of another word in the Noun Phrase,
how ever, they show the b eh avio r of p ho no log ical c litics.
Determiners act as clitics also with respect to other rules. For
the clitic-like behavior of b a t, evidence internal to the Baztan
d ia le ct is given by the fact that its initial /b / can delete
intervocalically, a process that takes place only within word-
boundaries, eg.: /buru bat/ [buruet] 'one head.' As shown in (20),
the assim ilation of the vowel of the determ iner is independent
of whether the triggering word is a noun such as buru 'head' or an
adjective such as a u n d i 'big.' The examples in (20) also show
that there is no assimilation in other contexts such as between
a noun and its adjective, as in esku aundi 'big hand' (not *e s ku
e u n d iT Assim ilation across words within the Noun Phrase is
thus lim ited to cliticization, an operation which has precisely
the effect of erasing a word-boundary.
 62

Certain auxiliary verbs also undergo assim ilation. Among

them are the intransitive present indicative forms dâ. (3sg), o a ra
(1pl) and z a ra (2 sg formal) and the transitive imperative z a z u
(2 sg formal):

(21) torri de 'he has come' of. g an da 'he has gone'

gain de 'he will go' of. torri ko da 'he will come'
torri gera 'we have come' of. gan gara 'we have gone'
torri zera 'you have come' of. gan zara 'you have gone’
in zezu 'do it' of. jan zazu 'eat it'

Again, these auxiliaries do not form a lexical word with

the main verb, from a syntactic point of view. They can be
separated from the main verb by a modal as in (22a) and they
are preposed to the main verb in negative clauses, as in (22b):

(22) a. torri omen da 'he has apparently arrived'

b. ez da torri 'he has not arrived'

The auxiliaries can also be used as main verbs in the

clause. In these instances, the forms susceptible to undergo
assim ilation will do so even if the trigger is not a verb, but
some nominal expression:
 63

(23) patxi de 'it is Patxl'

oi de 'it is that one'

O ther form s of the auxiliary, on the other hand, never

undergo a ssim ila tio n , including the o th e r persons of the
intransitive present indicative not mentioned above (naiz (1sg),
aiz (2 sg, informal) and zate (2 pi).; E.g. [tori nays] *[tori neys] 'I
have come.'
We must conclude that only a few auxiliary form s can
behave like clitics and thus undergo Vowel Assimilation.
A th ird elem ent th a t undergoes A ssim ila tio n is the
co n ju n ctio n fe U a 'and,' which loses its initial vowel when it
becomes a clitic:

(24) ardi te auntz ' sheep and goat' of. zabal ta eder 'wide and
pretty'
gorri te urdin 'red and blue'

2.2.6. Exceptions
To com plete the presentation of the data on Vowel
Assim ilation in Baztan, I should mention that there is a small
number of morpheme -Internal exceptions to the rule. These are
Items presenting a morpheme-internal diphthong [wa] which is
not raised to [we]:
 64

(25) [e^kwara] ' Basque language'

[kwarto] 'room'
[gwarda] 'guard'
[ëanwanak] 'St. John's day's festivities'
[kwadPi/Ce] 'group of people'
[kwadro] 'picture'

For all these items (all of them but the firs t Spanish
loanw ords), underlying /o a / can be safely assum ed, since
m orphem e-internal /oa/, /ua/, /w a/ never contrast in Basque.
Underlying /oa/ becomes [ua] by Mid Vowel Raising, arule that,
as we saw in section 2, is ordered after Vowel Assimilation in
Baztan. The sequence /ua/ then becomes surface [wa] by gliding
of the high vowel. Aside from those cases, no true exceptions to
Vowel Assim ilation seem to exist. A putative exception would
be [guratëo] 'parent.' But this item is probably better understood
as a compound containing /gure/ 'our.' In Basque, one's own
relatives are usually referred to with the plural first person
possessive, e.g. cure anaia means 'my brother' but literally it is
'our brother.'
Some apparent mono-morphemic exceptions are found both
in N'Diaye (1979) and Salaburu (1983). Pello Salaburu, in a
personal com unication, however, points out to me that these
items do not constitute genuine exceptions to the rule. Instances
of i- a . u - a . sequences in the trancriptions have one of two
 65

sources. Firstly, some Baztan speakers do not apply Vowel

Assim ilation in overcareful pronunciation. In this respect, this
rule of Vowel Assim ilation in Baztan has a sim ilar status to
Flapping in American English.
Another sort of occasional violation of the assim ilation
rule in contexts where it can apply is of a phonetic/phonological
nature. When the targ e t of Vowel A ssim ilation im m ediately
follows a front glide, a palatal consonant or an affricate, quite a
bit of variation seem s to be possible, the m ost com m on
realization being a vowel of intermediate value between [a] and
[e], which I will represent as [e]. Eg. : [itéeéo] 'sea,' [g ilts s w f]
'w a ln u t,' [m in tse yre ] 'language,' [byeram un] 'fo llo w in g day,'
[ih e w te ] 'c a rn iv a l,' [ko n ]fys ntsa] 'c o n fid e n c e ' (P. S alaburu,

p.C.).®

2.3. Analysis
Summarizing the data in the preceding subsection. Vowel
A s s im ila tio n a p p lie s in u n d e rive d fo rm s, in in fle c tio n a l
m orphology, in m ost d erivatio na l suffixes and also affects
determiners, certain verbal auxiliaries and the conjunction 'and.'
On the other hand. Vowel Assimilation does not generally apply
across-word boundaries, across the members of a compound, or
with certain derivational suffixes.

9 In the dialects of Ondarroa and Arbizu, to be studied in following sections,

neither of these two sorts of variation are found.
 66

A first approach to trying to make sense of these data

within the postulates of Lexical Phonology would be to assume
that a single rule of Vowel Assim ilation applies repeatedly in
the phonology of the language, in connection with d ifferent
m orphological operations. Using the term cliticization to refer
to determ iners, the conjunction and the verbal auxiliaries that
undergo the rule, and by distinguishing two types of derivational
s u ffix e s , we can su m m arize the a p p lic a tio n of V ow el
Assim ilation to different m orphological levels or operations as
in (26):

(26) morphology Vowel

underived forms yes
derivation 1 yes
derivation II no
compounding no
in fle c tio n yes
c lit ic iz a t io n yes

Now, we cannot allow the rule to apply to underived lexical

items before inflection takes place. This is because, as argued in
section 2, in w ords such as /a ita -a / [ayte] 'the fa th er,' or
/m uga-a/ [m uge] 'the limit,' the rule that reduces the sequence
of two low vowels created by the suffixation of the article must
apply before Vowel Assim ilation gets a chance to apply. Were
 67

Vowel Assim ilation to apply cyclically , first to the stem, then

to stem plus suffix, we would obtain /aite-a/, /m uge-a/ by the
application of Vowel Assim ilation at the stem level. Then Mid
Vowel Raising would get a chance to apply after affixation of
the Inflection and the unattested *[aytia], *[m ugia] would result
as surface forms. We must conclude that the first application of
V ow el A s s im ila tio n m ust be p o stp o n e d u n til a fte r the
m orphological operation of inflection, but then be allowed to
apply to form s not derived by that operation, to account for
Vowel Assim ilation within stems."* ^
The lexical distribution of Vowel Assim ilation in Baztan is
rather typical if we look at other Basque dialects with the same
rule. Txillardegi (1984) reviews the Vowel Assim ilation facts of
several other Basque dialects spoken in Navarre, Biscay and
Guipuscoa. I have also examined the High Navarrese variety of
Lakuntza which fits into the general pattern."* "* The results are
that in no Basque dialect is Vowel Assimilation allowed to apply
across word boundaries or across a compound-boundary. There is
also in every case a list of derivational suffixes which fail to
undergo the rule. This list of exceptional su ffixes varies

Salaburu (1983 ) proposes the existence of two rules. O ne acts as a

Morpheme Structure Constraint, disailowing a high-iow vowei sequence which
w iil, h o w eve r, not apply to ste m -fin a l /a /, being re stric ted to a
preconsonantai environment. Then another ruie, identicai to the Morphem e
Structure Constraint but without this restriction, wiii appiy to ail other such
sequences after Low Vowel Coalescence.
^ ^ I am grateful to Edurne Arregi for the Lakuntza data.
 68

somewhat from dialect to dialect but typically includes -c a rre n .

-o a r ri. -ta s u n . and -ta r and, often, -aoo and the adverbial -ka.- In
some High Navarrese varieties (Lakuntza, as well as the
w ritten texts examined by Txillardegi (1984, 215-18)), some
inflectional suffixes such as the adlative -ra, (e g. [mendira] 'to
the m ountain') are also excluded. The d e te rm in e r b a t,
dem onstratives, and the conjunction feUa vary in their behavior
from dialect to dialect. A few lexical exceptions in underived
items are also frequently found.
Txillardegi (1984) proposes to mark those suffixes that do
not undergo the assimilation rule with a "# " boundary, the same
boundary that is assigned to compounds. O ther suffixes (and
clitics undergoing assim ilation) w ould be m arked with a "+"
boundary. Vowel Assim ilation can then be form ulated as a rule
that can operate across a "+" boundary, but not across a stronger
boundary. This SPE-type analysis clearly works. The derivation
of sample Baztan examples is given in (27). To obtain the correct
results in cases such as /m uga+a/ 'the limit' we simply need to
order S tem -Final Low Vowel D eletion (LVD) before Vowel
A ssim ilation (VA):

(27)/m uga/ /m uga+a/ /p a t§ i+ d a / /tra n k il# a g o / /b u ru # a u n d i/

LVD 0
VA e e e BLOCKED BLOCKED
[muge] [muge] [patSiée] [trankilago] [buruawndi]
 69

An im portant goal of Lexical Phonology has been to

e lim in a te th e d e g re e of u n re s tric te d a b s tra c tio n in
representations that the use of d iffe re n t boundary-m arkers
a llo w s. Instead, the e ffe cts th a t the d iffe re n t types of
boundaries in the SPE model can produce are derived from the
interaction of m orphology and phonology. D ifferent m orpho-
phonological levels or strata are proposed to exist. In each
stratum a set of phonological rules operate in conjunction with
the m orphology created at that level. For Baztan Basque and
other dialects where Vowel A ssim ilation applies in a sim ilar
manner, two strata can be proposed: a First stratum where the
attachm ent of the majority of suffixes takes place (in Baztan all
inflectional and m ost derivational suffixes), and where also
underived forms may undergo rules; and a second stratum where
com position and exceptional su ffixatio n is realized. Vowel
Assim ilation applies only in stratum I:

(28)
Stratum I (non-cyclic)
M orphology Phonology
Stem-Final Low Vowel Del.
Underived items and Vowel Assim ilation
suffixation (and other stratum I rules)
 70

Stratum II
Morphology Phonology
Exceptional suffixation and
Compounding No V.A.

When clitic groups are formed In the syntax, these groups

are sent back to stratum I.

It must, perhaps, be pointed out that there is nothing

im plausible in Basque about assigning com pound-form ation and
even the attachm ent of some or all derivational suffixes to a
later stage than the suffixation of inflectional m arkers. As
shown in chapter 1, only the last word in a Noun Phrase is
in fle c te d . In fle c tio n a l s u ffix e s w ill a tta ch to the second
member of a compound before compound form ation takes place
and to a derivational suffix before this suffix is added to the
stem . C o n sid e r the mo rpho-pho no log ical d e riv a tio n of a
compound or a word containing one of the suffixes which we
have suggested are attached at the same level as compound-
formation, when compared with a Noun Phrase containing a noun
and an adjective:

(29)/m util buru-aundi-a/ /iru -g a re n -a / /mendi g o r i- a /

boy head big the three nth the mountain red the
'the big-headed boy' 'the third' 'the red mountain'
 71

Stratum I Morpho-phonology:
[aundie] [garena] [gorie]
Stratum II Morpho-phonology:
[buruaundie] [irugarena]

I will now offer data from other Basque dialects which, in

addition to Vowel Assim ilation, possess other rules that support
a bi-stratal organization of the lexicon. In all dialects, all or
m ost in fle ctio n a l m arkers are attached at stratum I and
compound formation takes place in stratum II. Dialects differ in
the assignm ent of derivational suffixes to one stratum or the
other.

3. Arbizu Lexical Phonology

3.1. Vowel Assim ilation
In the dialect of Arbizu, Vowel Assim ilation takes place
essentially as in Baztan except for a couple of details. A first
difference is that in Arbizu the rule has the constraint that in
order to undergo the rule a target /a/ must be followed by a
consonant. As the exam ples in (30a) show, the final /-a/ of
singular absolutive forms does not undergo the rule, but the
plural a bsolutive/ergative form s show the application of the
rule. Singular ergative form s, where the ergative marker /-k/

12 1 am indebted to Xabier Arbizu for all the data from the dialect of the town
of Arbizu used in this dissertation.
 72

follow s the /-a / of the absolutive form s, are identical to the

plural absolutive/ergative form s, showing that the addition of
the /-k / suffix makes singular /-a/ into a target of the rule.
Being excluded from Vowel Assim ilation is thus not a property
of the singular determiner /-a/. This affix is excluded only when
w ord-final. The same restriction is shown in the exam ples in
(30b) with other su ffixes: the adla tive /-ra /, the adverbial
/ - k a / and the distributive /-na/:

(30)
a. U n in fl abs sg abs/erg pi & erg sg
[aëto] [adtua] [aëtuek] 'donkey'
[m endi] [m en d ija ] [m e n d ije k] 'm o un ta in'
[edku] [e^kuba] [edkubek] 'hand'
[t^ a k u r] [tSakura] [tS akurek] ’dog'
[alaba] [alaba] [alabak] 'daughter'
[gison] [gisona] [gisonak] 'man'

b. [m endira] 'to the mountain'

[mendireho] 'towards the mountain'
[biha] 'two for each'
[biheka] 'two by two'
[a rik a ] 'throwing stones'
[a rike so] 'throwing of a stone'
 73

There are two ways to capture the restriction to Vowel

Assim ilation illustrated in (30). One way would be to say that
the target must be followed by a consonant. The other way would
be to assume that the target must not be word final, i.e. must be
follow ed by some other segment, either vowel or consonant
w ithin the word. There are facts that suggest that the first
interpretation is the correct one. The singular dative /-ai/ and
the clitic iz in g dem onstrative /a u/ 'this one' do not show
application of the rule, as shown in (31). The explanation seems
to be that there is no follow ing consonant. This is w ithout
considering possible functional explanations in the case of the
dative singular (i.e. its distinction from the plural dative /-ei/):

(31) [mendijay] 'to the mountain'

[e^kubay] 'to the hand'
[mendijaw] 'this mountain (abs)'
[e^kubaw] 'this hand (abs)'

The rule of Vowel Assim ilation in Arbizu must therefore

have the condition that the target be followed by a consonant.
Notice that this is not equivalent to saying that the target must
be in a closed syllable. As the exam ples in (30b) show, the
following C may be in the onset of a following syllable. A linear
formulation is given in (32):
 74

(32) Vowel Assim ilation in Arbizu

V > [-low] / [+high] C q C

As we will imm ediately see, Vowel Assim ilation is limited

in its application to Stratum I morphology, as in Baztan. The
effects of following suffixes in creating the environm ent for a
low vowel to undergo the rule require Vowel Assim ilation to
apply in a noncyclic fashion at the end of Stratum I.
A second difference that Vowel A ssim ilation in Arbizu
presents with respect to the application of the same rule in
Baztan is that in Arbizu, Vowel Assim ilation follows Mid Vowel
Raising. High vowels created by Mid Vowel Raising trigger
Vowel Assim ilation in Arbizu, unlike in Baztan. In both dialects
the underlying q ua lity of the stem -final vowel is, however,
retrievable, since in Arbizu an epenthetic consonant is inserted
after stem-final high vowel and another vowel (see 3.3. below):

(33) U ninfl. Abs pi

A rb iz u Baztan
a. [beéo] [beduek] [be^uak] •the arms'
[berde] [berdiek] [berdiak] •the green ones'
b. [edku] [eëkubek] [e^kuek] •the hands'
[m endi] [m e n d ije k] [m endiek] •the mountains'
 75

Aside from these differences, Vowel A ssim ilation takes

place in Arbizu in fundamentally the same way as in Baztan. As
in Baztan the rule applies m orphem e-internally; [iC oa], [uC oa]
sequences are not found and new borrow ings are quite
c o n s is te n tly m o d ifie d to co m p ly w ith th is re q u ire m e n t.
Examples are provided in (34):

(34) Standard A rb iz u
ik a s - i [ik e é i] 'to learn'
b e la u n ik a -tu [b e la w n ik e tu ] 'to kneel'
ila r g i [id e rg i] 'moon'
iz a r [is e r] 's ta r '
in d a r [io ge r] 's tre n g th '
b iz a r [b ise r] 'beard'
izan [isen] 'to be'
ip a r [ip e r] 'north (wind)'
m a ita -tu [m a yte tu ] 'to love'
p in ta -tu [p in te tu ] 'to paint,' Sp. pintado
ig u a l [ig w e l] 'same' Sp. igual
k o n fia n tz a [k o rg f ij e n t s a ] 'confidence,' Sp. confianza
k r is t a l [k r ië te l] 'g la ss'

As the examples in (34) above show, the rule also applies

to inflectional and some derivational suffixes, subject to the
restriction that the ta rg e t be follow ed by a consonant, a
 76

restriction that also applies to m onom orphem ic w ords, e.g.

[fa b rik a ] 'fa c to ry ' not *[fa b rik e ]. A lso, to illu s tra te the
application of the rule in inflectional m orphology, com pare for
in s ta n c e [p a a te -ta é ] 'about/by means of the w alls' with
[m e n d i-te ^ ] 'about/ by means of the mountains' or [paate-tan] 'in
w alls' with [m enti-ten] 'in mountains.'
As in Baztan, Vowel A ssim ila tio n is excluded from
applying between the members of a compound or to derivational
suffixes such as /- ta r/ 'native of,' a fact whose generality for
all Basque dialects was already noted by Azkue (1923: 48). The
dialect of Arbizu is in this respect more restrictive than that of
Baztan since the vast m ajority of derivational affixes are in
fact excluded from the rule. Derivation, as a rule, thus takes
place at Stratum II, a fact that the nonapplication in these
contexts of other rules such as Palatalization and Consonant
Insertion confirms (see sections 3.2 and 3.3 immediately below):

(35) Compounding and Derivation

[buu -aw n di] 'big headed' *[b u u -e w n d i]
[m e n d i-a ld e ] 'm o u n ta in -s id e ' *[m e n d i-e ld e ]
[m e n d i-s a le ] 'm o un ta ine er' *[m e n d i-s e le ]
[a rb is u -a r] 'native of Arbizu' * [a rb is u -e r]
[b a ^ e ri-ta r] 'fa rm e r' * [b a ^ e r i- te r ]
[a rg i-ta ^ u n ] 'c la r it y ' * [a rg i-te ^ u n ]
[a ré i-ta ld e ] 'flock of sheep' * [a r d i- te ld e ]
 77

[a r i- g a r i] 's u rp ris in g ' * [ a r i- g e r i]

[b i-g a re n ] 'second' *[b i-g e re n ]

As in Baztan, [^u k e ld e ] 'kitchen' shows assim ila tio n in

spite of being originally a compound of l i u l 'fire' plus /aide/
'side.'
A n o the r d iffe re n ce betw een the a p p lica tio n of Vow el
Assim ilation in Baztan and in the dialect of Arbizu is that in
this la tte r d ia le c t there are no clitic iz in g elem ents w hich
undergo the rule. The com plem entizers /la / 'that' and /lakes/
'because' do not show assimilation , e.g. [duula] 'that we have it,'
[duulakos] 'because we have it.' A reason for clitic groups to be
excluded from cliticization is that in most cases of cliticization
in other dialects, such as Baztan, the target vowel would be
w ord-final and thus would not meet the environm ent that the
rule has in Arbizu.
The determ iner bat 'a/one' has developed two allomorphs
[at], used after mid or low vowel, and [bet], after high vowel or
consonant. Compare the Baztan examples, where the alternation
[bat]/[bet] is determined by the rule of Vowel Assim ilation, with
the A rbizu exam ples w hich show [bet] a fte r c o n so n a n t
regardless of the quality of the preceding vowel;
 78

Standard Baztan A rb iz u
alaba bat [alababat] [alabaat] 'a daughter’
baso bat [ba^obat] [ba^oat] 'a forest’
gizon bat [gisom bat] [gisombet] 'a man’
txakur bat [tS akurbet] [tSakurbet] ’a dog'
mendi bat [m endibet] [m endibet] 'a mountain'
hots bat [o^bat] [o^bet] ’a sound'

3.1.1. Exceptions to Vowel Assimilation

Aside from a group of system atic e xceptions to be
discussed imm ediately below, other exceptions are rare. I have
found [g u ra tso ], for w hich the sam e co m p o u n d -s tru c tu re
explanation as in Baztan is available.
A system atic exception to Vowel Assim ilation is presented
by items with a long low vowel. In most Basque dialects, long
vowels are rare or nonexistent. A peculiar characteristic of the
Arbizu dialect is its welldeveloped length contrast in vowels.
The historical origin ofthe long vowels is as follow s: A
common optional rule of Basque is the deletion of [b, b, g, r]
intervocalically inside a word. This rule became obligatory for a
large num ber of item s in A rb izu , w ith the co n se q u e n t
d is a p p e a ra n c e of th e consonant from the u n d e rly in g
representation. This historical process created long vowels in
items such as /irabasi / > /iraasi/ ’to w in.' Other long vowels
were created by the subsequent total assim ilation of the two
 79

vowels in contact if they were both non-high and of different

qua lity: /egon /> /eon/ > /o o n / 'to stay'; /parete/> /paete/>
/p a a te / 'wall.'"’ ^
A set of system atic exceptions to the rule of Vowel
A ssim ilation is constituted by instances where the possible
ta rg e t low vowel is long. Long or gem inated /a a / fa ils to
undergo Vowel Assim ilation both morpheme internally, and in
inflectional m orphology:

(37) [ira a s i] 'to win, earn'

[yaanon] 'day before yesterday'
[m endijaan] /m endi-aan/ 'of the mountain'
[e^kubaaki] /e^ku-aaki/ 'with the hand'
[m enditaaho] /m endi-taa-ho/ 'up to the mountain'

13 This is only an account of the most general process of long vowel creation.
There are other more restricted processes in the dialect as well. A striking
case is the compensatory lengthening of the first vowel when the second vowel
in the sequence formed after consonant deletion was high, as in /lag un/ >
/la a u n / 'friend'; /te g i/> /te e i/ 'place' (cf. Hualde (in prep, a) for details of
historical change). W hat is surprising about examples like these is that the loss
of a consonant in onset position would seem to have triggered vow el
lengthening, against the general assumption that onsets are weightless and will
not trigger compensatory lengthening when they are lost (cf. Hyman 1984).
The explanation for these examples seem to be found in a gliding process with
conservation of moras. The derivation seems to have proceeded as follows:
(i) m m m m mm
I I I I 1/
l a g u n > I a u n > I a w n
The reason for the lengthening of the first vowel is that in Arbizu [au] does
not contrast with [aw]. A high vowel must obligatorily glide (i.e. lose its mora)
after another vowel.
 80

G iven the fa c t th a t in A rbizu the rule of Vow el

Assim ilation (rule (30) above) requires a follow ing consonant,
an explanation for the absence of assim ilation in (37) would be
that the possible target /a / is not followed by a consonant but
by a second vowel /a/. This explanation would force us to do
away with the now standard representation of long vowels as
single segm ents linked to two tim e units. Such a move is,
however, not necessary. We may instead assume that the rule of
Vowel A ssim ilation w ill not apply to a long vowel due to
gem inate blockage (cf. Hayes (1986), Schein and Steriade
(1986)). The rule of Vowel A ssim ilation fo r Arbizu can be
schematized as in (38a). This rule will accept (38b) as an input,
but not (38c):

(38) a. Co V C
I
a —> e / i ___

b. V C
I I
/mend i a k / [mendijek]

c. W O
V I
/mend i a n/ [mendijaan]

To sum m arize the Arbizu facts with respect to Vowel

A ssim ilation, this rule is, as in Baztan, lim ited to Stratum I
m orphology which includes the m orphem e-internal domain and
 81

some affixation, mostly inflectional. There is a sm aller number

of actual instances of application of the rule in Arbizu due to
two facts: on the one hand, there is a constraint that requires a
consonant following the target low vowel; on the other hand,
some derivational affixes that undergo the rule in Baztan are
excluded in Arbizu and there are no instances of Vowel
A ssim ila tio n in c litic groups. We concluded th a t Vowel
assim ilation is a noncyclic stratum I rule.
Now we will study two other rules of the dialect that make
refe ren ce to the sam e m orph o lo g ica l dom ains as Vowel
A s s im ila tio n : P a la ta liz a tio n and C onsonant In s e rtio n .
Interestingly, there are affixes that present the phonological
context for more than one of these three rules. As I will show, in
general, stratum I affixes undergo all rules for which they meet
the phonological requirem ents and stratum II suffixes fail to
undergo any of them.

3.2. Palatalization
In Arbizu /t/ /I/ and Ini are palatalized after /i/, syllabic
or nonsyllabic. This Palatalization rule applies both morpheme-
interna lly and with most inflectional suffixes. Exam ples are
given in (39):

(39) / a i t a / [a yt'a ] 'fa th e r'

/ a it o n a / [ayt'ona] 'g ra n d fa th e r'
 82

/m a lte / [m ayt'e] 'beloved'

/Ito / [it'o ] 'to drown'
/b e r i- tu / [b e rit'u ] 'to renew'
/d - it- u - t/ [d it'u t] '1 have them'
/ e g i- t e n / [eyt'en] 'to do/make (imp.)'
/ilu n/ [i/Cun] 'da rk'
/Ib ill/ [ibi/Ci] 'to walk'
/M e / [i/ce] 'm onth'
/n iu til- a / [muti/Ca] 'the boy'
/a u to m o b il- a / [awtomobi/Ca] 'the car'
/ d a d i- la / [dabi/Ca] 'that it be'
/m ln - a / [m iha] 'the pain'
/ x a k ln - ik / [x a k lh ik ] 'know n'
/s e ln - e n / [sehen] 'of which'

M on om o rp he m ic e xce p tio n s are re s tric te d to rece nt

borrowings such as [kilo] 'kilo,' [piloto] 'pilot,' [minuto] 'minute'
and [ka^ino] 'casino.'
The exam ples in (39) show that Palatalization operates
both m o rp h e m e -in te rn a lly and a c ro s s s u ffix b o u n d a rie s.
Derivational suffixes are however system atically excluded from
P a la ta liza tio n . The su ffixes /- ta r / 'native, inhabitant' and /-
ta ^ u n / 'quality' are not affected by Palatalization:
 83

(40) / b a ë e r i- t a r / [b a ë e rita r] ' farm er'

/ b is k a i- t a r / [b is k a y ta r] ' Biscayan'
/a r g i- t a ë u n / [a rg ita ^ u n ] ' cla rity'
/a u n d i- ta ^ u n / [aw nditaëun] ' greatness'

Given our bi-stratal conception of the m orpho-phonology of

the language, the failure of Palatalization to apply to /-ta r/ and
/-ta ^ u n / would indicate that these are stratum 11 suffixes and
th at P alatalization, like Vowel A ssim ilation, is restricted to
stratum I morphology. Observe that these suffixes also fail to
undergo Vowel Assimilation in the examples in (40).
As e xpected, com pounds, w hich never show Vow el
Assim ilation, are also totally excluded from Palatalization:

(4 1 ) /b e g i- lu s e / [begiluse] 'long-eyes (i.e. curious)'

/ a r d i- t a ld e / [a rb ita ld e ] 'flock of sheep'
/ t S e r i - t e e i / [tS e rite e y ] 'pig farm '

A problem atic case for our bi-stratal conception of the

lexicon where stratum I includes the morphological operation of
inflection and the phonological rules of Vowel Assim ilation and
Palatalization is presented by the locative inflectional markers.
In A rbizu, the ablative /-ti/ does not undergo P alatalization
(e.g. [m enditi] 'from the m ountain,' *[m endit'i]). This fact could
be captured by postulating that exceptionally this inflectional
 84

affix is added at stratum II. A more difficult case is presented

by the locative/instrum ental pluralizer /-ta -/ which undergoes
Vowel Assim ilation but does not undergo Palatalization:

(42) 'w a ll' 'm ountain'

/p a a te / /m e n d i/
[p a a te -ta s ] [m e n d i-te s ] instrum ental pi
[p a a te -ta n ] [m e n d i-te n ] inesive pi
[p a a te -ta ti] [m e n d i-te ti] ablative pi
[p a a te -ta k o ] [m e n d i-te k o ] g e n itiv e -lo c a tiv e pi

The forms in the right column show that /-ta-/ undergoes

Vowel Assim ilation but falls to be palatalized. Since otherwise
Palatalization and Vowel Assim ilation present the exact same
distribution, the most adequate way to capture the facts in (42)
is to mark /-ta-/ as an exception for the rule of Palatalization.
Given the fact that the pluralizer /-ta -/ and the ablative /-ti/
are the only locative markers that present possible targets for
P alatalization, the generalization that arises is th at locative
markers are marked as exceptions to the rule of Palatalization.
A n o th e r case of seem ingly Irregular e x c e p tio n a lity to
P a la ta liz a tio n Is p resented by the hom ophonous g e n itive
indefinite and superlative (or "excessive") forms:
 85

(43) [p atS iina ] 'the one of Patxi'

[m endiina] 'the one of mountains'
[b e riin a ] 'the newest one'

The explanation of this exceptionality seems to be of a

ve ry d iffe re n t nature from the e x c e p tio n a lity of loca tive
markers. As will be argued below when we consider Consonant
Insertion, a stratum I rule for which these m arkers also
c o n s titu te e x c e p tio n s , the g e n itiv e in d e fin ite and the
superlative have an underlyingly unspecified vowel /-V n/ (see
also chapter 1, section 1.2.3). The surface features of this vowel
are obtained by spreading from the vowel to its left. Thus from
/m endi-V n-a/ we obtain [m endiina] 'the one of m ountains' and
from /e ^k u -V n -a / [e^kuuna] 'the one of hands.' Sim ilarly with
the superlative: /b e rd e -V n -a / [b e rb e e n a ] 'the greenest one';
/alu-V n-a/ [aluuna] 'the most stupid one.' To account for the
absence of Palatalization with these two homophone suffixes we
can sim ply assume that the rule of spreading that gives the
u n s p e c ifie d vo w e l its fe a tu re s is o rd e re d la te r than
P a la ta liz a tio n .
An alternative that seem s less preferable w ould be to
argue that a long /ii/ cannot trigger Palatalization. This result
could be form ally attained by interpreting association lines
e xh a u stiv e ly in the s tru c tu ra l d escrip tio n of the rule of
Palatalization (of. Hayes (1986)).
 86

3.3. Consonant Insertion

In m any central and w estern d ia le cts a co nso na n tal
se gm en t is inse rted a fte r a high fro n t vow el w hich is
im m ediately follow ed by another vowel. The precise phonetic
realization of this consonant varies. In many Guipuscoan dialects
(e.g. Errenteria, Azpeitia) it is usually a palatal [y] with little
fric tio n . In B iscayan and som e G u ipuscoan d ia le c ts the
re a liza tio ns [1] (e.g Gernika, Bermeo) and [§] (e.g. Markina,
Ondarroa) are found. In some of these dialects the behavior of
the front high vowel is paralleled by the insertion of [b] or [w]
between a high back vowel and another adjacent vowel. In Arbizu
a palatal stop or fricative [j] is inserted after /i/ and [b]
(norm ally realized as a stop in spite of being in intervocalic
position) after / u/.^^
The operation of this rule can be clearly observed in
inflectional form s and also with the cliticizing dem onstratives:

(44) / b u u / /m e n d i/
[buuba] 'the head' [m e n d ija ] 'the mountain'
[buubek] 'the heads' [m e n d ije k] 'the mountains'

14 Non-epenthetic intervocalic /b / is realized as a fricative [b] in Arbizu, as

in other Basque dialects (see chapter 4). There is thus a contrast between the
nonepenthetic intervocalic [b] of words such as [a la b a] 'daughter' or [kuba]
'Cuba' and the intervocalic [b] that results from Insertion after /u / In a word
such as /e^ ku -a/ [e^kuba] 'the hand.' I will assume that inserted [b] is assigned
its [-cent] value by the same type of rules that assign all other consonantal
features to these epenthetic segments, and not by the rules responsible for the
assignment of continuancy to underlying voiced obstruents.
 87

[buuben] 'in the head' [m en d lje n] 'in the mountain'

[buubaw] 'this head' [m e n d ija w ] 'this m ountain'

C onsonant Insertion can be understood as a sort of

gem ination process in which a skeletal slot is inserted in an
empty onset position after a high vowel and associated with the
features of the high vowel to its left:

(45)

a
/ \ /\
O R O R
/ ? ?
N N
I ' 7
X X X : X
\ I
R 0 I o
? ?
m 1 i d I a

Being in onset position, the [I] or [u] thus created will be

hardened, and dialect-specific redundancy rules w ill apply to
give the different consonantal segments [y], [j], [1] ,[S] and [w],
[b], [b] that obtain in different dialects (cf. Kamprath (1986)).
Secondary high vowels created by Mid Vowel Raising do not
trig g e r C onsonant Insertion. Thus we have, fo r instance,
/e r le - a k / [e rlie k ] 'the bees' (not *[e rlije k ]) and /b a é o - a k /
[b a ^ u e k ] 'the forests' (not *[b a ^ub ek]). The rule ordering in
 88

Arbizu is thus as indicated in (46). Only this order of the rules

will produce the correct results:

(46 ) 1. Consonant Insertion

2. Mid Vowel Raising
3. Vowel Assim ilation

From w hat we have seen about other rules, we would

expect this rule to also apply m orphem e-internally and not to
apply in com pounding and derivation. These expectations are
confirmed by the data. The examples in (47a) show that the rule
applies morpheme-internally. The examples in (47b) show that
the rule does not apply in com pounding and derivation. The
d e riva tio n a l su ffix /- ta r / 'native,' which, as we saw, did not
undergo either Palatalization or Vowel A ssim ilation, presents
an allomorph /-a r/ after most place names ending in a vowel. As
the exam ples show, this suffix also fa ils to cause Vowel
Insertion to apply:

(47)a. [ f a m ilija ] 'fa m ily ' Sp. fam ilia

[k o n ]fije n ts a ]'c o n fid e n c e 'S p . confianza
^[ero^aayjo] 'rosary' Sp. rosario
[b a lijo ] 'value' St. B. balio, Sp. valia
[b ijo ts ] 'heart' St. B. bihotz
 89

[b ije r ] 'to m o rro w ' St. B. bihar

[e r ijo ts a ] 'death' St. B. eriotza
b. /a r b is u - a r / [arfeisuar] 'native of Arbizu'
/d o n o ë ti- a r / [d o n o ^tia r] 'native of Donosti'
/m e n d i-a id e / [m endialde] 'm o u n ta in -s id e ]
/ ir u s k i- a ld e / [iru s k ia ld e ] 's u n -s id e '
/b u u -a u n d i/ [buuawndi] 'big-headed'

These facts thus confirm the structure of the lexicon that

we have been assuming in Basque: there is first stratum that
co m prises the dom ain of the lexical entry and inflectional
m orphology. In this stratum rules such as Vowel Assim ilation,
P a latalization and C onsonant Insertion apply. In a second
stratum , where these rules do not operate, the m orphological
o p e ra tio n of com pounding ta ke s place. The b eh avio r of
derivational affixes is less uniform . In Arbizu they are quite
generally excluded from undergoing stratum I rules, showing
that they are attached at stratum II. In Baztan, on the other
hand, derivational affixes are split in two groups.
A group of affixes in the treatm ent of which dialectal
v a ria tio n o bta ins is the a d je c tiv a l a ffixe s. In A rbizu the
com parative /-oo/ (St. B. -a a o ) 'm ore' and the superlative
absolute (or "excessive") /-ei/ (St. B. -e g i) 'too' are stratum I
affixes, as their behavior with respect to Consonant Insertion
shows. As already mentioned with respect to Palatalization, the
 90

superlative relative (/-en-/ in Standard Basque) 'the most' has

peculiar behavior. Consider the examples in (48):

(48) plain co m p a ra tiv e sup. abs. super, rel.

[askar] [askaroo] [askarey] [askarena] 'quick'
[bero] [beruoo] [beruey] [beroona] 'hot'
[berée] [b er^ioo ] [berdiey] [beréeena] 'green'
[b e ri] [b e rijo o ] [b e rije y ] [beriina] 'new'
[alu] [aluboo] [alubey] [aluuna] 'stupid'

As the exam ples show, the com parative and superlative

absolute trigger both Mid Vowel Raising with stems ending in a
mid vowel and Consonant Insertion with stems ending in a high
vowel. The superlative relative, on the other hand, does not
trigger either of these two rules (and, as the example [b e riin a ]
shows, does not undergo Palatalization either). One way to
interpret the surface superlative relative form s w ould be to
assum e th a t the superlative relative m arker is underlyingly
/_ e n -/, w hich is the way it surfaces w ith co n so n a n t-fin a l
stems, and to explain its failure to undergo stratum I rules by
stipulating that this suffix is actually attached at stratum II. To
obtain the co rrect surface form s with vow el-final stem s we
would require a rule that in a sequence of two vowels spreads
the features of the first vowel rightwards, giving a long vowel,
as in (49):
 91

(49) Total Vowel Assim ilation

This, however, is not a general stratum II rule. It does not

apply in com pounding or other derivational affixes. The only
affixes that show this behavior are the superlative relative and
the genitive indefinite, which, as we saw, produced homophonous
surface forms. We could assign these two affixes to their own
stratum and let the rule of Total Vowel Assim ilation apply only
in this special stratum. A more economical solution is to assume
that the superlative relative suffix (and the genitive indefinite)
are attached at stratum I together with all other inflectional
affixes but that they fail to undergo certain rules and cause the
application of a different rule because of their structure. The
evidence is all in favor of assigning an initial empty segment to
these affixes. This empty segment will receive its surface value
by spreading if there is a vowel to its left:

(50)Vowel Spread

R
y
O "''
 92

If the segment to its left is a consonant, on the other hand.

Vowel Spread w ill be blocked and the em pty segm ent w ill
surface as [e], which is the default vowel.
C onsider the derivation of the genitive plural [b e rije n ] 'of
the new ones' on the one hand, and the homophonous genitive
Indefinite and superlative absolute [b e riin ] 'of new ones/the
newest' on the other. We must assume that the genitive plural
marker is /-en / and that the genitive indefinite and superlative
relative are /-X n/ (where X is a tim e-slot devoid of feature
co nte n t). For com parison the derivatio n of g e n itive plural
/a s k a r - e n / and genitive ind e finite /a bso lutive relative /a s k a r-
Xn/ is also given. In this latter example the surface forms are
identical to each other:

(51) /beri-en/ /b e ri-X n / /a s k a r-e n / /askar-X n/

Cons. Insertion berijen (n.a.) (n.a) (n.a.)
Vowel Spread (n.a.) beriin (n.a.) (n.a.)
Default (n.a.) (n.a.) (n.a.) askaren
[berijen] [beriin] [askaren] [askaren]

G iven this analysis, su p e rla tive re la tive and g en itive

indefinite forms can be taken to be attached at stratum I like all
other inflectional affixes.
 93

4. Ondarroa
4.1. Vowel Assim ilation
The dialect spoken in Ondarroa also presents a rule raising
the low vowel to [e] after a high vowel or glide, as the
comparison of the pairs of examples in (52a,b, c) shows:

(5 2 )a ./g i$ o n -a / [giSona] 'the man'

/la g u n - a / [lagune] 'the friend'
b./pelota-ka/ [pelotaka] 'throwing a ball'
/a r i- k a / [arike] 'throwing stones'
c. /bat-na/ [bana] 'one by one'
/b i- n a / [bine] 'two by two'

B efore a na lyzin g the m o rph o lo g ica l and p ho no log ical

conditions for the application of Vowel Assim ilation in Ondarroa
we m ust b rie fly exam ine o the r rules a p p lyin g to vow el
sequences.

4.1.1. The treatm ent of vowel sequences

Consider the singular and absolutive forms of items ending
in a vowel:

(53) /a la b a -a / [alaèi] 'the daughter'

/a la b a -a k/ [ala&ak] 'the daughters'
/etS e - a / [etSi] 'the house'
 94

/e t§ e -a k / [etSik] 'the houses'

/b a s o - a / [basu] 'the forest'
/b a s o - a k / [basuk] 'the forests'
/m e n d i- a / [mendiSe] 'the mountain'
/m e n d i- a k / [mendiSak] 'the mountains'
/b u ru -a / [buru] 'the head'
/b u r u - a k / [buruk] 'the heads'

Stems ending in /-a/ give rise to radically different forms

in the singular and the plural. The different results are caused
by a rule, general to all Biscayan dialects, that raises the final
vowel of stems ending in /-a/when immediately followed by the
singular determiners, and only in this case:

(54) Singular Low Vowel Raising

/a/ > [-low] / sg articles

Rule (54) creates sin g u la r abso lutive form s such as

[neskea] form /neska-a/ 'the g irl,' and [alabea] 'the daughter'
from /a la b a -a /, w hich are final in some Biscayan d ialects
(Arratia, Orozko, etc. Cf. Angulo et. al., 1983) and contrast with
the corresponding plural absolutive [neskak] 'the girls,' [alabàk],
'the daughters,' with simple reduction of the two low vowels to
one by Stem -Final Low Vowel Deletion, as in non-Biscayan
dialects.
 95

In Ondarroa, the forms /neskea, /alabea/ are not final; but,

from that point on are phonologically derived exactly as /e/-
final stems, such as /etSe-a/, as we will see immediately.
It seem s that in the first decades of this century, the
Ondarroa dialects presented forms with long high vowels such as
/b uru -a / [buruu], /b aso -a/ [basuu] and /etS e-a/ [etSii]. Azkue
(1923, 23) te lls us th a t in o rd er to make fun of the
pronunciation of Ondarroans, their neighbors to the west used
the formula: O n da ruu . e ri beruu - Bai ereten eauuniin 'Ondarroa,
hot town. Yes, when it was burning' ; which in Literary Biscayan
(that is, in a dialect where underlying vowel qualities are kept
in these sequences ) would be O n d a ro a . e ri beroa - Bai e re te n
eaoanean. Nowadays, those long vowels have been reduced to
norm al l e n g t h . 15 We would then have the follow ing ordered
rules: Singular Low Vowel Raising —> Mid Vowel Raising —>
Total Vowel Assim ilation —> Vowel Shortening:

(55) /a la b a - a / /e tS e - a / /b u ru -a /
Sing Low V R e (n.a.) (n.a.)
Mid V Raising i i (n.a.)
Total V-Ass. ii ii uu
V-Shortening 0 0 0
[alabi] [e tS i] [buru]

15 Som e other Biscayan dialects still present long vowels in these cases (cf.
deRijk (1 9 6 9 )). _________________________________________________________
 96

The d erivatio n can be s im p lifie d , since th ere is no

synchronic evidence for long vowels. Instead of a rule creating
long vowels and another rule shortening them, we can assume a
rule deleting the second vowel in a sequence, ordered after Mid
Vowel Raising; e.g.; /et§e-a/ > /et$ia/ > [e tS i]:

(56) Vowel Deletion

V -> 0 / V 16

The effects of Vowel Assimilation can be observed only in

stems ending in a consonant (e.g. /egun-a/ [egune] 'the day') or in
a high front vowel (e.g. /m endi-a/ [m endiée] 'the mountain'). In
this second case, a rule inserting [S] bleeds the application of
Vowel Deletion. The low vowel of the determ iner can and does
then undergo Vowel Assimilation."*?
Given the complementary application of Vowel Deletion and
Vowel A ssim ila tio n , in O ndarroa there is no evidence fo r
o rd e rin g Vow el A s s im ila tio n w ith resp ect to o th e r rules
affecting vowels. The crucial ordering requirem ents that exist
in Ondarroa are expressed in (57):

15 Notice that In this case it is the second vowel in the sequence that deletes,
unlike in the rule of Stem -Final Low Vowel Deletion, where the first vowel in
the sequence deletes, as shown by forms such as /alaba-ok/ [alabok].
1 ^ The fact that the plural [mendiSak] does not undergo Vow el Assimilation
will be considered below.
 97

(57) Singular Raising

[S ]-ln s e rtio n
Stem -Final Low Deletion
Mid Vowel Raising
Vowel Deletion
unordered: Vowel Assimilation

Singular Raising must precede Vowel Deletion, since were

Vowel Deletion to apply to an underlying form such as /alaba-a/,
we would obtain [alaba]. The rules of Singular Raising and Mid
Vowel Raising that give the final form [a la b i] 'the daughter'
would be bled and we would obtain the incorrect *[alaba] as
surface form.
[§]-lnsertion must precede Mid Vowel Raising because [S] is
only inserted after underlying /i/ and not after interm ediate [i]
resulting from Mid Vowel Raising.
[S]-lnsertion must precede Vowel Deletion, since it is the
p resence of the e p e n th e tic co n s o n a n t w hich b locks the
a p p lica tio n of Vowel Deletion to stem s ending in /i/, thus
accounting for the difference in the treatm ent of the underlying
vowel sequences in form s such as /m endi-a/ [m endiée], where
Vowel Deletion does not apply, and /etSe-a/ [etSi], where Vowel
Deletion applies.
That S tem -F inal Low Vowel D eletion precedes Vowel
Deletion is shown by forms such as /alaba-ok/, which could be
 98

potential inputs for either of the two deletion rules. The fact
that it is the first of the two vowels in the sequences that
deletes, giving [alabok] 'these daughters,' shows that it is the
more sp e cific rule of S tem -Final Low Vowel D eletion that
applies. Both of these rules being vowel deletion rules, they are
intrinsically ordered. The more specific rule of Stem -Final Low
Vowel Deletion must be ordered before the more general rule of
Vowel Deletion.
Mid Vowel Raising must precede Vowel Deletion to get a
chance to apply, since its trig g e r is elim inated by Vowel
Deletion. To obtain [etSi] from /etS e-a/ we must go through
interm ediate [etëia].

4.1.1. Domain of application of Vowel Assimilation in Ondarroa

In the dialect of Ondarroa, the morphological conditions for
the application of the rule of Vowel Assim ilation are much more
re s tric tiv e than in Baztan. As fo r a d d itio n a l p ho no log ical
conditions, they are in a sense the opposite of those needed in
Arbizu.
To s ta rt w ith, all co nte xts w here Vowel A ssim ila tion
applies are m orphologically created. M orphem e-internally there
is no constraint against high vow el-low vowel sequences in
adjacent syllables. Unlike what we saw was the case in Baztan
and Arbizu, in Ondarroa Basque these sequences are not at all
rare. Some examples are give in (58):
 99

/b ia r / [b iS a r] 'tom orrow'
/b is a r / [b isa r] 'beard'
/ in d a r / [in ja r ] 'strength'
/u s a in / [usayh] 'smell'
/ s ik a - tu / [sikatu] 'to dry'
/tS i m is t a / [tS im ista] 'lig h tn in g '
/m u g a / [m uga] 'limit'
/ i k a s - i/ [ikasi] 'to learn'

By comparing the examples in (58) with those in (52) we

can conclude that the assimilation rule in Ondarroa is subject
to the following condition:
A. Vowel A ssim ilation only applies in derived e nviron ­
ments
From this restriction of the rule to derived environm ents
we can directly deduce that this is a lexical rule (cf. Kiparsky,
1985).
As in Baztan and Arbizu, the application of the rule across
a com p ou n d-bo u nd ary is to ta lly excluded, eg. /b u ru -a u n d i/
[buruaw ndi] 'big-headed' not *[buruew ndi]. O nly suffixes and
clitics can in fact be affected by the rule, but not all suffixes
and clitics either. Observe the absolutive singular and plural
forms in (59). As the examples in (a) show, the singular article
/-a / undergoes the rule, but the plural article /-ak/ does not.
This is thus exactly the opposite of the situation in the Arbizu
 100

dialect shown in (30). The exam ples in (b) do not meet the
minimal phonological requirements for the rule to apply and are
given for comparison (i.e. to show that the singular and plural
articles are in fact /-a/ and /-ak/ underlyingly) •

(59) stem absol. sg. abs. pi.

a. / u r/ [ure] [uràk]"*® 'w a te r'
/la g u n / [lagune] [lagunàk] 'frie n d '
/m u t i l / [mutl/Ce] [muti/Càk] 'boy'
/tS a k u r / [tSakure] [tS akuràk] 'dog'
/m e n d i/ [mendiée] [mendiSàk] 'm ountain'

b. /g iS o n / [giSona] [giSonàk] 'man'

/p la te r / [p la te ra ] [p la te rà k ] 'd is h '
/a r/ [ara] [aràk] 'w o rm '

The fa c t th a t the p lu ral a rtic le d oes not undergo

assim ilation does not constitute an exceptional behavior of this
suffix. In fact, only m onosyllabic suffixes ending in /a / (i.e.
/- C o a/ suffixes) show the effects of assim ilation. That is, the
target of the rule must be morpheme-final. Inflectional suffixes

A grave accent indicates a drop in pitch in that and following syllables in

the word (see Chapter 6). Unless marking of suprasegm ental features seem s
relevant for the clarification of exam ples, as is the case in this section,
suprasegm ental information will be omitted in the phonetic representations,
with the exception of Chapter 6, where suprasegmental patterns are studied in
d etail.
 101

that do not conform to this phonological shape do not undergo

the rule. Thus, the prelative /-ts a t/ does not show vocalic
alternation, e.g. [laguntsat] 'as a friend,' not *[laguntset]. Neither
do the dative suffixes singular or plural /-ai/; e.g. [lagunay] 'to
the friend,' not *[laguney], and [lagunay] 'to the friends,' not
*[la g u n è y ]. N either is there alternation with locative markers
which are disyllabic or end in a segm ent other than /a/; e.g.
/to k i-ta n / [tokitSan] 'in place,' not *[tokitS en].
The d e te rm in e rs /b a t/ 'o n e /a ' and /a u / 'th is ,' w hich
assim ilate in Baztan, are also excluded from the targets of the
rule in Ondarroa, although the application of a rule such as [$]-
Insertion shows that /a u / is a c litic also in Ondarroa. The
difference in the way these determ iners are affected by Vowel
Assim ilation in the two dialects does not depend on their status
as cliticizing elem ents, but rather on the restriction of Vowel
Assim ilation in Ondarroa to /-C oa / suffixes; to suffixes where
the low vowel is final:

(60) /lagun bat/ [lagumbat] 'a friend'

/mendi bat/ [mendibat] 'a mountain'
/lagun au] [lagunaw] 'this friend'
/mendi au/ [mendiéaw] 'this mountain'
 102

On the other hand, the adlative /-ra/, which conform s to

the s tru c tu ra l re q u ire m e n ts for the a p p lic a tio n of Vowel
Assim ilation, does indeed undergo the rule:

(61) a. /bilbo-ra/ [bilboa] 'to Bilbao'

/etë e-ra/ [et^ea] 'to the house'
b. /ondaru-ra/ [ondarure] 'to Ondarroa'
/mendi-ra/ [mendire] 'to the mountain'

The structural restriction m entioned is totally general and

also holds of ve rb al s u ffix e s , d e riv a tio n a l s u ffix e s , and
c litic iz in g elem ents. The exam ples in (62) show th a t the
derivational adverbializer /-ka / 'by means of, throw ing, hitting
w ith ,' which conform s to the /-C o a / p attern, does indeed
undergo the rule (62a); as do the distributive /-na/ (62b), the
co m p le m e n tize r /-n a / 'th a t' (62c), the clitic iz in g conjunction
/ta / (/(e)ta/) 'and' (62d), the intransitive auxiliary /da/ '3 sg
non-past indicative' (62e) and the participial /-ta/(62f). In (63)
I show that derivational suffixes in which the low vowel is not
final, such as /-g a ri/ 'worthy of, able to,' /-lai/ 'maker, player'
and /sale/ 'fond o f do not show vocalic alternation; and this is
also true of the com plem entizer /-lako/ 'because':

(62) a. /a ri-k a / [arike] 'throwing stones' (Sp. 'a pedradas')

/u k a b il-k a / [ukabilke] 'hitting with fists,' of.
 103

/pelota-ka/ [pelotaka] 'throwing a ball'

b. /bi-na/ [bine] 'two for each'
/iru-na/ [irune] 'three for each,' cf.
/bat-na/ [bana] 'one for each'
c. /ba-du-na/ [barune] 'that he goes,'"*® cf.
/daue-na/ [dabena] 'that they have it'
d. /p a ts I ta koldo/ [patSite koldo] 'Patxi and Koldo,' cf.
/koldo ta pats i/ [koldota patSi] 'Koldo and Patxi'
e. /p a tS i da/ [patSire] 'It is Patxi'
/baso-a da/ [basure] 'it is the forest'
/obe-a da/ [obire] 'it is better'
/xun da/ [xunde] 'he has gone'
/e to ri da/ [etorire] 'he has come,' cf.
/koldo da/ [koldora] 'it is Koldo'
/amen da/ [amenda] 'here it is'
f. /g e rta -tu -ta / [g erta tu te ] 'having happened'
/ikus-i-ta/ [ikusitSe] 'having seen'

[du] 'he goes' historically derives from /doa/, found in other dialects and
the Standard language, as the name of the town in the dialect [ondaru] derives
from /o n d a ro a /. The reduction of underlying /o a / to [u] takes effect by means
of synchronically motivated rules (cf. section 1), as can be seen, for exam ple,
in /b a s o + a / [basu]. In the absence of clear synchronic evidence, given by
alternations within the dialect to justify representing [u] as /o a / in particular
cases , such as in [ondaru] or [du], I will adopt phonological representations
closer to the phonetic substance.
As Rotaetxe (1979 , 578 ) points out, Vowel Assimilation only affects the
conjunction /(e )ta/ in rapid speech. In careful speech, the conjunction does not
cliticize, but constitutes a separate word, and appears as [ta] or [eta] even
after words ending in a high vowel.
 104

/egin-ta/ [eyn^e] 'having made,* cf.

/emon-ta/ [emonda] 'having given'

(63) /ik u s - g a r i/ [ik u z g a ri] 'w orth-seeing'

/tS istu-lai/ [tS istulay] 'flo u tis t'
/m endi-sale/ [m endisale] 'm ountaineer'
/du-lako/ [dulako] 'because he goes'

We can thus state a second condition on the application of

Vow el A ssim ila tio n :
B. The ta rg e t of Vow el A ssim ila tio n m ust be right-
adjacent to a morphological bracket: / ]
We m ust now dete rm in e w hat kind of m orph o lo g ica l
boundary this condition makes reference to. Or, in the terms of
Lexical Phonology, at what point in the derivation of the item
this bracket must be present.
It can be shown that those suffixes which undergo
assimilation in the examples above fail to undergo the rule when
follow ed by another suffix. We can reach this conclusion by
com paring absolutive singular form s, which bear the singular
m arker /-a / with e rgative sin g u la r form s, w hich ca rry the
suffix /-k / after the singular /-a/:
 105

(64) absolutive sg. ergative sg

[ure] [urak]
[lagune] [lagunak]
[m u ti/e ] [m u ti/a k ]
[t§ a k u re ] [t$ aku ra k ]
[m endiée] [mendiSak]

It is, by the way, possible to show that the ergative

singular form s are created by adding an ergative /-k/ to items
bearing the singular /-a/, and not by adding a d ifferent affix
/-a k / to the bear stem, as in the plural absolutive. Clear
evidence is provided by /a/-final stems. It is to the form s that
have undergone Singular Raising (rule (53)) that the ergative
marker is added. Thus, there is a contrast with /a/-final stems
betw een e rg a tiv e s in g u la r form s such as /n e s k a -a -k / >
/neskeak/ (by Singular Raising (53)) > /neskiak/ (by Mid Vowel
Raising (4)) > [neskik] (by Vowel Deletion (55)) 'the girl (erg),'
/a lab a-a -k/ [alabik] 'the daughter (erg)' and the plural absolutive
form s /n eska -ak/ [neskak] 'the g irl,' /a la b a -a k/ [a la b à k] 'the
daughters.' If the stem has a different ending, however, the
contrast is only at the suprasegmental level, e.g. [giëonak] 'the
man (erg) vs. [giSonàk] 'the men (abs, erg).'
It must be pointed out that the rule of Singular Raising
also operates before /au/ 'this' and /ori/ 'that,' as the examples
in (65) show. The operation of the rule is apparent in the
 106

presence of a stem -final [i] from /a/. There are no other

triggering elements .In Ondarroa, as in many Biscayan varieties,
dem onstratives are both preposed to the noun and in the
determ iner position (cf. section 1). Postposed dem onstratives
are suffixes and not mere clitics as in non-Biscayan dialects. In
fact /a/ is simply a third degree dem onstrative:

(65 ) /au neska-au/ [aw neskiaw] 'this g irl'

/o ri n e s k a -o ri/[o ri neskioi] 'th a t g irl'
/a neska-a/ [a neski] 'that girl (over there)'
/n e s k a - a / [neski] 'the girl'

S ingular Low Vowel Raising does not operate in the

corresponding plural form s, w here proposed and postposed
dem onstratives have radically different forms. Instead, the final
vowel of the stem is deleted:

(66 ) /oneik neska-ok/ [onèyk neskok] 'these girls'

/o re ik neska-ok/ [orèyk neskok] 'those girls'
/areik neska-ak/ [arèyk neskak] 'those girls (over there)'
/n e s k a - a k / [neskak] 'the girls'

The situation illustrated in (65) for the singular m arker

follow ed by the ergative marker is generalizable to any other
case in which a potential target of assim ilation is follow ed by
 107

another suffix. This is shown for the singular article /-a / in

(67a), for the adlative /-ra / in (67b), for the distributive /-na/
in (67c) and for the adverbializer /-ka/ in (67d) when followed
by other suffixes:

(6 7 ) a. / m u t i l - a / [muti/Ce] 'the boy'

/m u til-a -k/ [muti/Cak] 'the boy (erg.)'
/m u til-a -g a i-tik / [m u tix a g a y tS ik ] 'because of the boy'
b. /o n d a r u - r a / [ondarure] 'to Ondarroa’
/o n d a r u - r a - k o / [ondarurako] 'bound for
Ondarroa'
c. / b i - n a / [bine] 'two for each'
/bi-na-ka/ [bihaka] 'two by two'
d. / a r i - k a / [a rik e ] 'throw ing stones'
/ari-ka-da/ [a rik a ra ] 'throwing of a stone'

In all the examples in (67) the two suffixes in contact are

both inflectional suffixes (in (a) and (b)) or both derivational
suffixes (in (c) and (d)). Given the constrained character of the
rule, it is not easy to find examples which contain a derivational
su ffix follow ed by an infle ctio na l suffix. Assum ing th at the
participial /-ta / is a derivational suffix, it can offer the type of
evidence that we are looking for, since participial form s can
then take nom inal inflection. The result is that the rule of
a ssim ila tio n is also blocked w hen the p a rtic ip ia l suffix is
 108

follow ed by an inflectional suffix. Thus, for instance, when the

ergative /-k/ is added to /ikusi-ta/ [ikusitSe] 'seen,' we obtain
[ik u s itS a k ], not *[ikusitS ek]. Another te st case would be the
dim inutive /-Ska/ found in Baztan. We would predict that a form
such as /m endi-S ka/ [m endiSke] 'little m ountain' would fail to
show assim ilation when inflectional suffixes are added, e.g.
[m e n d iS k a tik ], not *[m endiS ketik] 'from the little m ountain.'
U nfortunately this dim inutive suffix is not used in Ondarroa.
From the evidence that we do have, however, it seems that we
can safely conclude that for the low vowel of a suffix to be an
eligible target for vowel assim ilation it is not enough for it to
be w ord-final at some stage in the derivation; but, rather, it
must be word final after all suffixation has taken place. If we
equate suffixation with the erasure of boundaries between stem
and suffix, we can qualify condition B above by stating that this
condition m ust be met "at the output of all m orphological
operations of suffixation." The operation of the phonological
rule of Vowel Assim ilation (VA) with respect to the morphology
can be represented as in (68):
 109

(68) Stratum I
Morphology Phonology
suffix a
suffix b

suffix z ■>Vowel Assim ilation

This derivation is illustrated with two exam ples in (69),

[tSakure] 'the dog' and [tSaku rak] 'the dog (erg)':

(69) [tS akur] a] [tSakur] a] k]

dog sg dog sg erg
S uffixational Stratum
su ffixa tio n of /-a / [tSakura] [tS akura]
su ffixa tio n of /-k / [tS aku ra k ]
output of Morphology [tS akura] [tS akurak]
Vow el A ssim ila tio n [t^ aku re] S.D. not met

In [t§ a k u r a k ] the s tru c tu ra l d e s c rip tio n of V ow el

Assim ilation is not met because the possible /a / target is not
adjacent to a morphological boundary at the point when the rule
applies.
I am assuming that lexical phonological rules form blocks
or com ponents th at are associated with p a rticu la r m orpho­
logical components. Each pair of phonological and morphological
 110

com ponents form s a m orpho-phonological stratum . Halle and

Mohanan (1985) argue that m orpho-phonological strata can be
cyclic, if the phonological rules in the stratum apply after
every m orphological operation, or noncyclic, if phonological
rules apply only once, after all m orphological operations in the
com ponent. The w ay our rule of Vowel A ssim ilation applies
indicates that all suffixation is done in a single, noncyclic
stratum in Ondarroa Basque.
A question that arises is how exactly the blocking of the
rule in a non-derived form such as /m uga/ is accomplished. One
way to ensure this result would be to include two boundaries in
the structural description of the rule, before and after the
target. The rule w ould not apply to /m ug a / or /t§ i m is t a /
because there is no boundary between trigger and target, but
would apply to /[t§akuT]a]/. The alternative would be to treat the
restriction to derived environm ents as a condition on particular
rules. It is usually assum ed that being restricted to derived
environm ents is a property of cyclic rules (K iparsky (1985),
Halle and Mohanan (1985), Booij and Rubach (1987)). I have
shown, how ever, th at Vowel A ssim ila tion in O ndarroa is a
noncyclic rule restricted to derived environments.
We m ust now ask if c lltlc-a tta ch m e n t also blocks the
application of Vowel A ssim ilation to an otherw ise w ord-final
/a/ in a suffix, and, therefore, if cliticization takes place at the
same time as suffixation. The answer is: no. Clitics do not block
 111

the a p p lica tio n of Vowel A ssim ila tion to S u ffixe s, as the

examples In (70) show:

(70) /lagun-a da/ [lagunera] 'it is the friend'

/mendi-a da/ [m endiSera] 'It is the mountain'

On the other hand, the application of other suffixational

phonological rules creates outputs for Vowel A ssim ilation to
apply to clitics:

(71) /buru-a da/ [burure] 'it is the head'

/baso-a da/ [basure] 'it is the forest’
/etSe-a da/ [e tS ire ] 'it is the house'
/alaba-a da/ [a la b ire ] 'it is the daughter'

All lexical rules a ffe cting su ffixes apply before c litic-

attachm ent. Vowel A ssim ilation m ust thus apply tw ice: once
after suffixation and one after cliticiza tion . This is evidence
th a t c litic g ro u p s are fo rm e d at a la te r sta ge than
m orphologically com plex words, as one would expect. Once a
clitic group is formed by syntactic concatenation, it is sent back
to stratum I of the lexicon, where the rules of this stratum get a
second chance to apply. Consider the derivation of [etSire] 'It is
the house' and [lagunera] 'it is the friend' (irrelevant rules are
ignored):
 112

(72) [et§e] a] [da] [lagun] a] [da]

s u ffix a tio n etSea laguna
Mid Vowel Raising e t^ ia---------------------------------
Vowel Deletion e tS i-----------------------------------
V Assim ilation^"* ............ lagune

c lit ic iz a t io n etS ida laguneda

V A ssim ila tio n etSide .................
[e tS ire ] [lagunera]

We conclude that clitic groups, which are form ed in the

syntax, revert to Stratum I and undergo the rules of this
stratum . A w ord dom ain is thus d efin ed tw ice: a fte r all
suffixation and after cliticization.
There is more evidence for making syntactically form ed
clitic groups revert to the stratum I, instead of adding clitics
directly at stratum I at the same stage that suffixes are added.
This evidence is provided by the rule of Underlying Low Tone
Assignment. As we will see in chapter 6, in Ondarroa underlying
low tones are linked to the rightm ost syllable of the w ord-
domain. This word domain does not include clitics:

Notice that Vowel Deletion and Vowel Assimilation could be also ordered
with each other in the opposite way. The intermediate form /et $ le/ created by
Vowel Assimilation would also give /e tSi/ after Vowel Deletion.
 113

(73) /lagun-àn-a da/ 'it is the one of the friends'

L
Tone Assoc lagunana da

Underlying Tone Association is, thus, a rule that applies

before c litic iz a tio n . We m ust ask w hy the rule of Tone
Association does not apply again when the clitic group reverts
to the lexicon. The answer seems to be in the fact that the
effect of this rule is to associate a free tone. A fter the rule
applies once, there will be no free tone to associate.
In Ondarroa, unlike other dialects seen above, the Vowel
A ssim ila tio n fa cts do not p ro vide e vid e n ce fo r a ssigning
compounding to a different stratum. The rule does not apply to
com pounds, but there is an independent ju stifica tio n for this,
since in compounds the boundary- adjacency requirements on the
ta rg e t of the rule w ould never be met. Two o the r rules.
Palatalization and Flapping, however, clearly show that, as in
other dialects, com pounds are form ed at a d iffe re n t stratum
from both s u ffix a tio n and c litic iz a tio n . The d o m a in of
Palatalization and Flapping in O ndarroa is studied in the next
subsections.
 114

4.2. P alatalization
As w ill be shown in detail in chapter 3, w here the
P alatalization rule is form ulated, in O ndarroa, /!/, /n/, /t/ are
realized as [/C], [h], [tS] after /i/, vowel or glide. Palatalization
also affects the groups /ilt/, /in t/, /ild/, /ind/.
In all dialects having a rule of P alatalization, this rule
applies at the level of the underived lexical item. At this level ,
P a la ta liz a tio n u s u a lly p re s e n ts som e e x c e p tio n s , m ainly
(perhaps exclusively) among recent borrowings. Dialects vary in
the list of exceptions. In dialects that palatalize after syllabic
/i/, /kilo/ 'kilogram ' undergoes Palatalization for some speakers
who pronounce [ki/Co], but for some other speakers this item
constitutes an exception to the Palatalization rule. There are
even exam ples of words tw ice borrow ed from Spanish, that
constitute m inim al pairs for some speakers, w here the older
borrow ing regularly shows palatalization and the more recent
borrowing does not. Iverson and Ohederra (1985) mention [pixoto]
'pilot of a ship' vs. [piloto] 'pilot of a plane'; [makina] 'a lot' vs.
[m akina] 'm achine.' In Ondarroa, there is Palatalization in the
word for 'm achine,' but [kilo] and [kasino] are exceptions. For
exam ples of m orphem e-internal application of Palatalization in
Ondarroa, see chapter 3.
As is true for all Basque dialects. Palatalization does not
apply across word-boundaries, as the examples in (74) show:
 115

(74) [saldi totu] *[saldi tSotu] 'fat horse'

[mendi lusi] *[mendi Xusi] 'long mountain'
[saldi nekatu] *[saldi nekatu] 'tired horse'

There is no palatalization across the members of a compound

e ith e r:

(75)
/begi-luse/ [begiluse] *[begiXuse] ' long-eyes'
/a rd i-ta ld e / [aréitalde] *[aréit§alde] 'sheep-group'
/tS a ri-te g i/ [tS a rite g i] *[tS a ritS e g i] ' pig-sty'
/m util-ne ska k/ [m utilneskak] *[m utixheskak] 'boys and girls'

On the other hand, In both verbal and nominal inflectional

m orph o lo g y P a la ta liz a tio n a p p lie s q u ite re g u la rly , as the
examples in (76) show:

(76 ) / m u t i l - a / [m utixe ] 'the boy'

/m in-a/ [m ine] 'the pain'
/ m e n d i- t ik / [m e n d itS ik] 'from the mountain'
/to k i- ta n / [to k itS a n ] 'in place'
/a la i- tu / [alaytS u] 'to make happy (perf.)'
/b a r i- tu / [b a ritS u ] 'to renew (perf.)'
/a I eg in - t u / [a le g ih ju ] 'to make an effort (perf.)'
/ e g i- t e n / [etSen] 'to make (im perf.)'
 116

A surprising exception to Palatalization is the genitive

suffix /-n/:

(77) / p a t S i- n - a / [p a t^in e ] *[pat^ihe] 'that of Patxi'

/etS e -a -n -a /[e tS in e ] *[etSine] 'that of the house'

This exceptionality is most remarkable, since it holds true

across all dialects for which I have been able to obtain data
(Gernika, Errenteria, Arbizu, Lakuntza). One way to account for
this exceptional behavior would be to assign the genitive suffix
to the Com pounding stratum . Another explanation Is that the
presence of an empty segm ent blocks Palatalization with this
suffix. The standard dialect has /-ren/ as singular genitive. The
examples above are thus P a tx ire n a and e tx e a re n a in Standard
Basque. Perhaps, then, a palatalization blocking empty segment
must also be posited for dialects like Ondarroa.
The application of Palatalization in derivational m orpho­
logy is much more variable across dialects. We saw that in
Arbizu derivational suffixes are quite generally excluded from
the dom ain of P a la ta liza tio n . P a la ta liza tio n in d e riv a tio n a l
morphology is fairly regular in the Ondarroa dialect, on the other
hand:
 117

(77) Palatalization in derivational m orphology

/-ta r/ 'dw eller, native of
[b ase ritS a r] 'fa rm e r'
[m enditS ar] 'm ountain dw eller'
of. [o n d a ru ta r] 'Ondarroan'

/-n a / 'tim e s '

[biha] 'tw ic e '
[saspiha] 'seven times'
[b e d e ra ts ih a ]'n in e tim es'
[s o fts ih a ] 'eight tim es'
cf. [iru n a ] 'three tim es'

/-ta s u n / 'q u a lity '

[a fg itS a su n ] 'c la r it y '
[bardihtS asun] 'sameness, equality'
[xakihtS asun] 'know ledge'
o f. [askatasun] 'fre e do m '

The only example that I have found in the Ondarroa dialect

of a derivational suffix that does not undergo Palatalization is
/-ta ri/, which indicates profession, eg. [a fin ta y ] 'leader' (from
/a g in -tu / 'to give orders'). In section 2.3 we proposed that
exceptional suffixation, which is not affected by rules such as
Palatalization and Vowel Assim ilation, takes place in the same
 118

stratum as Com pounding. Which suffixes are assigned to the

C om pounding stratum , however, seems to be subject to a
certain amount of variation among dialects.
C liticizing elem ents, which include the conjunction /(e)ta /
'and' and a num ber of conjugated verbal form s, undergo
P a latalization in O ndarroa (observe also the a pplication of
Vowel Assim ilation to final /a / in the clitics):

(78) P alatalization of clitics:

/g o ri eta berde-a/ [gofitSe b e ré i] 'red and green'
/saldi eta kato-ak/ [salditSe katuk] 'horses and cats'
/il da/ [ix^e ] '(s/he) has died'
/egin dau/ [eyhJaw] '(s/he) has made (it)'
/m artin da/ [m a rtih je ] '(it) is Martin'

P alatalization is thus a lexical rule w hose dom ain of

a pp lica tio n includes underived lexical item s, su ffixa tio n and
cliticization. Excluded from the domain of the rule are certain
exceptional suffixes. Compounding and the Postlexical stratum.
We thus assign Palatalization to the stratum I of the lexical
phonology. C litic groups, as argued above in relation with
several other processes, are sent back to stratum I.
 119

4.3. Flapping
In the variety of Ondarroa, intervocalic /d/ can optionally be
realized as [r]. There is free va ria tion between these two
allo ph o ne s. This a lte rn a tio n , how ever, is lim ited to certain
le x ic a l d o m a in s, in clu d in g c litic iz a tio n of /d /-in itia l th ird
person conjugated verbal form s; see (79a). On the other hand,
across w o rd -b o u n d a rie s, o utsid e of c litic iz a tio n , and even
across the members of a compound, only [é] and never [r] can be
the surface realization of intervocalic /d/, as shown in (79b):

(7 9 )
a. /bide/ [bire], [bl^e] w ay
/abade/ [abare], [ababe] 'p r ie s t'
/koldo da/ [koldora], [koldoba] 'it is Koldo'

b. /saldi danak/ *[s a ld ira n a k ] 'all the horses'

[saldlbanak]
/saspi domeka/ *[s a s p iro m e k a ] 'seven Sundays'
[saspibom eka]
/e s p a ta -d a n S a ri/*[e s p a ta ra n ts a y ] 'sword dancer'
[espatabantsay]

The exam ples show that in this dialect there is a rule that
o p tio n a lly sonorlzes intervoca lic /d/, w ithin certain dom ains.
This rule can be given a simplified, formulation as In (80):
 120

(80) Flapping
/d/ > [+son] / V_ V (optional)

The process in (80) should probably be understood as the

optional spread of the feature [+sonorant] to intervocalic /d/
from one of the two neighboring vowels; that is, as either (a) or
(b) in the sim plified representations in (81). Cole (1987, 52)
makes the claim that the feature [sonorant] never spreads in any
language; this rule is a counterexample to that claim; ^2

a. V d V b. V b V
I 1 1
0 0 R
1 1 t
1
0 0 0 SL
/

[+son] [+son]

T his is an o ptio na l rule th a t a p p lie s w o rd -in te rn a lly ,

including phonological words created by cliticization, but not in
compounding or across non-reduced word boundaries. The rule
thus has stratum I as its domain of application.

22 Som e Dominican Spanish dialects possess a sim ilar rule, although in

Spanish it applies across the board. For an analysis of the Dominican Spanish
facts that m akes use of hierarchical representations, see N ühez C edeho
(1 9 8 7 ).
 121

4.4. Summary
To sum m arize, we have justifie d an organization of the
lexical m orpho-phonology of O ndarroa Basque in two strata.
S tra tu m I, w h e re ru le s such as V ow el A s s im ila tio n ,
P a la ta liz a tio n and F lapping apply, inclu de s p ra c tic a lly all
suffixation. The syntactic form ation of clitic groups triggers a
second application of stratum I phonology. C om pounds are
formed at stratum II and are not affected by any of these rules.

5. Is stratum II the postlexical stratum ?

All the lexical rules examined so far in this chapter have
excluded compounding from their domain of application. We must
now ask if there are any rules that affect com pounds and
com pound-like suffixation, w ithout applying across the board,
and thus grant the setting of a lexical stratum II. The answer is
that rules of this type do exist. Such a rule is Stop Devoicing. In
Basque, oral stops are devoiced after a voiceless segm ent (in
some dialects the rule affects /b/ and /d/ but not /g/). This rule
applies in all m orphological dom ains: in fle ctio n a l a ffixatio n ,
derivatio na l affixation and c litic iz a tio n , but not across other
boundaries. The rule is form ulated in (82) for dialects devoicing
all oral stops:
 122

(82) Stop Devoicing

Operation: Spread
Argum ent: [-voice]
Direction : Rightwards
Target: [-son], [+voice]

If the trigger of Stop Devoicing is an oral stop it will

itself be deleted by an independent process of Stop Deletion (see
chapter 5). The result of a sequence of two oral stops will then
be a voiceless stop with the point of articulation of the second
stop in the original sequence; for instance, /kb/ will give [p]. As
m en tio n e d , the rule of S top D evoicing o n ly a p p lie s in
m o rp h o lo g ic a l c o n te x ts (both s tra tu m I and s tra tu m II
m orphology), but not p ostle xically across the board. A very
interesting exam ple is given by T xillardegi (1984, 229). He
m entions that the phrase o o itik beh era 'from the top to the
bottom' is always pronounced w ithout devoicing the /b/, but the
com p ou n d c o itik - b e h e r a 'a type of funicular' is pronounced
[guitipera]. Other examples showing the domain of application of
Stop Devoicing are given in (82) (most of them from Txillardegi,
(1980; 1984), M ujika (1978)):

(82) a. inflection:
/gison-ak-gas/ [gisonakas] 'with the men (Bisc.) '
 123

b. derivation:
/b a t-g a r/ [bakar] 'only'
/a rits-d i/ [aristi] 'oak grove'
/bîots-dun/ [biostun] 'courageous (heart-having)'
c. compounding:
/e re t-b id e / [erepide] 'highway (king's road)'^^
/ats-gura/ [askura] 'itch'
/bat-bateko/ [bapateko] 'sudden'
d. c litic iz a tio n :
/onak dira/ [onatira] 'they are good'
/nik besala/ [nipesala] 'like me (erg.)'
/nik be es/ [nipes] 'me (erg.) neither'
e. across the board ( rule does not apply):
/es beste-a/ *[espestea] 'not the other'^4
/ur-es bete/ *[urespete] 'full of w ater'
/peru-k batu ditu/ *[perupatu ditu] 'Peru (erg.)
united them'

The rule of Stop D evoicing thus offers evidence for

distinguishing a stratum II in the phonology different from the
p o s tle x ic a l level w here m o rp h o p h o n o lo g ica l in fo rm a tio n is

23 Erret is the derivational form of e rre g e 'king' (cf. chapter 3).

After a sibilant the rule is optional in morphological contexts (including
cliticization, and com petes with a voicing rule operating in the opposite
direction, e.g. /ik u ^ - g a r i/ [ik u é k a r i], [iku zg ari] 'worth seeing '; /e s da/
[esta], [e zb a ] 'is not.* In non-m orphological contexts, how ever, Devoicing
never applies, as the exam ples in the text indicate.
 124

ignored in the application of rules. At this stratum , where the

m orphological process of com pounding (and some derivational
s u ffix a tio n , d e p e n d in g on the d ia le c t) ta k e s p lace, the
phonological rule of Stop Devoicing (which also applies at
stratum I) w ill apply.
In the next ch a p te r, som e fu rth e r a sp e cts of the
interaction of the phonology with the m orphological operations
of derivation and compounding will be examined.
 125

Chapter 3
Stem Alternations

0. Introduction
There is a set of phonological rules w hose dom ain of
a pp lica tio n cuts right across the m orpho-phonological strata
established in the previous chapter. These are those phonological
rule s w hich are in vo lve d in the fo rm a tio n of a lte rn a te
d erivatio na l stem s of nouns and adjectives. These alternate
stems are used with derivational suffixes and in sub-compounds.
W hereas the p ho no log ical rules of Vowel A ssim ila tio n ,
Palatalization and Consonant Insertion pointed tow ard a basic
d is tin c tio n betw een s u ffix a tio n and co m p o u n d in g as two
different domains for the application of phonological rules in all
Basque dialects studied (although allowing for the exceptional
inclusion of certain suffixes in the domain of com pounding), the
Stem A lternation rules group derivation and subcom pounding
together in one domain with associated phonological rules and
exclude inflection and cocompounding.
I w ill argue that there is no contradiction here. The rules
c re a tin g a lte rn a tin g stem s a p p ly to le xica l e n trie s . The
operations of derivation and subcompounding sim ply create the
context for the insertion of these form s. Derivational stems, I
will argue, are inserted in nonhead position of a complex word.
 126

1. Phonological rules governing stem alternations

Only a limited subset of Basque nouns and adjectives present
an a lte rn a tiv e d e riva tio n a l form w hen bearing d e riv a tio n a l
suffixes or as first member of a subcompound. Traditionally the
form ation of derivational stems has been understood in terms of
a quite large number of processes of addition, suppression and
mutation of phonemes affecting very reduced groups of items
(of. Villasante (1974), Azkue (1923)). It is possible, however, to
extract a small number of generalizations that can account for
the vast m ajority of alternating stems.
A lthough the set of p a rticu la r stem s th at undergo a lte r­
nations varies som ew hat from d ia lect to dialect, the general
p ro c e s s e s and re s tric tio n s fo r the fo rm a tio n of th ese
d e rivatio na l stem s are com m on to all Basque dialects. The
examples in this section will thus be from Standard Basque.
A first group of alternating noun-stem s is given in (1). For
each example, the free nominal form is given in the left column
and a derived or compound form is given in the right column.
Stems are separated from suffixes by a hyphen, as are compound
members (all exam ples are taken from the same dictionary of
the standard language, Kintana et al. (1980)):
 127

(1) free stem derived form

base 'forest' b a s a -ti 'savage'
lore 'flower' lora-tegi 'garden (flower-place)'
etxe 'house' etxa-jabe 'home-owner'
uso 'pidgeon' usa-kaka 'pidgeon guano'
oilo 'hen' oila-sko 'young chicken'
aho 'mouth' aha-men 'mouthful'
erle 'bee' erla-untza 'bee-hive'
hauzo 'neighbor' hauza-tu 'to become a neighbor'
asto 'donkey' asta-keria 'irra tio n a l act'
lasto 'hey' lasta-tegi 'barn (hay-place)'
suge 'snake' suga-mahats 's n a k e -v in e '
arto 'corn' arta-buru 'ear of corn (corn-
head)'

The group of words to which the exam ples in the left

column of (1) belong are all bisyllabic stems ending in a non-
high vowel, which changes to /a / in the d erivational fo rm J
The rule is form ulated in (2) (num ber of syllables m ust be
interpreted e xhaustively):

1 Although ending in a high vowel, katu (kato. in some dialects) also changes
its final vowel into /a / in its derivational form, which is k a ta . E.g. kata-kum e
'kitten,' k a ta -a o r r i 'squirrel (red -cat).'
 128

(2) Lowering

Stem
/ \
a a
I
[-high] — > [+low] (in derivational contexts)

If the basic form has more than two syllables or ends in a

high vowel, the last vowel is lost:

(3) harri 'stone' harr-etxe 'stone-house'

iturri 'fountain' iturr-buru 'source (fountain
head)'
eliza 'church' eliz-koi 'devout'
herri 'country' herr-kide 'compatriot'
itsaso 'sea' itsas-gizon 'seaman'

Alternating stems which do not fall under rule (2), because

they do not meet one or both phonological requirements of being
bisyllabic and ending in a non-high vowel, undergo the final
vowel loss rule illustrated with exam ples in (3) and form ulated
in (4). This rule is an "elsewhere rule" (cf. Kiparsky (1973)) with
respect to (2):

(4) Vowel Truncation

V —> 0 / ___ ] (in derivational contexts)
 129

Some rules operate then on the output of the Truncation

rule. A num ber of the consonants that become final after the
loss of the final vowel of the stem undergo some rule-governed
change.

If the final consonant after Truncation is an oral stop, it

is changed to /t/:

ogi 'bread' ot-apur 'bread-crumb'

begi 'eye' bet-ile 'eye-lash'
errege 'king' errepide
(erret-bide) 'highway (king's road)'
ardi 'sheep' art-ile 'wool (sheep-hair)'
argi 'light' art-izar 'Planet Venus (light-star)
sagu 'mouse' sat-or 'mole'
idi 'ox' it-zain 'ox-driver'
zaldi 'horse' zalt-zain 'horse raiser'
ipurdi 'buttocks' ipurt-ada 'fall onto the buttocks'

This phenomenon finds an explanation once we realize that

/t/ is the only oral stop that is found stem -finally in Basque (in
nouns, adjectives or verbs), although /k/ is found in inflectional
affixes (absolutive plural /-ak/, ergative /-k/ and second person
 130

informal m asculine ergative verbal affix, e.g. d a k ik 'thou (man)

knowest'). The loss of the stem-final vowel by rule (4) places an
oral stop in stem -final position in the exam ples in (5). Stem-
final stops are then neutralized as /t/:

(6) [-cont, -son] - > [t] /___ ]

Rule (6) expresses only part of a w ider generalization.

There is a constraint that restricts the possible consonants in
stem -final position: only coronal consonants may appear stem-
finally. The only consonants found stem -finally are /t, s, é, é, ts,
i é, I, n, r, r/. This generalization can be captured by means
of the rule in (7), which applies to lexical entries:

(7) Stem -Final N eutralization

[+0 0 ns] > [cor] / ___ ]

The coronal /d / is e xcluded from ste m -fin al position,

because in Basque only oral stops in an onset may contrast in
voice. Obstruents in a rime are always underlyingly voiceless.
The rule of Vowel Truncation has the effect in some cases
of creating "illegal" stems containing other types of stem -final
stops (e.g. /og-/ 'bread'). To make the so-created stems conform
to the general principles of the Basque lexicon, the "illegal"
features must be deleted and replaced by the default values.
 131

If, after final-vow el truncation, the final consonant is /r/,

it is changed to /I/: ^
(8) abere 'cattle’ abel-buru 'cattle-head'
gari 'wheat' gal-buru ' shaft (wheat-head)'
afari 'supper' afal-ondo 'after-supper'
aizkora 'hatchet' aizkol-begi 'hatchet-eye'
atari 'hall' ^ atal-aurre 'doorway (hall-front)'
euskara 'Basque Ig.' euskal liburu 'Basque book'
erdara 'foreign Ig.' erdal-tzale 'xenophile'
m erkatari 'm erchant' m erkatal-untzi 'com m ercial ship'
meatzari 'miner' meatzal problemak 'miners'
problem s'

The consonantal change in (8) reflects a historical change

in the opposite direction, which changed i to r intervocalically:

(9) I> r / V V

This h is to ric a l ch an ge is re a d ily o b s e rv a b le in the

treatm ent of Latin loans (the examples are from Mujika (1982)):

2 1 know of one exception: buru head,' whose derivational form Is bur- and
not biit-.
3 Three of the left column forms are. In their turn, derived from other
roots: a ta ri 'hall' from aiû , with Final Vowel Lowering, 'door,' m e rkatari
m erchant' from m e r k a tu m arket,' with Truncation, and m e a tz a ri 'miner'
from m eatz 'mine.'
 132

(10) angélum > aingeru 'angel'

coelum > zeru 'sky'
colum > goru 's p in n in g -w h e e l'
voluntatem > borondate 'will'
gulam > gura 'd e s ire '

In contem porary Basque, however, rule (9) not only is not

productive, but does not even reflect a lexical regularity, since
w ords w ith in te rvo ca lic [I] are quite com m on. The lexical
generalization that holds is, rather, that stems ending in /VrV/,
not fa llin g into the ca teg o ry to w hich rule (2) (Low ering)
applies, change their /r/ into [I] in their derivational form. This
is thus a case of historical rule inversion (of. Vennem ann
(1972): see also Hyman (1975,176-178)). This rule, like the
other rules discussed above, can still be applied to new words
w hich fu lfill the necessary requirem ents, even though its
application is by no means obligatory.
A small number of nouns form their derivational form in a
seem ingly irregular manner, adding a final d i, som etim es also
with change in vocalic quality:

(1 1 ) kate(a) 'chain' katen-begi 'link (chain-eye)'

ardo 'wine' ardan-degi 'tavern (wine-place)'
gazta 'cheese' gaztan-bera 'cream cheese'
 133

bale(a) 'whale' balen-untzi 'whaling ship'

koro(a) 'crown' koron-gile 'crown-maker'

H istorically the basic stems in these exam ples underwent

a rule of intervoca lic nasal stop deletion (orig in a lly accom ­
panied by the sim ultaneous nasalization of the vowel to its left;
of. Michelena, 1985):

(12) Intervocalic Nasal Deletion

n > 0 / V_V

Subsequently, when the last vowel in the resulting vocalic

sequence was -a , this was often interpreted as the singular
article and was lost in the lexical entry. Other vocalic sequences
were also sim plified:

(13 ) Lat. catena > kate(a) 'chain'

Lat. corona > koro(a) 'crown'
Lat. ballena > bale(a) 'whale'
nat. *ardano > ardao > ardo 'wine'
nat. *gaztane > gaztae > gazta 'cheese'

The rule of vowel deletion affected the basic stems of

these words. On the other hand, the nasal consonant was not lost
 134

in the derivational forms, because here the nasal was not in

intervoca lic position (the final vowel being rem oved by the
Vowel Truncation rule (4)).
At som e h is to ric a l p oin t, b e fo re In te rv o c a lic N asal
Deletion came to operate, the words in (11) presented the same
pattern as those in (3). Being trisyllabic, their last vowel was
lost in the derivational form (Vowel Truncation); e.g.: basic form
a rd a n o . derivational form a rd a n . When the language acquired the
rule of Intervocalic Nasal Deletion, its structural description
would be met by the basic form, but not by the derivational form.
This created the alternation a r d a o /a r d a n 'w ine' (o rig in a lly
a rd a o /a rd a n ). Then the vocalic sequence was reduced in the basic
form in many dialects, giving a rd o .
From a co nte m po ra ry, syn ch ro n ic p oint of view , the
sim plest w ay to account for the a lterna tio ns in (11) is to
assum e that the historical rule of Intervocalic Nasal Deletion
still applies in the lexicon to underlying forms such as /katena/,
/ko ro n a /, etc. Nouns such as a rd o /a rd a n 'w in e ' (basic
stem /derivational stem), and g a z ta e /q a z ta n 'cheese' which
also present a vocalic alternation, m ust sim ply be treated as
irregular. There are Biscayan varieties, however, where basic
form s a rd a o , g a z ta e are found. In these dialects the relation
between basic and derivational stem is also straightforw ard for
these items.
 135

The same h istorical explanation can be given to the

irregularity found in items such as a rtz a i(n )/a rtz a n _ 'shepherd'
(e.g. a rtza n -tza 'shepherdy,' a rtza n -o r 'sheep-dog (shepherd-dog')
and a r r a i( n ) /a r r a n 'fish' (e.g. a r r a n - tz a 'fis h in g ,' a rra n -tza le
'fisherman' a rra n -d e q i 'fish-m arket'). M ichelena (1985) proposes
reconstructed form s such as *a rtz a n i , *a rra n i. The derivational
form s, by Vowel Truncation, w ould be a r tz a n . a r r a n : that is,
those still found nowadays. M ichelena suggests that, in the
reconstructed basic stem s the final vowel /I/ palatalized the
nasal stop to Its left and was lost. The result would be pairs
a rtz a h (baslc)/âIlZ âIL (derivational), a r r a h /a r r a n : form s which
are presently found In some d ialects. N evertheless, in the
absence of convincing evidence for postulating a synchronic rule
of re g re ssive p a la ta liz a tio n -c u m -v o w e l d e le tio n , such pairs
must sim ply be taken as irregular from a contem porary point of
v ie w .
A very small group of Items ending in -a change this nasal
into a flap in th e ir derivational stem. I have found three
examples with this alternation. One is jaun 'lord,' which becomes
ja u r in a number of derived forms; e.g. : ja u r-e q i 'palace (lord-
place),' ia u r-e rri 'shire (lord-country),' ja u r-e ts i 'to worship as
a lord.' A second example is oihan 'forest'; e.g.: o ih a r-b ld e 'forest
trail,' olhar-tzain 'forest ranger.' A third example Is e qun 'day,'
 136

which presents a fro me e a u r In a few derived items; e.g.: e o u r-

aldi 'whether,' e a u r-a ts 'athm osphere.'
Finally, there is an item, also ending in -n_, that presents a
to ta lly p e c u lia r a lte rn a tio n w hich has no s y n c h ro n ic or
diachronic explanation known to me: g iz o n 'man (either male
human being or human being, in general),' whose derivational
from is o iz a : e.g.: g iza -se m e 'man, male person (man-son),' o iza -
koi 'm an-loving.'
The rules that intervene in the form ation of derivational
stems can be summarized as in (14):

(14) Derivational Stem Formation

a. Irregular form ation. E.g: /gizon/ - > /giza/ 'man'
b. Lowering
ste m
/ \
[CT a]
I
[-high] - > [+low]

E.g: /b a s o /-> /basa/ 'forest'

c. Vowel Truncation (elsewhere rule wrt (b))
V > 0/ __]
E.g.: /h a rri/ —> /h a rr/ 'stone,' /abere/ —>/aber/ 'cattle,' /ogi/-
-> /og/ 'bread'
 137

d. Final Stop Neutralization (ordered after (c))

[-cont, -son] > [t] / __]
E.g.; /og/ - > /ot/ 'bread'
e. Lateralization (ordered after (c))
/r/ — > [I] /__]
E.g. : /aber/ — > /abel/ 'cattle'

2. Alternating and non-alternating stems

As mentioned, it is not the case that all Basque nouns and
adjectives possess a derivational stem distinct from their free
stem. It is, in fact, only a relatively small group of items that
participate in the alternation, if the whole lexicon is considered.
The group of nouns and adjectives that are affected by the rules
in (14), that is, that have a derivational form distinct from the
basic form , is not easily characterizable. Most of them are
originally Basque roots; but there are some nouns of Latin origin
that present an alternate derivational form, such as a b e re /a b e l
'liv e s to c k ,' from Latin h a b e r e , lo r e /lo r a 'flow e r' from Latin
f l o r e -, e r r e q e / e r r e t 'king' from Latin r e g e - . and e liz a / e liz
'church' from Greek via Latin e c c le s ia . On the other hand, many
purely Basque roots, as far as this fact can be discerned, do not
present an alternate derivational form. For instance J iix i 'town'
does not have a derivational form *h il. A good example of the
u np re d ictab ility of possessing a d istin ct derivational form are
the two stems ero and z o ro . both meaning 'crazy.' Whereas zoro
 138

has a derivational form z o ra (e.g. z o ra -tu 'to become crazy,'

z o ra -q a rri 'wonderful'), ero does not possess a derivational form
e ra (e.g. e ro-tu 'to become crazy' , not *e ra -tu ).‘^
We may attribute to all and only the alternating roots a
feature such as [^alternating], sim ilar to the [+Latinate] feature
proposed for certain English items in SPE, the [+Yamato] feature
th at Ito and M ester (1986) propose fo r Japanese, or the
[+ D ravidian] feature that Mohanan (1986) p ostulates in the
M alayalam lexicon. A fact that justifies the use of the feature
[+Latin ate] in English is that only [+Latinate] roots combine with
[+Latin ate] suffixes. Thus, s a lin itv . which combines a [+Latin ate]
a d je c tiv e w ith a [+ L a tin a te ] s u ffix is w e ll-fo rm e d ; but,
* s w e e tity is ill-form ed. Latinate roots d iffer from other items
in that they both undergo d iffere nt rules and make use of
different affixes. In Basque, on the other hand, alternating items
are affected by different rules, those in (14), which regulate the
alternations, but these stems make use of exactly the same set
of a ffix e s th a t is g e n e ra lly a v a ila b le to o th e r nouns or
a d je c tiv e s .
The alternate derivational stems of [+alternating] form s
will not be listed in the lexicon, since once we know that a basic
stem has a d iffe re n t d e riv a tio n a l stem (i.e. th a t it is
[+alternating]), the shape of its derivational form is predictable

4 This example was suggested to me by Jon Ortiz de Urbina.

 139

and will be given by the regular operation of the Derivational

Stem Form ation rules in (14). Only irre gu la r item s such as
g iz o n /g iz a 'man' will need to have both basic and derivational
form s listed in the lexicon.
It would be redundant to have pairs of connected lexical
entries in the lexicon listing the d erivatio na l stem together
with the basic stem as in (15):

(15) basic derivational

atari > atal 'doorway'
bazkari > bazkal 'midday meal'
atari > afal 'supper'
m erkatari > m erkatal 'm erchant'
aizkora > aizkol 'hatchet'
erdara — > erdal 'foreign'
nekazari > nekazal 'farm er'
azeri > azel 'fox'
abere — > abel 'cattle'

To have such pairs listed in the lexicon would miss the

generalization that once we know that the basic stems in the
lefthand colum n have a d is tin c t d e riva tio n a l stem we can
predict the shape that this derivational stem will have. On the
other hand, we need an indication that those items do possess a
derivational form , as opposed to phonologically sim ilar items
 140

which do not have a different form in derivational contexts, such

as lib u r u (*lib u n 'book,' tx is tu la r i (*tx is tu la l) 'flutist,' o p a r i
f * o p a h 'gift, present,' or p o z k a ri (*p o z k a l) 'joy.' We sim ply
need an indication in the respective lexical entries that w ill tell
us that, say, bazkari 'midday meal' undergoes Derivational Stem
Formation, but pozkari 'joy' does not.

3. Lexical dom ain of the D erivational Stem Form ation rules

We must now determ ine which are the lexical strata in
which the rules of Derivational Stem Formation apply.
The derivational form of a noun has two uses: in derivation
and in adjectival use (subcom pounding). The derivational stem
can be used whenever a derivational suffix is attached, as in
b a s a -ti 'savage,' b a s a -k e r ia 'savagery' from b a s o 'forest,' or
g iz a - k i 'human being,' and g iz a - k o i 'm an-loving' from g iz o n
'man.' It can also be used when the noun is used adjectivally,
m odifying another noun as in b a s a - k a tu 'bobcat' or g iz a -
eskubideak 'human rights.'
Subcom pound form ation is a very productive process in
Basque, as it is in English; and, as in this language, sub­
com pounds may present d iffe re n t degrees of le xica liza tio n .
U nlike English, a d je ctive s and nouns used a d je c tiv a lly in
subcom pounds appear on different sides of the qualified noun.
A d je ctive s fo llo w nouns, but in subcom pounds the m odifier
 141

precedes the modified noun. E.g. : liburu berri 'new book' vs.
euskal liburu 'Basque (language) book,' where e u s k a l is the
derivational form of the noun e u s k a ra 'B a squ e la n g u a g e .'
G enerally speaking, derivational stem s are available for the
formation of new subcompounds.
On the other hand, derivational form s cannot be used in
cocom pounds. Cocompounds, like subcompounds, are formed by
the concatenation of two noun stems. Unlike subcompounds, they
have the further requirement that they must be marked as plural.
In a plural compound of the form X-Y-pl , where X. and Y. are
nouns, am biguities between a cocompound reading ('X's and Y 's')
and a subcom pound reading ('X -like Y 's') may arise if the first
noun, X, does not possess a distinct derivational form . Thus,
s a q a r - m a d a r ia k 'apple-pear-pl' can be interpreted as either
'a p p le lik e pears' (subcom pound) or as 'ap ple s and pears'
(cocompound). There will be no ambiguity, on the other hand, if X
is a derivational stem. In that case, only the subcom pound
reading will be available. A form g iz a -e m a k u m e a k 'm an-w om an-
pl.,' where the derivational form of g iz o n 'man' has been used,
cannot have the cocompound interpretation 'men and women.' Its
only interpretation is as a subcompound, where the first member
is being used adjectivally; that is, som ething like 'm anlike
women' or 'human women.' A cocompound with the meaning 'men
and women' must be formed with the basic form of 'man': g iz o n -
 142

e m a k u m e a k . To give another example, the interpretation of basa-

lu rre ta n 'forest-land-ines. pi.' (b a s o /b a s a 'forest') must be 'in
wild lands' (subcompound). The cocompound 'in forests and lands'
w ould be b a s o - l u r r e t a n . with the basic form of 'forest.'
There are thus two m orphological operations associated
w ith D e riv a tio n a l Stem F o rm a tio n : d e riv a tio n and sub­
com pounding. In two other m orphological operations, inflection
and cocompounding. Derivational Stem Formation does not apply.
I have ju s t shown that cocom pounds cannot be form ed with
d e riv a tio n a l ste m s. D e riv a tio n a l ste m s can never ta ke
inflectional affixes. Form s like *a r t- a k (of. a r d i/a r t ’sheep')
'the sheep, pi.' or *q iz a k (cf. o iz o n /a lz a 'man') 'the men,' where
the d e riv a tio n a l stem has been e rro n e o u s ly p ro vid e d in
in fle ctio na l m orphology, are totally ruled out. A derivatio na l
stem cannot im m ediately precede an inflectional suffix.^

^ A word must be said about the actual use of derivational stem s in those
m orphological contexts w here their use is perm issible. The actual list of
alternating stems varies from dialect to dialect and from speaker to speaker.
More importantly, the fact that a noun has an alternate derivational form does
not m ean in general that this form must necessarily be used, but only that it
can, in the appropriate morphological contexts. In other words, the blocking
effects that Kiparsky (1982) finds in the English lexicon are only very w eak in
Basque. Thus the possibility of obtaining a verb g iza -tu 'to becom e/m ake (like)
a man' using the derivational form of gizon 'man,' does not exclude from the
lexicon a verb o izo n -d u with the same meaning using the basic form of the
noun. What is not possible is to use a derivational form of a noun in contexts
other than derivation and subcompounding.
This difference between the English and Basque lexicons is to be expected,
given the fact that English is a highly standardized language and Basque only
 143

It would seem that the existence of the Derivational Stem

Form ation rules would force us to recognize the existence of
tw o m o rp h o -p h o n o lo g ic a l s tra ta w he re the ru le s a p p ly:
D erivation and Subcom pounding; or, perhaps, a single stratum
c o m b in in g both m o rp h o lo g ic a l p ro c e s s e s . In fle c tio n and
C ocom pounding would constitute different strata, with d ifferent
associated phonological rules but not including the Derivational
Stem Formation rules in their phonological components.
This is all in contradiction to the structure of the Basque
lexicon which rules such as Vowel A ssim ilation, Palatalization
and C onsonant Insertion require in any dialect (cf. chapter 2).
These rules and Derivational Stem Formation seem to give us
two totally different models of the lexical m orpho-phonology of
Basque. The phonological rules examined in chapter 2 require a
fundam ental distinction between inflection, on the one hand, and
all compounding, on the other; derivation being generally treated
in the same way as com pounding. The rules of Derivation and
Stem F orm ation, how ever, establish a d is tin c tio n betw een
derivation and subcom pounding, on the one hand, where the
processes apply, and inflection and cocompounding, on the other
hand, where Derivational Stem Formation does not apply.

recently has started to be codified and extended in its use to a number of

sociological contexts and intellectual fields from which this language had been
traditionally excluded. This has caused a rapid creation of vocabulary with
often several competing forms for the sam e meaning.
 144

Is there a way out of this contradiction? I w ill try to

show that indeed there is a solution. I w ill suggest that the
Derivational Stem Formation rules, although associated in some
m anner with derivation and subcom pounding, do not apply in
conjunction with these m orphological o perations; but, rather,
apply to bare stems. Derivational Stem Formation constitutes an
o peration applying to bare stem s in the lexicon, with the
associated phonological rules in (14), which is undergone only by
entries marked for it. The output form s created by Derivational
Stem F orm ation are p ro vide d w ith in s tru c tio n s fo r th e ir
insertion in certain m orphological structures created on latter
strata. This situation is parallel to what Case marking involves
in languages with overt Case. Forms marked for a given Case, in
many languages, must be created at earlier strata by morpho-
phonological operations (e.g. because the phonological rules
involved in the affixation of Case markers cease to operate at an
early stage). Instructions must then also be provided to ensure
that C ase-m arked w ords w ill be inserted in the appropriate
syntactic contexts (e.g.: "insert in contexts where accusative
Case is assigned").
The generalization that underlies the use of derivational
stem s is that they are used in m orphological positions where
they do not determ ine the lexical category of the word. If we
take the determ ination of the lexical category as the relevant
 145

criterion for being the head of a word (of. W illiam s (1981)), we

can say that derivational stems are used in m orphological non-
head p o sitio n . This includes th e ir use in d e riv a tio n and
subcompounding.
In a subcompound, it is the second member that determines
the category of the word. A derivational stem can thus be used
as firs t m em ber. Also it is the d e riv a tio n a l a ffix th a t
determ ines the category of the word. Derivational stems can be
used with derivational suffixes because the affix is the head of
the word. In cocom pounding, both stems must be of the same
category (i.e. two nouns, two adjectives or two verbs) and the
resulting com pound w ill be of the sam e ca teg o ry as th eir
components. Both members of the cocompound can be considered
heads. D erivational stem s can then not be inserted as any
mem ber of a cocom pound. Finally, inflectional suffixes do not
determ ine the category of the word. The plural absolutive -a k .
for instance, attaches to a noun in lib u ru -a k 'the books' but to
an adjective in liburu gorri-ak 'red books,' w ithout changing the
ca teg o ry of the word. Since in a structure ste m -infle ctio na l
affix it is the stem that is the head of the word, derivational
stems cannot be used in this context.W e can thus propose that
derivational stems created by the rules in (14) are associated
with the instruction "insert in non-word-head position."
 146

The w ord -fo rm a tion o perations th a t take place in the

lexicon c o n s is t not only in b rin g in g to g e th e r s trin g s of
segm ents, elim inating boundaries between them , but also in
c re a tin g w o rd -s tru c tu re , w h ich w ill in c lu d e c a te g o ria l
labelling, using the notion of ’w ord-head.’® A derivational stem
will be insertable at any moment when the appropriate structure
is m o rp h o lo g ic a lly cre a te d , in d e p e n d e n tly of the m orpho-
phonological stratum where this takes place.
According to this, the derivational stem of c iz o n can be
correctly inserted in (16a), (16b), (16c) or (16d), where it will
be in non-head position; but not in (16e) or (16f), where c iz o n is
a head. (R epresentations modeled after Selkirk, 1982. Arrows
indicate headship):

(16) a.
/ A
N V-aff N A-aff

gizon -tu gizon -koi

giza giza
/gizatu/ ’ humanize' /gizakoi/ 'man-loving'

6 This is excluding cliticization. A clitic group will behave like a word only
with respect to phonological rules, but often not at all in the morpho-syntax.
 147

N ^ N
/ \ / \
N N-aff A N
I I I I
gizon -ki N I
I I
giza gizon lan

/gizaki/ 'human being' giza

/gizalan/ 'man's work'

e. N f.
0N
/ \
det det

gizon -a gizon emakume -ak

'giza *giza
/gizona/ 'the man' gizon-emakumeak
'the men and women'

In (e) and (f), o iza . the derivational stem of c iz o n cannot be

inserted because it would be in a head-position.
There is a third type of com pound that we have not
considered yet; noun-adjective compounds, in their normal order
of syntactic collocation. Most compounds with this structure are
exocentric compounds, e.g. b u ru -haundi fburu 'head,' haundi 'big')
'big-headed, one who has a big head,' e s k u -z a b a l fe sku 'hand,'
za b a i 'wide') 'generous, one who has wide hands.' Derivational
stem s do not seem to be available in the form ation of these
 148

exocentric com pounds. This is unlike the situation in Spanish

w here com pounds of this type present the only exam ple of a
regular change sim ilar to those involved in the form ation of
derivational stems in Basque. In Spanish, the last vowel of the
noun is re p la c e d by /i/ in n o u n -a d je c tiv e "d e s c rip tiv e "
compounds (consonantal changes are purely orthographic);

(17)
pelo rojo 'red hair' peli-rrojo 'red-haired (man)'
barba cana 'grey beard' barbi-cano 'grey-bearded'
ojo-s negro-s 'black eyes' oji-negro 'with black eyes'
cabeza dura 'hard head' cabeci-duro 'hard headed'
pata-s larga-s 'long legs' pati-largo 'long-legged'
boca abierta 'open mouth' boqui-abierto 'with an open mouth'

Villasante (1975) suggests a possible Basque influence for

the existence of these compounds in Spanish.
There are also a few Basque examples of compounds with
this noun-adjective structure, which are not exocentric and that
have meanings that are not totally transparent or com positional.
These are often formed with derivational stems, e.g. ; a b e l-g o rri
'free roaming cattle' (cf. abere g o rri 'red cattle'), k a ta - g o r r i
'squirrel (cf. katu gorri 'red cat'). Perhaps these two exam ples
show an early meaning 'free, wild' of the word g o r r i. which it
 149

does not have nowadays. I have nothing to say about compounds

of this type.

4. A lternating adjectives
There are a small num ber of adjectives that participate in
the stem a lterna tio n process. The phonological a lte rn a tio n s
shown by these adjectives follow the same rules th at were
given for nouns:

(1 8 ) maite 'beloved' maita-tu 'to love'

m aita-garri 'lo va b le '
luze long' luza-tu 'to lengthen'
luza-dura 'lengthening'
sendo 'strong' senda-tu 'to heal'
senda-ro 'firm ly'
ugari 'abundant' ugal-du 'to multiply'
ugal-garri 'productive'

W hat is in te re s tin g about th ese a d je c tiv e s is the

distribution of basic and derived stems in the m orphology. The
derivational stems are not used with the set of affixes that can
be considered regular adjectival inflection; those affixes which
can be used unrestrictedly with any adjective: the com parative
-ago 'more,' the superlative -en 'most' and -egi 'too.' With these
 150

suffixes basic stem form s must be used, as is illustrated with

m a ite /m a ita and lu z e /lu z a in (19);

(19) m a ite -a g o 'dearer,' m a ite -e n 'dearest,' m a ite -e g i 'too dear'

lu ze -a q o 'longer,' luze-en 'longest,' lu ze -e g i 'too long'

The affixes in the examples in (19) are regular and totally

p roductive a ffixes which can attach to any adjective in the
language virtually exceptionlessly (only q r 'good' presents some
irre g u la ritie s). They occur interna lly to noun infle ctio n, e.g.
m a ite - e n - e i (stem -sup erlative -plura l dative) 'to those most
d ear,' and are best understood as adjectival inflection. Like
nom inal inflectional affixes, they do not trigger the insertion of
derived stems.
Most compound and derived words (perhaps all?) containing
the derivational stem of an adjective require the transform ation
of the adjective into a verb by a com pletely productive rule of
zero-derivation (of. Ortiz de Urbina (1986b)). A form such as
luza-garri 'lengthenable,' which carries a suffix which norm ally
attaches to verbs (e.g.: uler-garri 'understandable, ikus-garri
'worth seeing,' etc.), would require the transform ation of the
adjective luze/luza 'long' into a verb:

(20) [[[luze a ] V ] garri a ]

 151

The a ttachm ent of the zero-affix w hich determ ines the

category of the word as a verb, is what triggers the insertion of
the derivational stem.
We can conclude that the contexts for the insertion of the
derivational form of an adjective are parallel to those where
derivational form s of nouns are inserted. Derivational form s are
not used in adjectival inflection, where the adjective-stem is
the head, and are used in derivation, where the affix is the head.
There is a final problem concerning the use of derivational
stem s of adjectives. There are two affixes w hich attach to
adjectives to form adverbs: -k i and -r o : both with the functions
and meaning of English adverbial -ly.. As Ortiz de Urbina (1986b)
points out, -ro. takes derivational stems, but -!si does not:

(2 1 ) maite-ki maita-ro 'lo v in g ly '

luze-ki luza-ro 'le n g th ly '
sendo-ki senda-ro 'f ir m ly '

An im p o rta n t d iffe re n ce between these two su ffixes is

that -k i is fully productive and can be attached to any adjective
to p ro d u c e an a d ve rb w hich w ill have a p re d ic ta b le
com positional m eaning, w hereas -rg. is much more restricted in
its use and less tran sp aren t in its in te rp re ta tio n .7 We may

7 G iven the lack of tra n s p a re n c y of - i û ., form s co n tain in g both

adverbializers, m a ite -k i-r o or m a ita -ro -k i 'lovingly' are attested.
 152

conclude that -k i, in spite of the fact that it is used to create

adverbs from adjectives, and thus determ ines the category of
the word, is to be treated as an inflectional, non-head affix.
 153

Chapter 4
Rule Interaction in a Hierarchical Model of Phonological
Representations and in Lexical Phonology

0. Introduction
In this chapter, I exam ine a group of three interrelated
p h o n o lo g ic a l p ro c e s s e s of B a sq u e : P lace A s s im ila tio n ,
C o n tin u a n c y A s s ig n m e n t and P a la ta liz a tio n , w ith in a
hierarchical model of phonological representations. I w ill show
that we can gain a better understanding of the ways in which
p ho no log ical rules may interact and the co n stra in ts in the
a p p lic a tio n of rules by c o n s id e rin g how the g e o m e tric a l
structures created by one rule are then utilized by rules applying
subsequently. It will also be shown that, whereas some of the
rules studied operate only lexically or only postlexically, some
other rules have an application in each of the two components.
Two of the ru le s s tu d ie d in th is c h a p te r. Place
Assim ilation of nasals and laterals to a following consonant and
C ontinuancy Assignm ent, m anifested in the alternation of stop
and fricative allophones of the voiced obstruents, are processes
that Basque shares with Spanish. Previous analyses of these
p ro ce sse s proposed fo r S panish both w ithin lin e a r and
autosegm ental fram eworks are thus considered. It is shown that
the Hierarchical Model provides the means for an improved, more
 154
I

insightful treatm ent of these phonological processes. Then, I

dem onstrate that a treatm ent of these processes w ithin the ■
H ierarchical fram ew ork not only is more adequate but also
makes some interesting predictions for the way a rule operating
on the output of the two rules of Place A ssim ilation and
C on tin ua ncy A ssignm ent, nam ely, P a la ta liza tio n in certain
Basque dialects, can apply to different segm ental structures. It
is shown that some seem ingly puzzling effects of the rule of
Palatalization are in fact predicted by the consideration of the
geom etrical structures that serve as input to the rule.

1. Place Assimilation of Nasals and Laterals

In Basque, nasal segments in a rime assim ilate in point of
articulation to a follow ing consonant. Laterals also assim ilate
in place, but in a more restricted way; they do not assim ilate if
the follow ing consonant is a velar or a labial (of. Salaburu,
1983). Assim ilation of nasals and laterals thus takes place in
the same manner as in Spanish (cf., for instance, Harris, 1969,
Cressey, 1978). As in Spanish, nasal and lateral assim ilation
applies both w ord-internally and across word boundaries. Basque
examples are given in (1) and Spanish examples in (2). In each
case, assim ilated forms are given under (a) and examples where
assimilation does not take place in (b):
 155

(1) Place Assimilation of Nasal and Laterals in Basque

a. egu[m] berri 'new day'
egu[n]] fresku 'cool day'
egu[n] totu 'fat day' a ta [l] totu 'fat section'
egu[n] ttiki 'small day' ata[/C] ttiki 'small section'
egu[Q] gorri 'red day'

b.
ata[l] berri 'new section'
ata[l] fresku 'cool section'
ata[l] gorri 'red section'

(2) Place Assimilation of Nasals and Laterals in Spanish

a. [um pwérto] 'a port'
[urg foko] 'a focus'
[u n tio] ' an uncle' [el tio] 'the uncle'
[un' Elko ]'a boy' [el' Elko] 'the boy'
[uh y ate] 'a yacht' [e/C yàte] 'the yacht'
[ur] gorila] 'a gorilla

[el pwérto] 'the port'

[el foko] 'the focus'
[el gorila] 'the gorilla'
 156

The examples show that, whereas a nasal will assim ilate in

place of a rtic u la tio n to any fo llo w in g co n s o n a n t (b ila b ia l,
labiodental, dental, prepalatal, palatal or velar), laterals fail to
assim ilate to bilabial, labiodental and velar articulations.
I will consider the more straightforw ard process of Nasal
A ssim ilation first. Nasal assim ilation, as attested in Spanish
and B asque, am ong m any o the r languages, c o n s titu te s a
p a ra d ig m a tic e xam ple of a n a tu ra l, c o n c e p tu a lly sim p le
phonological process whose naturalness is all but obliterated in
its standard SPE-type form ulation. Such a form ulation is given
in (3) (from Harris (1984)):

(3) ■+ cons'
a cor a cor
[+nas]— > p ant / —
P ant
Y dist Y dist
6 back .5 back

Under the com m on assum ption that there should be a

c o rre la tio n b etw e en the n a tu ra ln e s s and c ro s s -lin g u is tic
generality of a process and the sim plicity of its form ulation, the
com plexity of rule (3) points to a clear inadequacy of the SPE-
fram ew ork. W orking w ithin this fram ew ork, in w hich it was
generally assumed that the value that is assigned to a rule by
the grammar is a function of the number of features whose value
 157

is changed, Cressey (1978) proposes to replace the mention of

the features whose value is changed in rule (3) with the notation
PA, for 'point of articulation.' Rule (3) can then have a sim pler
form ulation as in (4):

+00 ns
(4) [+ nasal] > a PA/ a PA

The theoretical significance of the PA notation, qualified

by C ressey (1978, 65) as 'abbreviatory device' is, however,
unclear. One position to take could be th at 'PA' is sim ply
shorthand, with no theoretical implications. On the other hand, it
could be understood as a real archifeature grouping together all
place features. The development of such an approach could give
as a result a theory in which features which tend to act together
fo r p h o n o lo g ic a l p ro c e s s e s a re g ro u p e d in d iffe r e n t
archifeatures, perhaps some more inclusive than others. These
archifeatures would thus be com parable to the nodes in the
Hierarchical Model. In the Hierarchical Model, where the Place
Node (P) constitutes a perm anent elem ent in the structure of
segm ents, the process of Nasal Assim ilation illustrated in (1)
and (2) can be captured as a sim ple operation of P node
spreading; a process whereby a P node spreads leftwards to a
nasal in a rime, as expressed in (5a). Individual place features
need not be expressed in the form ulation of the process.
 158

Assim ilation will take place in all and only those features that
are dom inated by the Place Node. The process is graphically
illustrated in (5b):

(5) Nasal Assim ilation

a. Operation: Spread
Argument: P
Direction: leftwards
Target Conditions: [+nasal], rime

rime
X X
6 R
[+nasal] o SL

Let usconsider Lateral Assim ilation now. As indicated, the

a ssim ila tio n of la te ra ls to the p oin t of a rtic u la tio n of a
fo llo w in g co nso na n t d iffe rs from Nasal A ssim ila tion in that
there are some a rticu lato ry points to w hich laterals do not
assim ilate. If the follow ing consonant is a labial or a velar,
assim ilation in place will not take place.
The reason for these restrictions seems to be the universal
im p o ssib ility or d iffic u lty to produce a lateral a rticu latio n at
ce rta in places. A lateral m anner of a rtic u la tio n g e n e ra lly
 159

requires the presence of the tip or blade of the tonge as active

a r tic u la to r :

G e ne rally speaking, late ra ls are made w ith the tip,

b la d e , or fro n t of th e to n g u e . T he y m ay be
e ithe r dental (as in French), alveolar (as in English),
retroflex (as in M alayalam and other Indian languages),
or palatal (as in Italian). Velar laterals do occur in a
few languages spoken in Papua New Guinea, but they are
so uncom m on th a t we w ill not co nsid er them here.
(Ladefoged (1982, 156)).

Given these universal phonetic restrictions, it seems that

we should treat both Nasal and Lateral Assim ilation in Spanish
and Basque as instances of the same process. The process will
be blocked, in the case of the laterals, when the output
co n fig u ra tio n w ould vio la te universal co n stra in ts establishing
the co m p a tib ility of fe a tu re s on a segm ent. One of these
constraints, acting as an output filte r for Lateral Assim ilation,
is the sta te m e n t of a co m p le te ly u niversal in c o m p a tib ility
between the features [lateral] and [labial]. This is expressed in
(5):

(5) A segment X cannot bear both [labial] and [lateral] (Universal)

T his c o n s tra in t accounts fo r the b lockage of Lateral

A s s im ila tio n in th e case of b ila b ia l and la b io d e n ta l
a rtic u la tio n s .
 160

Given the rarity of velar laterals, we may propose a near-

universal filte r analogous to (5) to block the creation of these
segm ents:

(6) A segm ent X cannot bear both [dorsal] and [lateral] (near-
u n iv e rs a l)

There are, however, two problems with (6). A first problem

with such a co n stra in t is the fact that it is not absolutely
universal; but seems to be violated by some languages of Papua-
New Guinea. Secondly, the dark or velarized [t],fo u n d in coda
position in languages like English and Catalan, where the main
articulation is coronal but where the back of the tongue is
raised tow ards the velum (Ladefoged (1982, 61-62) MacKay
(1987, 105-06)), would seem to require its characterization as
bearing a [dorsal] secondary point of articulation. These two
problem s are avoided if, rather than a negative co n stra in t
prohibiting segm ents bearing both features [+lat] and [dorsal],
we have a positive constraint that requires lateral segm ents to
have a coronal articulation.
We may assume that every segm ent in a language like
Basque or Spanish will bear one of three m onovalent place
features, [labial], [coronal] or [dorsal], depending on whether the
active articulator is the lower lip, the edge of the tongue, or the
body of the tongue (of. Sagey (1986)). Then we can unify the two
 161

negative constraints in (5) and (6) and also avoid the problems
that (6) presents by means of a positive constraint requiring
the presence of the feature [coronal] in a [lateral] segment:

(7) If a segment X bears [lateral] then it must bear [coronal].

The filter in (7) can be built into the hierarchy by making

the feature [lateral] depend hierarchically on a [coronal] node
(of. Sagey (1 9 8 6 )). A s s u m in g su ch a c o n s tra in t on
rep re se n ta tio n s, we can then fo rm u la te Nasal and Lateral
Assim ilation in Basque and Spanish as a single rule whereby a
n o n -c o n tin u a n t s o n o ra n t in a rim e a c q u ire s the place
specifications of a consonant to its right by P node spreading :

(8) Nasal and Lateral Assim ilation

a. Operation: Spread
Argument: P
Direction: leftwards
Target Conditions: [-cont], [+son] , rime

rime
\
X X
[-cont] o 6 R
[+son ]--------------L 9 SL
I P
 162

2. Continuancy Assignm ent

In Basque, as in Spanish, there is a conditioned allophonic
alternation between voiced stops [b, d, g] and voiced fricatives
[b, b, g]. Although in both languages there is a certain amount of
dialectal and free variation (cf. Lozano (1979) for Spanish), the
basic distributional pattern of the two series of allophones is
the same in the two languages: voiced stops are found word-
initially, after nasal and, in the case of [d], also after a lateral;
voiced fricatives are found elsewhere (of. for instance, Harris
(1969), Lozano (1979), for Spanish, and H olm er and Holm er
(1968), for Basque);"*

(9) D istribution of Stop & F ricative A llophones of Voiced

O b stru e n ts
a. Spanish
[bote] [umbote] [elfeote] [rebote] [rezbala] [arbol]
'boat' 'a boat' 'the boat' 'bouncing' 'slides' 'tree'
[daho] [undaho] [eldaho] [reéâfio] [dézée] [ârbe]
'damage' 'a damage' 'the damage' 'caul' 'from' 'burns'
[gato] [urjgato] [elgato] [agatas] [dezgrana] [ergfbo]
'cat' 'a cat' 'the cat' 'crawling' 'threshes' 'upright'

1 Here I will ignore the epenthetic [b] in the dialect of Arbizu, which is
usually realized as a stop in spite of being in intervocalic position (of. chapter
2, section 3.3). As indicated in chapter 2, I assume that the assignment of
continuancy to this epenthetic segment is brought about by the same process
that assigns it all other consonantal features.
 163

b. Basque
[beri] [embora] [alfeoan] [afeere] [dezberdin] [arbi]
'new' 'trunk' 'beside' 'cattle' 'uneven' 'turnip'
[dore] [mendi] [talde] [adar] [ezbuyn] [arbi]
'tower' 'mountain' 'group' 'horn' 'unworthy' 'sheep'
[gori] [at] go] [algodoy] [lagun] [dezgogo] [argi]
■red' 'of there' 'cotton' 'friend' 'reluctance' 'light'

As the examples show, in both Spanish and Basque, voiced

stops have a more restricted d istrib u tio n than the frica tive
allophones. They normally appear only after pause and after a
hom org an ic nasal or lateral. R ecall that, w hereas nasals
a s s im ila te to the p o in t of a rtic u la tio n of any fo llo w in g
consonant, laterals fail to a ssim ila te to n on -co ron als (see
section 1). The distribution observed in (9) thus suggests that
the assignm ent of continuancy to voiced obstruents is a process
o p e ra tin g a fte r N asal and L a te ra l A s s im ila tio n . V o ice d
obstruents surface as stops in those cases when Nasal and
Lateral Assim ilation has created a hom organic cluster and as
fricatives when the formation of such clusters is blocked, i.e. in
the case of [lb] and [If] (second and third columns in (9)).
The continuant/non-continuant alternation could be viewed
as resulting from either a spirantization process turning stops
into frica tive s, or a fo rtitio n process m aking frica tive s into
sto p s. In the a bse nce of o th e r a rg u m e n ts, m arked n ess
 164

considerations would lead one to posit voiced stops and not

fricatives as basic segments, with a spirantization rule creating
fric a tiv e a llo p h o n e s in ce rta in e nviro n m e n ts. T hat is the
solution proposed, for instance, by Harris (1969, 40) for Spanish
and Txillardegi (1980) for Basque. Harris' rule is given in (10):

( 10)
[+ obst]
+ obst ■+ cont ; [+ cont] \ (#)
- tense — > -strid /i < [a cor]>

As stated in (10), /b,d,g/ are spirantized after obstruents,

a fte r c o n tin u a n t segm ents, and also a fte r n o n -c o n tin u a n t
sonorants (i.e. laterals and nasals) if there is disagreem ent in
the value for the feature [coronal]. This spirantization analysis
has the obvious disadvantage of the great com plexity required in
the sta te m e nt of the e nvironm ent of the rule. The form al
com plexity needed in the form ulation of the spirantization rule
is one reason that led Lozano (1979) to reject this analysis in
favor of a fortition or stop-form ation rule. Another reason that
Lozano gives for preferring a fortition rule is that the existence
of such a fortition process is independently attested in Spanish
in the case of [w] and [y], w hich present non-continuant
allophones in roughly the same contexts where [b,d,g] are found.
Lozano actually proposes underlying /B, D, G/, unspecified for
the feature continuant. To these underspecified segments a rule
 165

(S to p -fo rm a tio n ) w ill assign the fe a tu re [-c o n tin u a n t] in

certain environm ents; and a com plem ent rule w ill fill in the
feature [+ continuant] elsewhere.
W ithin the autosegm ental fram ework, Mascaro (1984), who
considers Basque and Catalan in addition to Spanish, proposes
th a t, in th e se lan g ua ge s, vo ice d o b s tru e n ts , w hich are
u n d e rly in g ly u n s p e c ifie d fo r c o n tin u a n c y , re c e iv e th e ir
specification for this feature by spreading from a segm ent to
th eir left. If there is no segm ent to the left of the voiced
obstruent (i.e., after pause), the feature value [-continuant] is
inserted. The core of Mascaro’s analysis thus consists of a rule
spreading both values of the fe a tu re co n tin u a n t, w ith an
insertion rule applying in contexts where the spreading rule
cannot apply. Mascaro's analysis is summarized in (11):

(11) Mascaro ’ s (1984) analysis:

a. Spread [a cont] rightward from any [a cont] to a sequence
of nonstrident voiced obstruents, (p. 296)
b. Assign [-cont] after pause.

How Mascaro's analysis in (11) would work is shown with

examples in (12):
 166

(12) I+ct] [-et] [+ct] [-et]

f\ i'\ i\ \ :
reBote umBote aBDika Bote

[rebote] [umbote] [abbika] [bote]

In /reB ote/ the segm ent to the left of the under-specified

o b s tru e n t bears the fe a tu re [+ co n tin u a n t] and rightw ard
autosegm ental spreading of this feature by rule (11a) produces
a co n tin u a n t [b ]. In /u m B ote/ the same rule (1 la ) spreads
[-c o n tin u a n t] from the nasal to the left of /B/, giving a non­
continuant [b]. in /aB D ika/ the sequence of unspecified voiced
obstruents receives a [+continuant] specification from the vowel
to their left. Finally, in /Bote/, there is no segm ent to the left
of /B / from which it could receive a value for continuancy, and
(11b) applies inserting [-continuant].
The problem with this analysis is presented by sequences
where the voiced obstruent is preceded by a lateral. Recall that
in this environm ent /□ / is non-continuant, but /B / and /G / are
continuant. Since, according to rule (1 la ), the continuant value
of the voiced obstruent is spread from the lateral both in [Id] and
in [lb], [Ig], the only way to derive the correct specification for
continuancy in each case would be to postulate that /I/ bears
the fe a tu re [-co n tin u a n t] when it is fo llo w e d by /□ /, but
[+continuant] in the sequences /IB/, /IG/. This is, in fact, the
solution that Mascarô suggests: "One might suggest that laterals
 167

are stops with respect to some segm ents (dentals, alveolars,

p a la ta ls ), but fric a tiv e s w ith re sp e ct to o th e r segm ents
(labials, velars)." (Mascarô (1984, 293)).
In the two exam ples in (13), /I/ w ould bear opposite
specifications for continuancy. It would be [-continuant] before
/D /, but [+continuant] before /G/. Only once the [continuant]
value of the lateral is determined, can the spreading rule (11a)
assign the correct value to each of the voiced obstruents:

(13) [-ct] [+ct]

i\ I\
k a I D o [kaldo] ka IBo [kalbo]

There are very serious problems with this approach, both

em pirical and theoretical. These are problem s that stem from
the fact th a t M ascaro does not consider the hom organicity
between the /n/ or /I/ and /B ,D ,G / as a factor that determ ines
w hether the voiced obstruent w ill surface as a stop (if the
c lu s te r is hom organic) or as a fric a tiv e (if th ere is no
hom organicity). This hom organicity condition was encoded in
both Harris' (1969) and Lozano's (1979) form ulations of the rule.
In doing away with the hom organicity condition, Mascaro must
explain the contrast between the stop realization in the group
[Id], on the one hand, and the fricative in the sequences [lb], [Ig],
on the other, in some different manner. The solution that he
proposes for this distribution is, however, untenable for both
 168

factual and purely theoretical reasons. Before expanding on

these problem s w ith M ascaro's analysis, I w ill present the
considerations that may have led Mascaro to prefer an analysis
that does not make m ention of hom organic clusters. These
considerations would seem, again, to be both of a factual and of
a theoretical nature.
The theoretical reason could be the relative d ifficu lty of
c a p tu rin g a co n d itio n on h o m o rg a n icity w ith in the A u to ­
segmental Theory in which he casts his analysis.
The empirical reason is that Mascaro wants to account also
for the facts of Catalan, Basque and Spanish by means of the
same basic rule of autosegmental spreading; although the more
com plicated Catalan facts would require some additional rules.
In C atalan, nasals need not be hom organic with follow ing
consonants, unlike Spanish or Basque where non-hom organic
clusters are not found in normal pronunciation. Crucially, voiced
o b s tru e n ts in C a ta la n are e x c e p tio n le s s ly re a liz e d as
[-co n tin u a n t] after nasals regardless of hom organicity or lack of
it, as in the examples in (14):

(14) Catalan Non-homorganic Clusters

[som dos] 'we are two'
[som grans] 'we are big'
[sir] bins] 'five wines'
[sir] dits] 'five fingers'
 169

It is, however, not clear that we should account for the

d istrib u tio n of stop and fric a tiv e a llo ph o ne s of the voiced
obstruents in Catalan with the same basic system of rules as
for Basque and Spanish. The Catalan facts are different enough
from those of the other languages as to grant a differe nt
a na lysis; com pare Spanish /a B D ika / [a b b fk a ] w ith C atalan
/aBD ika/ [e b d fka ] ' he abdicates' (example from Mascaro (1984))
or Spanish /re g ia / [ré g la ] with Catalan /re g ia / [r é g g le ] 'rule'
(Catalan example from DeCesaris (1981)). In general, Catalan has
voiced stops in many more contexts than Spanish or Basque. I
will not attem pt an analysis of the Catalan facts here. Perhaps
in this language the rule of continuancy assignm ent must be
fo rm u la te d in such a w ay as to produce n o n -c o n tin u a n t
obstruents after any nasal but only after a homorganic lateral. In
any case, Basque and Spanish are susceptible to an analysis
which requires hom organicity between noncontinuant sonorants
and fo llo w in g voiced obstruents for the a ssig n m e nt of the
[—continuant] feature. It w ill be shown In section 3 that an
analysis requiring hom organicity for [-continuant] assignm ent
is well supported by Basque palatalization phenom ena. I will
now show that the alternative that Mascaro suggests to account
for the [Id] versus [lb], [Ig] contrast (to make the continuant
value of the lateral dependent on the co n te xt), is to ta lly
unacceptable for reasons that have to do both with other facts
of the languages in question and with theoretical argumentation.
 170

To begin with the em pirical problems, in Basque, laterals

consistently behave as non-continuant segm ents, with respect
to all other segm ents. Direct evidence is offered by a rule of
Standard Basque which voices voiceless stops after /I/ and /n/
in certain morphological contexts. As the examples in (15) show,
in Standard Basque, the initial voiceless stop of the suffixes
/ —ko/ 'o f and /-tik / 'from ' is voiced only after a nasal or a
la te ra l.^ A fter a vowel no change takes place and after any
other consonant an epenthetic [e] is inserted:

(18) Voicing of Stops in Standard Basque

a. Irun-go 'of Irun'
Brasil-go 'of Brazil'
Erroma-ko 'of Rome'
mendi-ko 'of the mountain'
Irak-eko 'of Iraq'
Aralarr-eko 'of Aralar'
b. Irun-dik ' from Irun'
Brasil-dik 'from Brazil'
Erroma-tik ' from Rome'
mendi-tik ' from the mountain'
Irak-etik 'from Iraq'
Aralarr-etik 'from Aralar'

2 These suffixes undergo voicing after lateral or nasal only with proper
names and adverbial expressions. With common nouns, there is epenthesis.
 171

Since the only triggers of Stop Voicing are laterals and

nasals, the triggering environm ent must be characterized as
"after a non-continuant sonorant." A linear form ulation of the
Stop Voicing Rule is given in (16):

(16) ■- cont ' + son ■

- son — > [+ voice] / _- cont

The segm ent /I/ m ust, therefore, be treated as a non­

continuant both with respect to the velar consonant of /-ko/ and
to the dental consonant of /-tik/.
Other rules of Basque such as Palatalization (of. section 3
below) also show that /I/ co nsiste ntly behaves as a non­
co n tin u a n t segm ent in Basque. For argum ents in fa vor of
characterizing /I/ as non-continuant also in Spanish and other
Romance languages, see Tatô (1981). The observed behavior of
/I/ with respect to other rules thus does not support Mascaro's
c o n te n tio n th a t the c o n tin u a n c y v a lu e of /I/ m ay be
environm entally conditioned. But the theoretical problem s that
M ascaro's suggestion w ould create, even if, for the sake of
argum entation, we ignore the Basque data presented in (15),
constitute a not less significant argum ent against his position.
It can be argued th a t late ra ls are not c le a rly non­
co n tin u a n t segm ents, since there is co ntin ou s flo w of air
through the sides of the tongue during its articulation. It would
 172

be therefore not surprising if languages were to differ on the

value for continuancy that is assigned to /I/. This possibility is
in fa ct discussed in SPE (pp. 317-8). H ow ever, to assign
a lte rn a tiv e c o n tin u a n t values to /I/ in the sam e language,
depending on the environment, is much more problematic. In the
languages that we are examining, Basque and Spanish, to obtain
the desired results in exam ples such as those in (9), a rule
would have to assign the value [+continuant] to /I/ before /B /
and /G / (again, ignoring the Basque facts in (15)). Elsewhere /I/
could receive the [-continuant] value by default. The rules
determining the continuant value of /I/ are given in (17):

(17) A ssignm ent of C ontinuancy to Laterals (rules im plicit in

M ascaro's analysis)

a. /I/ is [+cont] before /B / and /G/

- cor
- son
[+lat] - > [+ 0 0 nt] / + voice

b. Default: [+Iat] - > [ -cont]

Only then following /B, D, G/ will each receive the correct

[continuant] specification by spreading from the lateral.
This analysis is to ta lly circular. In order to spread the
correct value for [continuant] to a voiced obstruent to the right
/I/ we m ust firs t give th is 'p lu s ' or 'm in us' value to /I/
 173

according to whether /B, G/ or /D/ is to its left. There is no way

to avoid this circularity. We must therefore reject M ascaro's
(1984) analysis also on theoretical grounds.
This problem, which was not present in previous analyses,
arises, as indicated above, from the fact that Mascaro does not
consider the hom organicity or lack thereof between a lateral and
a voiced obstru en t to its right as a cond ition ing fa c to r to
dete rm in e w h e th er the la tte r w ill surface as a stop or a
fr ic a tiv e .
I will now develop an account of the distribution of voiced
fric a tiv e s and stops in Basque and S panish w ith in the
Hierarchical fram ework which does not present these problems.
In the H ierarchical Model rules apply to segm ents by
m odifying existing structures. The rule of Nasal and Lateral
Assim ilation seen in section 1, thus had the effect of creating
clusters of two segm ents with a shared Place Node. These
modified structures can then serve as input to other rules. That
is the case of the main rule that creates voiced stops in Basque
and Spanish.
I w ill assum e that the segm ents /B ,D ,G / are underlyingly
unspecified for continuancy, since their surface specification is
contextually determined. To determ ine surface values we clearly
do not need to spread both [continuant] values, as in Mascaro's
rule. It is enough to spread one of the values, the other one
 174

c o n s titu tin g the d e fa u lt s p e c ific a tio n . The co n te x ts w here

voiced obstruents surface as stops are more restricted than the
contexts of spirantization. As can be observed from H arris'
(1969) spirantization rule in (10) above, the statem ent of the
contexts where voiced obstruents surface as stops needs to be
rather com plex. This suggests that [+continuant] is the default
value. In two specific contexts /B,D ,G / are assigned the feature
[-continuant]: to the right of a homorganic nasal or lateral, and
after pause. In the first case, the [-continuant] value w ill be
assigned by spreading from the sonorant; in the second case by
feature insertion. If no value is otherwise assigned, /B ,D ,G / will
receive the feature [+continuant] by default.
We thus have a rule that spreads the feature [-continuant]
rightw ards from a non-continuant sonorant (i.e. a nasal or a
lateral) to a voiced obstruent within a homorganic cluster:

(18) [-continuant] spreading in Homorganic Clusters

a. Operation: Spread
Argument: [-continuant]
Direction: Rightwards
Target Conditions: [+voice], [-sonorant]
Trigger/Target Condition: Homorganic cluster (shared P)
 175

a.' [-cont]
N
\N [+voice]
N

[-son]
SL
P

Rule (18), which crucially requires the previous application

of Nasal and Lateral A ssim ila tio n , w ill assign the fe ature
[-co n tin u a n t] to the obstruent in the hom organic clusters [mb],
[aD], [t]G] and [ID]; but will not take as input the non-homorganic
sequences /IB/, /IG/.
There is only one other context where voiced obstruents
surface as [-continuant]: after pause. We thus need an additional
rule which will assign the feature [-continuant] to an utterance-
initial voiced obstruent:

(19) After Pause [-continuant] Insertion

Operation: Insert
Argum ent: [-cont]
Target Conditions: [-cont], H ___

It is justified to treat the form ation of stops after pause

by a d iffe re n t rule than the one that applies to hom organic
 176

clusters, (18), since the u tte ra n ce -in itia l position presents a

greater degree of variability. Although voiced stops are the most
frequent realization after pause, frica tive realizations are not
uncommon in this position (cf. Lozano (1979), Malberg (1955).
Elsew here a d efault rule assigns the feature [+continuant] to
/B ,D ,G /:

(20) Default
+VOice
-son —> [+cont]

Some sample derivations showing how this analysis works

are given in (21):

(2 1 )

/u N B ote ka I B o ka I D o D aho a G o ke D o/
NLA m I
-ctSpr b d
AP-ct d
Def. b g d
[umbote] [kalfeo] [kaldo] [dano] [ago] [kedo]

3. P alatalization
M ost Basque d ia le cts possess a rule of p ro gre ssive
p a la ta liza tio n w hich a ffects som e subset of the underlying
 177

coronal segments (cf. chapter 2). The concrete realization of the

palatalization rule shows considerable variation from dialect to
dialect. Basque dialects d iffer in w hether palatalization takes
place only after a fro n t glide /y/ or both after high front
vow els and glides. A second difference is in the requirem ent
that the target be follow ed by a vowel, which not all dialects
have. A third important difference is in the set of segm ents that
undergo the palatalization rule. In many dialects only In i and /I/
can be palatalized. In other dialects /t/ is also a possible target.
Finally there are dialects where 161 also palatalizes, but only in
the sequences /ind/, /ild /, w here /i/ is either a glide or a
vowel. For the geographical distribution of the different results
of palatalization see Echaide (1976). Here I w ill present data
from three d ifferent dialects: a restrictive dialect (Baztan), an
interm ediate d ia lect (D onostia-S an S ebastian), and a d ia le ct
w here the application of pala taliza tion is m axim ally general
w ith respect to the param eters of variation ju s t m entioned
(O nd a rroa ). I w ill show th at the rule of p a la ta liz a tio n is
e ssen tially the same for all Basque dialects, differences of
realization being caused by additional conditions on the trigger
or ta rg e t of the rule. I w ill sta rt with the d escriptio n and
a na lysis of the d ia lect of O ndarroa, which show s the least
re strictive pattern of p alatalization.
 178

3.1. Palatalization in Ondarroa

A geographical area where the process of palatalization is
particularly pervasive is the North of Biscay. C haracteristically,
in this area, /I/, /n / and /t/ undergo palatalization after vocalic
or n o n v o c a lic /i/ (cf. 2 2 a ).^ T he consonant /d / a lso
palatalizes, but only after /il/ or /in/, a co ntext where /t/ is
palatalized as w ell (of. 22b). D irectly after /i/, on the other
hand, /d / is not palatalized (of. 22c). The examples are from the
4
va rie ty of O ndarroa. Some variation is found w ithin the area
in the realization of the palatal allophone of /t/, w hich, in
O ndarroa, is, now adays, the a ffrica te [tS]. O lder O ndarroa
speakers have a palatal stop [f], which is still the result of the
palatalization of /t/ in other towns;

(2 2 )
a. Palatalization of IXJ , /I/ and /n / after /i/, /y /
/ b a r i- t u / [baritS u] 'to renew (perf.)'
/am ai-tu/ [am aytSu] 'to finish' (perf.)

3 In Basque there is a predictable alternation between high vowels and glides.

High vowels become glides when following another vowel. The distinction
between vowels and glides is, however, important for the rule of palatalization
in Basque since, as we will see, in some dialects (e.g. Baztan, cf.3.2), only
glides, but not vowels trigger palatalization. Front glides are further often
absorbed into the following palatal consonant, optionally or obligatorily,
depending on the dialect.
4 For the Ondarroa data I must thank especially Am aia Iparragirre and also
Josu Belaustegi and Goizane Zabala. Rotaetxe (1979) has also been used as a
source of data.
 179

/m e n d i-tik/ [m enditS ik] 'from the mountain'

/m util-a/ [muti/Ce] ' the boy' (abs.)
/il-a / [i/Ce] 'dead'
/m ila/ [mi/Ca] 'a thousand'
/ipin-i/ [ipini] 'to put' (perf.)
/m in-es/ [mines] 'with pain’
/ b i- n a/ [bine] 'two for each'
c. Palatalization of /d/ and /t/ after [h] and [/(]
/ in d a r / [in J a r] 's tre n g th '
/a le g in -tu / [a le g ih j u] 'make an effort'
/il da/ [ix je ] ' he died'
/egin dau/ [e y h ja w ] ' he made it'
/pinta/ [p ih t^a ] 'p a in t!'
/b a r d in - ta s u n / [bardihtS asun] 'sam eness'
/ ip i n - t e n / [ip ih tS en ] 'to put (imp.)'
d. N on-palatalization of 16/ after /i/
/ b id e / [bide], [bire] 'w a y '
/ b id e r / [b id e r] 'tim e '
/ id i/ [id i], [iri] 'ox'
/ id u r i / [id u ri] 'im age'
/a b a d e / [adade], [adare] 'p r ie s t'

As fo r the sequence /is/, p a la ta liza tio n seem s to be

sp re a d in g th ro u g h o u t the lexicon in O n d a rro a and o th e r
localities. In Rollo's (1925) description of the M arkina dialect,
 180

the palatalization of other coronals takes place very much as in

O ndarroa (with the difference that palatal /t/ is the stop [t']).
The coronal fricative /s / clearly does not undergo progressive
palatalization. Surface [is] is found quite unrestrictedly :

(23) [gison] 'man'

[gisen] 'f a t '
[isen] 'name'
[bisi] 'to live'
[b is e r] 'beard'
[isan] 'to be'
[is u r] 'w rin k le

In Ondarroa some of these items nowadays present [S], e.g.;

[g i$ o n ], [i$ e n ], [bi$ i]. But surface [is] is also found in other
examples, e.g.; [isa r] 'star,' [bisar] 'beard,' [ise rd i] 'sweat.' This is
in contrast with the palatalization of other segm ents, where the
only lexical exceptions are very recent borrowings. In the case
of /I/, /n /, /t/ and /d / the a pp lica tio n of the rule across
stratum I morphemes shows that there is a synchronic rule of
palatalization. In the case of /s/, however, I have not found any
re g u la r a lte rn a tio n betw een [s] and [$ ] show ing th a t /s /
undergoes palatalization synchronically. Our form ulation of the

rule of Palatalization will then exclude /s/ as a target.^

4 It should be mentioned, on the other hand, that surface [i§] is indeed much
more frequent than [is] given the fact that a rule inserts [S] between N and
 181

The palatalization rule can be form ulated as in (24):

(24) Palatalization in Ondarroa

Operation: Spread
Argument: [dorsal]
Direction: Rightwards
Trigger Conditions: [+high], [-back]
Target Conditions: [-continuant], [coronal]

This rule thus spreads the [dorsal] node rightwards from a

high, front segment (i.e. [i] or [y]) to a coronal stop, producing a
d o u b ly a rtic u la te d d o rs o -c o ro n a l se g m e n t (i.e . a p a la ta l
consonant, cf. Keating (1987)). The sequences /il/, /in/, /it/
co n stitu te le g itim a te inputs fo r the p ala taliza tion rule. They
will undergo the rule as shown graphically in (25):

(25) X ^
6 [-cont] R
SL
t
9 _ - - 0------ [cor] P
[d p ri]
[-b k ]
[+hi]

another vowel, e.g. /m en d i-a/ [m endiSe] 'the mountain.' Most instances of [iS],
therefore, do not represent the operation of the Palatalization rule but rather of
the rule inserting [S] after /i/.
 182

Of particular interest is the behavior of /d/ with respect to

p a la ta liz a tio n , s in c e th is segm ent does not u n d e rg o
palatalization when im m ediately preceded by a palatal vowel or
glide, but palatalizes after a palatal nasal or lateral. Given the
effects of the continuancy assignm ent rules and the fact that
only non-continuant segm ents can undergo palatalization, the
behavior of /d / is predictable. A fter a palatal vowel or glide,
/d / w ill fail to palatalize since in this context /d / does not
m eet the requirem ent of being [-cont]. Only after nasal or
lateral /d / is assigned the feature [-continuant] (by rule (18),
[-o n t] Spreading in Hom organic Clusters) and thus becomes a
p ossib le ta rg e t of p a la ta liz a tio n . T hat only n o n -co n tin u a n t
segm ents are possible targets of the rule can be deduced from
the fact that /s/ does not palatalize regularly and neither do the
continuant sonorants /r/ or / r / .
C onsider then the sequences w here a nasal or lateral
intervenes between the trigger of palatalization, a palatal vowel
or glide, and W or /d/. The two consonants in these sequences
w ill share a single Place Node as a consequence of the
application of Nasal and Lateral Assim ilation (rule (8)). The rule
of Palatalization will not distinguish these inputs from those in
(22a). In (26) it is shown how a sequence such as /in d/ is
affected by the palatalization rule:
 183

(26) n
X
o—[-cont]—o R
SL
P
[cor]
[dors]
[+high]
[-back]

S c h e m a tic a lly the e ffe c ts of P a la ta liz a tio n in the

sequences /it/ and /ind/ are compared in (27):

(27) n

,0
[dors] [dor]

It should be mentioned that, som ew hat unexpectedly, the

coronal a ffricate /ts / does not undergo palatalization, as the
exam ples in (28) with surface /its/ show:

(28) [its u ] 'b lin d '

[its a l] 'sad, huge'
[ i l t si ] 'the nail'
 184

[g ilts e ] 'the key'

[a rits u ] 's to n y '
[b a ré in ts u ] 'sim ilar'
[s a ld its a t] 'for a horse (pro!.)'
[mutiitsat] 'for a boy (prol.)'

T his segm ent bears both fe a tu re s [c o ro n a l] and

[—c o n tin u a n t] and should th e re fo re undergo rule (24) as
formulated. As will be argued in chapter 5 , where the geometry
of a ffrica te s is studied, the fa ct that the a ffrica te /ts / also
bears the feature [+continuant] blocks the application of the
P alatalization rule.

3.1.1. Lexical and Postlexical Rule Application

O ur a n a lysis of P a la ta liza tio n inhom org an ic clu ste rs
c ru c ia lly relies on the a p p lic a tio n of Nasal and Late ral
Assim ilation and [-cont] Spreading in Hom organic Clusters prior
to the rule of P alatalization. The palatalization of /t/ and /d /
after [h] and [x] has been made to follow from the existence of
hom organic clusters with a shared Place Node created by Nasal
and Lateral Assim ilation. The segm ent /d/ in these groups must
also have been previously assigned the feature [-cont] to be an
eligible target of Palatalization.
In chapter 2, it was shown that Palatalization is a stratum
I rule. The place assim ilation of nasals and laterals as well as
 185

the assignm ent of the [-cont] feature to /d/ in these clusters, on

the o the r hand, are processes th at operate across w ord-
boundaries and must, th ere fore , operate in the p o stle xical
module. It follow s that Palatalization must apply both after and
before these two other processes. This, rather than being a
contradiction, constitutes a not unexpected situation within the
theory of Lexical Phonology. It is an assum ption of this theory
th at the sam e rule can apply both in the Lexical and the
Postlexical M odules (cf., for instance, Mohanan (1986)). I will
conclude that N oncontinuant Spreading in Hom organic Clusters
and Nasal and Lateral A ssim ilation apply both at stratum I
(b e in g o rd e re d b e fo re P a la ta liz a tio n ) and p o s tle x ic a lly .
Palatalization, on the other hand, applies only at stratum I:

(29) Stratum I: Nasal and Lateral Assimilation

[-cont] Spreading in Homorganic Clusters
Palatalization
Postlexical: Nasal and Lateral Assim ilation
[-cont] Spreading In Homorganic Clusters

T his ord ering of the rules in lexical and p o stle xica l

derivations is illustrated in (30a) and (30b):
 186

(30)a. /IN D ar/

NLAssim. n
[-ct] Spread
Pal. nj
[ih ja r ] 'strength'

b. /m utiL Danak/
NLAss. 1
[-ct] Spread d
[m utiidanak] 'all the boys'

Since Palatalization does not apply postlexically, a form

like *[muti/Cj anak] will not arise. A clitic group like /m a rtin da/
'it is Martin,' on the other hand, will be sent back to stratum I
w here it w ill undergo P alatalization giving the surface form
[m a rtih je ].
The other two rules involved in C ontinuancy Assignm ent,
namely, After-Pause [-cont] Insertion and the Default Rule that
assigns the feature [+cont] will only apply postlexically. This
has to be necessarily the case with the first of these two rules,
since it must necessarily follow the insertion of pauses, which
takes place at the beginning of the postlexical component. The
final rule ordering is thus as in (31):
 187

(31) Rule Ordering:

Stratum I: 1. Nasal and Lateral Assim ilation
2. [-cont] Spreading in Homorganic Clusters
3. P alatalization
P o s tle x ic a l: 4. Pause Insertion
5. Nasal and Lateral Assim ilation
6. [-cont] Spreading in Homorganic Clusters
7. After-Pause [-cont] Insertion
8. Default [+cont]

(3 2 )/[a B e re ] / / [ e N B o r ] //[g iS o N ][B e r i] //[lo r e ][B e r i]//[ B e r i]/

L e x ic a l
NLAss m
[-ct]Spr b
P ostlex.
Pause H H H It It
In s e rtio n
NLAss m
[-ct] Spr b
After Pause b
Default b
1t[abere] 1t[ernbor] 1t[giSomberi] 1t[loreberi] 1t[beri]
'cattle' 'trunk' 'new man' 'new flower' 'new'
 188

Since Palatalization applies only lexically, its effects can be

undone in the postlexical component by Nasal and Lateral Place
A s s im ila tio n :

(33) /[m a r tiN ] [B a k a r ik ]/ /[ [m u tiL ] [n e s k a ] ak] ]/

L e x ic a l
A pplication
Palatalization n

P o s tle x ic a l
A p p lic a tio n
NLAss m
[-ct]Spr b
[m a rtim b a k a rik ] [mutilneskak]
' Martin only' 'boys and girls'

The le xica l a p p lic a tio n of Nasal and L a te ra l Place

Assim ilation also can bleed the application of Palatalization, as
in /egin-ko/ [eyr]go] (compare with /egin/ [eyh] in Ondarroa).

To sum m arize, in this subsection I have shown how the

apparently paradoxical conclusion that Palatalization m ust apply
both before other rules which create inputs for its application
and after those same rules, given the restricted m orphological
dom ain of application of Palatalization, can be handled by the
 189

L exica l P ho no log y M odel by m aking c e rta in rules apply

repeatedly in different components.
In th e fo llo w in g s u b s e c tio n s , I w ill p re s e n t the
p ala taliza tion facts in other Basque dialects to give a more
complete picture of this process in Basque. It will be shown that
P alatalization is essentially the same process on all Basque
d ia le c ts but a d d itio n a l c o n d itio n s are p la c e d on the
Palatalization rule with respect to the form ulation for Ondarroa.

3.2. A restrictive dialect: Baztan

In the High Navarrese dialect of Baztan, only a high front
glide, but not a high front vowel, can trigg er palatalization.
T argets are /n / and /I/; th a t is, the set of n on con tin ua nt
sonorant coronal segm ents. A fu rth e r condition for p a la ta l­
ization to apply is that the target must be followed by a vowel
(i.e. must be in an onset). The examples in (34a) and (34b) show
that /n / and /I/ are not palatalized after a front glide word-
finally or before a consonant, but undergo palatalization when a
vowel follows. As the examples show, palatalization causes the
absorption of the glide into the following palatal consonant (of.
chapter 4). The Baztan data are from Salaburu (1983):

(34) a. [seyn] 'which' b. [sehen] 'of which'

[a rts a y n ] 'shepherd' [a rtsa h e ] 'the shepherd'
[arayn] 'fis h ' [arahe] 'the fish'
 190
6
[er^o yl] 'ru s t' [er^O/Ca] 'the rust'

[er^oyldu] 'to rust (p)'

[e^keyntsen] 'offer (i)' [e ^keh i] 'offer (p)'

The examples in (35a) show that, unlike in Ondarroa, there

is no palatalization after a vocalic [i]. And the examples in (35b)
show that IXI does not undergo palatalization, again unlike in
Ondarroa, even if preceded by a front glide and followed by a
vo w e l:

(35)a. /s ik in - a / [s ik in e ] 'the dirt'

/ja k in - a / [ja k in e ] 'know n'
/u r in - a / [u rin e ] 'the sm ell'
/ la n g ile / [la n g ile ] 'w o rk e r'
/m u t il- a / [m u tile ] 'the boy'
b. [d eytu] 'to call (perf.)'
[a yte ] 'fa th e r'

Baztan palatalization is defined as in (36):

(36) Baztan Palatalization

Operation: Spread
Argum ent: [dorsal]
D irection: Rightw ards

6 As can be observed from the examples, and was mentioned In chapter 2,

there is some variation In the application of Vowel Assimilation after a palatal.
 191

Trigger Conditions: [+high], [-back], non-head

Target C onditions: [-continuant], [coronal], [+sonorant],
onset

The rule of Palatalization in Baztan differs from the same

rule in the dialect of Ondarroa in that it has three additional
co n d itio n s. One req uirem e nt is th at the ta rg e t m ust be a
sonorant, thus excluding IX/ and 161 from the rule. The other two
conditions refer to the prosodic geometry: the trigger cannot be
a syllable head and the target must be in an onset position.
The o p e ra tio n of P a la ta liz a tio n a u to m a tic a lly and
obligatorily causes the deletion of the glide. Assum ing that a
phonological rule can encode only a single operation, G lide
Absorption must be expressed by means of a different rule, as in
(37). This rule applies obligatorily in the Baztan dialect, and is
su b je ct to a ce rta in am ount of va ria tion in othe r Basque
dialects. The form ulation of the rule captures the fact that all
inputs for G lide Absorption are created by the application of
Palatalization by making reference to structures created by this
latter rule:

(37) G lide Absorption

Operation: Delete
Argum ent: leftm ost skeletal slot X
Target/trigger Condition: shared [dorsal] node
 192

Graphically, Glide Absorption is expressed in (37'):

(37') Glide Absorption

-O p f
T 1
R

Î 1 SL
P
r.
[dors] [cor]1

Salaburu (1983) points out that there is an additional rule

of P alatalization in the Baztan dialect. A fter w o rd -in itia l /i/,
and follow ed by a non-consonantal segm ent, /n / is palatalized.
The effects of this rule are apparent in a few gram atical forms
w here the initial [n] of question-w ords alternates with [h] in
q u a n tifie rs w ith a p re fixed /i/ (38a). In a d d itio n , su rfa ce -
initial [in] sequences are excluded, in favor of [in], (38b):

(38)a. [non] 'where' [ifion] 'somewhere'

[nor] 'who' [ihor] 'someone'
[noys] 'when' [ihoys] 'ever'
[nola] 'how' [ihola] 'somehow'
(38)b. [ihautiak] 'the carnival (pi.)'
[ihudia] 'the wet-nurse'
[ihera] 'swallow'
[in e ra ] 'a little (?)'
 193

There is every reason to treat these facts in (38) as being

produced by a different palatalization rule. First of all, because
in this case avowel /i/ triggers the process of palatalization,
unlike the case with non-initial vocalic /i/ (cf. the examples in
(35a)). Secondly, because only nasals, and not laterals are
targets. The exam ples in (39) show that a fte r initial /i/, a
lateral is not palatalized:

(39 ) [ilun] 'dark'

[ilia] 'the hair'
[ilua] 'the nephew, the grandson'
[ilebetia] 'the month'

Thus, in addition to the Palatalization rule in (37), there is

another palatalization rule affecting only the coronal nasal after
initial, vocalic /i/. The rule is form ulated in (40):

(40) W o rd -In itia l P a latalization

Operation: spread
Argum ent: [dorsal]
Trigger Conditions: [+high], [-back], adjacent to left-boundary
Target Conditions: [coronal], [+nasal]
 194

(40 ') ^

R
o— [nas] SL
,6—[cor] P
[dors]
[-back]
[+high]

Examples such as [gawihera] 'bat,' which is a compound of

[gaw] 'night' and [ihera] 'swallow ,' show that this rule, like the
main Palatalization rule, applies at stratum I. By w ord-initial
/i/ we m ust understand w o rd -in itia l before com pounding; in
fa ct, m o rp h e m e -in itia l.
Let us consider again the main process of Palatalization of
laterals and coronal nasals after front glides. Given the facts
that only glides trigger Palatalization in Baztan and that these
glides are then obligatorily deleted. Palatalization gives rise to
opaque forms. The trigger of Palatalization is never present in
surface forms. A consequence is that every intervocalic instance
of [/C] or [h] is interpretable as /il/ or /in/, including morpheme-
internal occurrences, where there is no alternation. In this way,
//C/ and /h/ can be com pletely elim inated from the underlying
phonem ic inventory of the dialect. The form s in (41), which
contain m orphem e-internal sonorant palatals, can be given the
u nd erlying re p re se n ta tio n s to th e ir left. The su rfa ce form s
 195

w ould then be obtained by the application of P alatalization

to g e th e r w ith G lide A b sorption (a fte r G lide F orm ation, cf.
above) :

[O/Co] 'chicken' / o ilo /

[te/Ca] 'tile ' /te ila /
[ée/Co] 'seal' /ë e ilo /
i/Ce] 'chair' /é\\\a/
[e^pi/Cu] 'm irror' /eép\\\u/
[€ohu] 'sound' / ë o in u /
[kehu] 's ig n ' / k e in u /
[eqgahatu] 'to deceive' /e n g a in a tu /
[éehale] 'signal' /€ e in a le /
[bahan] 'b u t' /b a in a n /
[etSeraho] 'up to the house' /e tS e - r a - in o /
[é ih e t^ i] to believe' /é iin e t^ i/
[ohase] 'to rm e n t' /o in a s e /

T h e re a re tw o p ro b le m s w ith th e s e u n d e rly in g
representations. The first one is that the sequence /iiC / that is
postulated for words such as [é\/Ce], [e^pi/Cu] and [€ ih e t€ i] is
otherwise un attested in Baztan Basque. Another problem is that
the vast m ajority of exam ples containing m orphem e-internal
p a la ta l s o n o ra n ts are lo a n w o rd s from R om ance, a lre a d y
borro w ed w ith the p a la ta l. The su rfa ce p a la ta ls do not.
 196

therefore, necessarily show the operation of a P alatalization

rule in Baztan Basque.
W ith the exception of the four final exam ples, all other
words in (41) are of Romance origin. This Is representative of
the readiness with which examples of one and the other origin
are found. These are not, however, all recent, unassim ilated
loanwords; on the contrary, in a number of cases the old age of
the borrowing can be dem onstrated with certainty.^
If we take these facts as arguing for the presence of
underlying palatal sonorants, an a lterna tive analysis for the
p la in /p a la ta l a lte rn a tio n illu s tra te d in (34) then becom es
available: Intervocalically, both plain [I], [n] and palatal [/(], [h]
can occur in the surface; witness pairs such as [katilu] 'bowl'
and [e^pi/Cu] 'm irror' or [Iana] 'the work' and [saha] 'the vein.'
W hen s y lla b le -fin a l in the surface rep re se nta tio n, how ever,
u n d e rly in g lAl and /h / are d ep ala ta lize d as [yl] and [yn]

^ The early date of the borrowing is apparent in the treatment of the initial
stop in [kehu] 'sign, wink' (of. Spanish ceho brow,' guiho wink' ). Examples such
as [te x a ] 'tile,' [espixu] 'mirror,' [ a r a x a ] 'crack, fissure' seem to have been
borrowed from the extinct Navarrese Romance. Castilian Spanish has non­
laterals in those items from the earliest attestations of the language; t e ja .
esp eio . raja , w here orthographic j_ represents [x], earlier pronounced as [§],
and as [2] in early Medieval Spanish. Navarrese Romance probably becam e
extinct soon after the occupation of the Kingdom of Navarre by Castilian troops
in the early sixteenth century, if [oxo] 'hen, chicken' is a borrowing (of. Spanish
polio), which is unclear, the unexplained deletion of the initial consonant would
also show the early date of the borrowing. The presence of many of these words
in the earliest Basque texts also guarantees this.
 197

re sp e ctiv e ly . F orm ally, th is d e p a la ta liz a tio n a n a ly s is w ould

involve the delinking of the [dorsal] node of a w ord-final /h / or
/X/ (giving [n], [I]) and its linking to its left to an inserted
skeletal slot, with autom atic creation of interm ediate nodes (of.
Archangeli and Pulleyblank (1986)), creating front glides:

(42) D epalatalization
X X ]s
o <j)— [-0 0 nt] R
o o — [+son] SL
o_ o [cor] P
^ q rs ]
[+him
[-back]
E.g.: /arah+e/ [arahe]
/a ra h / [arayn] by Depalatalization

In this analysis, Baztan Basque w ould have underlying

palatal / x / and /h/, a rule of initial palatalization, (40), and a
rule of final depalatalization, (42). In our first analysis of the
data all surface palatal sonorants were derived either by the
general rule of Palatalization after glide (37) or by W ord-Initial
Palatalization (40).
There is some evidence both internal to the Baztan dialect,
and com parative with other dialects, to m aintain our first
 198

analysis (the Palatalization analysis). The evidence is, however,

not overw helm ing.
The evidence internal to the dialect has to do with the
interaction between the palatalization process and the rule of
Vowel A ssim ilation, which raises /a / to [e] after vocalic or
non-vocalic /i/. As indicated in chapter 2 and footnote 5 of this
ch ap te r, a fte r palatal consonants there is va ria tio n in the
application of this rule of Vowel Assim ilation; i.e., we find both
[a ra h a ] and [arahe] 'the fish,' from /a ra in + a /; [gaha] and [gahe]
'the top,' from /g ain-a/. This variation was a ttributed to the
possibility of ordering Palatalization and Vowel Assim ilation in
different ways in chapter 2. This is also the explanation given by
Salaburu (1983, 169-70). On the other hand, if we postulate that
the palatals are found in the underlying representations, i.e.
/a ra h -a /, /g ah -a/. Vowel A ssim ilation should never apply to
such examples, since at no stage is there a triggering /i/. This
is an argum ent for the Palatalization analysis. The argum ent,
however, loses its force, once we consider the alternative of
having p a la ta l co n so n a n ts as o p tio n a l trig g e rs of Vow el
A ssim ila tio n . In th at case, we can have underlying palatal
sonorants.
The fa c t th a t o th e r Basque d ia le c ts p re se n t surface
palatals in other additional contexts, and that in these dialects
(e.g. Ondarroa) only a palatalization analysis is sensible, would
also lead us to postulate a palatalization rule for Baztan. From a
 199

com parative point of view, it makes sense to assume that both

B aztan and O ndarroa, and d ia le c ts in betw een, have a
Palatalization rule. This rule would have spread throughout the
Basque area, becoming less constrained in some dialects.
C o n s id e r now th e e v id e n c e fro m d ia le c ts th a t,
demonstrably, have lost the surface palatals that they once had.
There are dialects (e.g. Labourdin, Low Navarrese) where
su rfa ce p ala tals are e xtre m ely rare o utsid e of d im in u tive s
(a ffe ctive p ala taliza tion ). In these d ia lects Rom ance /xV, /h/
are represented as [il], [in]; e.g. [boteyla] 'bottle* (of. Sp.
[botexa]), [maynu] 'bath' (of. Sp. [baho]), [oylo] 'hen' (of. Sp. [poxo]
'chicken'), etc. (cf. Michelena (1985, 195-6)). Other instances of
m orphem e-internal palatal sonorants in dialects such as Baztan
also find the same correspondence; Baztan [bahan] 'but,' Low
Navarrese [baynan]; Bz. [gahera] 'm oreover,' LN [gaynera], etc.
An interpretation of these facts is that these dialects at an
earlier stage had a rule of palatalization after front glide and
before another nonconsonantal segm ent which gave surface
form s sim ilar to those of Baztan. Then the palatalization rule
w as lost, causing not only the loss of the p la in /p a la ta l
consonant alternation in forms such as [seyn] 'which' vs. [sehen]
'of which,' (which became [seynen]), but also making underlying
form s such as those proposed in (41) surface w ith plain
consonants preceded by a front glide. In this analysis we must
make the assum ption that all instances of palatal sonorants,
 200

in clu d in g th o s e in b o rro w in g s, had been re a n a lyze d as

re p re s e n tin g u n d e rly in g /in /, / il/ b e fo re th e ru le of
Palatalization was lost.
A gain, an a lterna tive analysis w ould be to claim that
dialects w ithout palatal sonorants, such as Labourdin, used to
have a s y lla b le -fin a l d e p a la ta liza tio n rule identical to (42),
w hich w as la te r sim p lifie d by e lim in a tio n of the pro sod ic
condition. In this w ay d ep ala ta liza tion w as extended to all
instances of /h/, /X/ regardless of their position in the syllable.
Interestingly, the most parsim onious analysis, which would
p o s tu la te u n d e rlyin g p a la ta ls in n o n -a lte rn a tin g in sta n ce s
(m orphem e-internally) and derived palatals only in cases of
a lte rn a tio n (m o rp h e m e -fin a lly ), w o u ld not e x p la in the
d e p a la ta liz a tio n observed in Labourdin and Low N avarrese
d ia le c ts .
The evidence for or against a Palatalization analysis, as
opposed to a Depalatalization analysis, in Baztan seems to me
inconclusive. I w ill assume that there is a Palatalization rule,
although granting that this analysis is not uncontroversial.

3.3. An intermediate dialect: Donostia-San Sebastian

T h e re are d ia le c ts w h e re the a p p lic a tio n of the
palatalization process is not as extensive as in Ondarroa but not
as restricted as in Baztan. This is the case of the dialect spoken
in the Guipuscoan capital. In this dialect, target segm ents are
 201

/n/, /I/ and optionally /t/ and both syllabic and non-syllabic /i/
trigger palatalization. As in Baztan, the target must be followed
by a nonconsonantal segment. The exam ples in (43) are from
Iverson and Ohederra (1985):

(4 3 ) [ il] 'die' (perf.) [ix e ta ] 'fu n e ra l'

[sa yltad u n] 'd if f ic u lt y ' [saxa] 'd i f f i c u l t '
[xa kin ] 'to know' [xakiha] 'know n'
[alegindu] 'to try'(p e rf.) [alegiha] 'e f f o r t '
[n ola ba yt] 'som ehow ' [nolabayt'eko] 'of some sort'
[nolabayteko]

Some conditions on both the trigger and the target of the

rule of palatalization are removed in this dialect with respect
to the more restrictive Baztan dialect seen above. Since both
syllabic and non-syllabic /i/ can trigger palatalization, prosodic
conditions on the trigger need not be present in the rule. As for
the ta rg e t, the rule a p p lie s o b lig a to rily to n o n co n tin u a n t,
sonorant coronal segments (e.g. /n/ and /!/) in an onset, but the
feature [+sonorant] can optionally be removed from the rule, thus
accounting fo r the optional palatalization ot /t/. The segm ent
161 never undergoes the rule. As Indicated in the discussion of
the Ondarroa facts, and also suggested by Iverson and Ohederra,
161 it is not a possible target of the rule after [i] or [y], because
in this position it is continuant. As for /ind/, /ild / and also
 202

/int/, and /ilt/, the prosodic requirem ents on the target block
the application of palatalization in this dialect.
Basque dialects thus differ in w hether some or all of the
follow ing conditions are added to a rule that is m inim ally as
form ulated for Ondarroa in (24):

(44) trigger: non-head

target: onset
target : [+sonorant]

3.4. A ffective Palatalization

From the c o n te x tu a lly c o n d itio n e d p a la ta liz a tio n rules
discussed in this chapter, we must distinguish the process of
Affective Palatalization. This is a process w hereby a more or
less large subset of the coronal segm ents (depending on the
dialect), become palatals. The process may apply at the word
level, in which case it is used to form dim inutives; or, at the
utterance level, in expressive speech, e spe cially directed to
children. The process is particularly active and productive in
dialects where the rule of conditioned palatalization studied in
the im m ediately preceding sections does not operate, as in
Labourdin (of. Lafon (1952)), or is very restrictive, as in Baztan.
In Biscayan dialects, such as Ondarroa, where contextually
conditioned palatalization is very pervasive, there is no process
of affective palatalization. Iverson and Ohederra observe that in
 203

the variety of Donostia (San Sebastian), the only segments which

undergo affective palatalization are the sibilants, which are not
affected by conditioned palatalization after /i/, and /tV, which
is only an optionally target of conditioned palatalization.
In dialects such as Baztan, the Affective Palatalization of
the coronals produces the follow ing changes (cf. Salaburu
(1983), Salaburu and Kintana (1984)):

( 45) P la in A ffe c tiv e

Z [s] — > 2L[S]

S.[é] —> X [5]

[ts] —> i)L m

Is. [t^] —> ÎX m

i —> 11 [ f ]

d —> û d [j]
I —> 11 [X]

a —> Ü

I —> Ë. 0

The only co ro na l th at is never a ffe cte d by A ffe ctive

P a latalizatio n is the trill [r ]. The other rhotic, the flap [r],
shows a variable behavior; it may become [j] or [X], it may be
deleted, or it may be unaffected. Exam ples of d im inu tive s
created by Affective Palatalization are given in (46):
 204

(46) Plain Diminutive

zezen xexen 'bull'

oizon oixon 'man'

m m 'mouse'

laaun llaaun 'f rie nd '

D o lita D o llitta 'p r e tt y '

P a latalized item s can be em ployed in o the rw ise plain

d isco u rse to convey a d im in u tive , a ffe ctio n a te , or scornful
meaning. In this respect, affective palatalization produces very
much the same effects as dim inutive affixes in a language like
Spanish. From this point of view, it is sensible to interpret a
word such as x e x e n 'little bull' as carrying a feature-size
p alatal m orphem e (an a utosegm ent) w hich is se m a n tica lly
e qu iva le n t to the affix -il- in the Spanish word to r ito 'little
bull' (of. to ro 'bull'). If the word contains palatals in its plain
form , these are not affected in any m anner by A ffective
P a la ta liz a tio n . The d im in u tiv e form of w ord s co n ta in in g
palatals (and not coronals) in their plain form such as e tx e
'house,' is formed by suffixation: e tx e tto 'little house.'
As m entioned. A ffective P alatalization may a ffect w hole
u tte ra n c e s . T his is c h a ra c te ris tic of c a re ta k e r ta lk . The
exam ples in (47), from the Baztan dialect, are from Salaburu
(1983, 118):
 205

(47)a. Plain: zer da

Affect. : xerdda, xedda 'what is it?
b. Plain: tortzen bazara
Affect. : ttortxen baxara 'if you come'
c. Plain: otz iten du
otx itteh ddu 'it is cold'
d. Plain: zazi ta erraiozu tortzeko
Affect. : xaxi tta erraioxu ttortxeko 'go and tell him to
come here'

U nderlying p alatals or palatals created by co n te xtu a lly

c o n d itio n e d p a la ta liz a tio n do not block the a p p lica tio n of
A ffective Palatalization to palatalizable coronals:

(48) Plain: zehendako for whom

Affect.: xehehddako

A ffective Palatalization can receive essentially the same

analysis as that provided by Cole (1987) for the sim ilar process
of G lottal Harm ony in Coeur d'Alene. I w ill assum e that the
palatalizing feature (which I will take to be [+high]) is situated
on a d iffe re n t plain, and is linked by the rule of A ffective
Palatalization to all coronal consonants in the dom ain (except
the trill and optionally the flap):
 206

(49) Otz iten du

Il UN II
X X XXXX XX

^ ^ [ + h ig h h " '^

4. Conclusion
In this chapter we have studied a set of interrelated rules
in Basque. Som e of these p rocesses (N asal and Lateral
A ssim ila tio n and C ontinuancy A ssig nm e nt) are shared with
S panish and have been the o b je c t of se ve ra l d iffe re n t
fo rm u la tio n s . I then show ed how the a n a lysis of these
processes, which can be attained by considering the geometrical
configuration of segments, is superior to previous form ulations.
It was also dem onstrated that the m anipulation of structures
assumed to be involved in these processes finds confirmation in
the way a rule of Palatalization selects and affects its inputs in
certain Basque dialects (Ondarroa). It was then argued that the
rule of Palatalization, a lexical stratum I rule, forces a lexical
a p p lica tio n of Nasal and Lateral A s s im ila tio n and a rule
assigning a value for continuancy to voiced o bstru en ts in
hom organic clusters, besides their postlexical application.
O ther problem s that the palatalization process raises In
different Basque dialects were also explored.
 207

Chapter 5
The Structure and Behavior of Affricates

0. Introduction
The complex behavior of affricates in Basque with respect
to a number of phonological rules justifies a detailed study of
th e ir p h o n o lo g ic a l s tru c tu re d In this chapter I will present
such a study. I will argue that affricates bear both the features
[-cont] and [+cont], which permits their grouping with stops for
the purposes of some phonological processes and with the
sibilant fricatives for some other rules. A common assum ption
is that the stop and continuant phases of an a ffrica te are
phonologically ordered. This view is expressed in an articulated
and reasoned m anner in Sagey (1986). Sagey argues that a
consequence of the ordering of features in the affricates is that
they will show "edge effects"; that is, they will be seen as
stops for processes in which the trigger or target is to the left
of the affricate (the [-cont] side) and as fricatives by rules
involving the right, [+cont], side of the affricate.
A ffrica te s in Basque do not show "edge effects" for a
number of phonological processes. In Basque, w hether affricates
w ill p a rtic ip a te in p ro ce sse s a ffe c tin g [-c o n t] or [+ cont]
segments cannot be determined from the position of trigger and

^ Som e of the m aterial in this chapter was first presented at W C C F L 6

(Hualde, 1987) and W C C FL 7 (Hualde, 1988).
 208

j ta rg e t of the rule. I w ill claim th at the two co n tra d icto ry

sp e cifica tio n s fo r the feature [cont] that a ffrica te s bear are
p h o n o lo g ic a lly unordered. The p re d icta b le o rd ering [-cont],
[+ 0 0 nt] is not encoded in the underlying structure of these
segm ents.

1. Affricates in Generative Phonology

The representation of affricates, like that of other non­
steady-state segm ents, has long been problem atic (cf. Ewen
(1982) fo r d iscu ssion ). In a linear system of phonological
re p re se nta tio ns, the decision has to be made w h e th e r to
re p re s e n t a ffric a te s as sequences of tw o e le m e n ts ,
corresponding to their two phonetic phases, or to ignore their
com plex articulation and treat them as simple single segments.
T ra n scip tio n s such as [tS], [d%] w ould seem to suggest a
■ conception of affricates as sequences of stop and fricatives,
whereas [6] and [ j] suggest a unitary conception of affricates.
St. Clair (1972; 1973) makes an attem pt at capturing the
fa c t th a t a ffric a te s p re se n t c h a ra c te ris tic s of both stop-
frica tive sequences and of single segm ents. For St. Clair,
affricates start out as two segments and are fused into one by a
transform ation that he calls the Compound Segment Convention.
In his 1973 article, he form ulates the C om pound Segm ent
Convention as follows:
 209

(1) [X] [Y] --> [ [X] [Y] ]

12 1+2

A problem with this derivational approach to affricates is

that there are languages that possess affricates w ithout having
in their inventory a corresponding fricative, contrary to w hat St.
Clair assumes. For instance, Spanish has [tS] but not [S]- In a
language like Spanish, it does not make any sense to derive [tS]
from an underlying sequence of [t] plus [S].
In a linear fram ework employing binary features, affricates
must be characterized either as stops by assigning to them a
feature such as [-cont], or as fricatives by giving them the
opposite value for this feature. Crucially, in a fram ework such
as SPE, affricates cannot receive both values for the feature
[cont] as the complex nature of their articulation would seem to
require, since they are given only one matrix. In the SPE system,
a ffric a te s are g ro u p e d w ith the sto ps as [-c o n t]. This
c h a ra c te riz a tio n im p lie s th a t a ffric a te s are e x p e c te d to
participate in phonological processes affecting stops and not in
processes affecting fricatives.
To distinguish affricates from other [-cont] segments, they
are given the feature [+delayed release] in the SPE system. This
fe a tu re has fre q u e n tly been used to uniquely ch a ra cte rize
a ffrica te s (cf. Hyman (1975), for instance), thus being the
equivalent of a [+affricate] feature.
 210

M ultitiered phonological frameworks provide the means for

a more conspicuous representation of affricates. In a framework
where feature matrices are linked to a CV tier, affricates can
be represented as two m atrices lined to a single C slot (of.
Clements and Keyser (1983)):

(2 ) 0
/ \
t ^

R epresentations such as that in (2) are still less than

satisfactory in that they do not capture the fa ct that the two
m atrices w hich, in this view , are contained in an affricate,
typically share all features, except for the [continuant] value.
W ithin the hierarchical model of phonological represent­
a tions, a ffric a te s can be represented as sing le segm ents
containing two specifications for the feature continuant. This
representation is adopted in Sagey (1986).
Sagey (1986) distinguishes two types of segm ents with
m ultiple articulations: contour segm ents which possess ordered
a rticu la tio n s and com plex segm ents, w hose a rticu la tio n s are
unordered. The a rticu latio ns will be ordered If the segm ent
contains incom patible features on the sam e tier. A ffricates,
which bear the two incom patible features [-cont] [+cont] are
th us co n to u r segm ents. As m entioned above, given this
 211
I

I representation, Sagey makes the prediction that affricates w ill

■ behave as stops fo r rules Involving th e ir left edge, but as
frica tive s for rules sensitive to th eir right edge. I w ill now
exam ine the behavior of a ffrica te s w ith respect to several
processes in Basque. In every case to be examined, the position
of the environ m e n t of the rule w ith respect to the ta rg e t
■ segm ent appears to be irrelevant in determ ining w hether the
affricate will or will not undergo the rule. It will be shown that
Basque presents both cases of overapplication of rules to
affricates with respect to their conception as contour segm ents
(i.e. cases w here affricates are not predicted to undergo the
rule, but they do) and cases of underapplication (i.e.. affricates
are predicted to undergo the rule and do not). The conclusion will
be that affricates are not contour segments.
I w ill firs t show th at a ffrica te s in Basque are single
segments and not mere sequences of stop and fricative.

2. Affricates are single segments

That affricates constitute single segm ents in Basque and
not mere sequences of two segments can be concluded from a
num ber of distributional facts.
! A first piece of evidence is given by the fact that groups
; such as Qs are not allowed in Basque.
If we consider the consonant sequences that may appear in
; a coda (cf. chapter 1, section 2.3.3), we also must conclude that
 212

affricates are single segm ents. Were affricates to be analyzed

as sequences of two segm ents, the sonorant-affricate groups
that appear in items like beltz 'black,' antz 'resem blance,' would
constitute the only examples of CCC clusters in a coda. T h is fact
suggests that affricates are single segm ents attached to only
one skeletal slot.
The same conclusion can be gathered from the examination
of onset clusters. The only consonant clusters that are allowed
in an onset in Basque are those formed by an oral stop and a
liquid (cf. ch ap te r 1, section 2.3.3). W ere a ffrica te s to be
considered sequences they would constitute rather anom alous
c lu s te rs .
It is interesting to note in this respect that, in some
d ia le cts, the sequences /ks/, /p s / in Latin and Rom ance
borrow ings are changed to an a ffrica te , as in e ts e m p I u
'e x a m p le ,' e ts a m in 'exam ,' s e ts u 'sex,' a ts o lu tu 'a b s o lu te '
(examples from Txillardegi (1980, 346-7)). In other dialects the
cluster is reduced to a fricative. In closely connected phrases
(between pronoun and conjugated verb) the change of [ks] into
[ts] across word boundaries constitutes a synchronic process in
the Baztan dialect, e.g.: /onek suen/ [onetsuen] 'this one had it'
(cf. Salaburu (1984, 279)). Why should â group [ts] be preferred
over [ks]? The answer seems to be that [ts] is not a group but,
rather, a single segment.
 213

I w ill now exam ine a num ber of processes th a t bear

directly on the structure of affricates. The first process to be
exam ined w ill be Stop D eletion, a rule th a t show s quite
co n vin cin g ly that affricates bear both values for the feature
[cont] and th a t the phonetic o rder of these values is not
phonologically relevant.

3. Stop Deletion
In, to my knowledge, all Basque dialects, a deletion rule
sim plifies sequences of two stops, where the first one is an
obstruent, by deleting the first stop in the sequence. This rule
applies o bliga to rily to word internal stop sequences and also
across word boundaries in certain syntactic contexts in rapid
speech. Examples are given in (4) (most examples from Salaburu
(1984, 286-87));

/bat paratu/ [baparatu] 'put one'

/bat traban/ [batraban] 'one stuck'
/bat kurri/ [bakurri] 'run one'
/guk pistu/ [gupistu] 'we light'
/guk kendu/ [gukendu] 'we take away'
/bat-naka/ [ban aka] ‘one by one*
/bat-batean/ [bapatean] 'at once'
 214

The clu s te r sim p lifica tio n rule also a ffe cts a ffrica te -
stop sequences. In this context, affricates do not delete, but
become fricatives. That is, they lose their noncontinuancy:

(5) /its - te g i/ [is te g i] 'd ic tio n a ry '

/its - k e ta / [is k e ta ] 'c o n v e rs a tio n '
/a r its - k i/ [a ris k i] 'oak wood'
/ a r it s - m e n d i/ [a ris m e n d i] 'oak mountain'
/o ts -tu / [o stu ] 'becom e cold'
(p e rf.)

Fricatives are not affected by this rule, and fricative-

stop se q u e n ce s are fre q u e n tly fo un d both m orphem e-
internally and across morphemes:

(6) /e ë k a / [e^ka] 'ask for'

/d a ^ ta / [d a^ta] 'ta s te '
/b e g i- s - ta / [begista] 'su rve y'

If affricates are characterized as [-cont, -del rel] it is

not possible to account for the process by means of a single
sim ple rule. We need two rules a pplying in the same
e n viro n m e n t: a rule dele ting stops, e xcluding a ffrica te
segments, and another rule changing affricate segments into
frica tive s :
 215

(7) a. ■cont
■son
■del rel — > 0 / [-cont]

4-cont

[4-del rel] — > -del rel [-cont]

Yet, both rules in (7) account for w hat arguably is a

single process, namely, the suppression of an oral occlusion
before a second occlusion. The consideration of affricates as
a type of stop would force us to make use of two separate
ru le s to account fo r one s in g le p ro c e s s o f c lu s te r
s im p lific a tio n .
In Hualde (1987) I argued that we can account for these
facts by means of a delinking rule, which in a sequence of
tw o a d ja c e n t s e g m e n ts w h e re the fir s t s e g m e n t is
characterized by the features [-cont, -son] and the second
segm ent also bears the feature [-cont], delinks the node
dom inating [-cont] in the first segm ent (see Hualde (1987),
for argum ents that the each of the [cont] values of an
affricate hangs from a different node): ^

2 In Hualde (1 9 8 7 ), I assum ed that the node that directly dom inates the
feature [cont] is the supralaryngeal node. The analysis can be m aintained
essentially unaltered under the assumption that [cont] is directly under the
root node. The location of the feature [cont] is explored in the following
section.
 216

(6) Stop Deletion

Operation: Delink
Argument: R node dominating [-cont]
Target: [-son] [-cont]
Trigger: _____ [-cont]

This rule w ill delink all features of a stop, but w ill

leave all the features of a frica tive if the input is an
affricate, if the representations in (7) are adopted:

(7) stop-stop sequence affricate-stop sequence

X X X

R
K
Fell [-ct] [+ct]
SL
P I I

Notice that if we were to consider affricates as mere

sequences, this process would violate the locality condition
on phonological rules in its application to affricates, since
the elem ent that is deleted and the trigger of the rule would
seem to be separated by a sibilant fricative (see chapter 1).
The question would be why fricatives can intervene between
 217

trig g e r and target, even though other segm ents, such as

vow els ca n n o t; e.g., w hy th ere is d e le tio n in /a rits -k i/
[ariski] but there is no deletion in /m utlko/ *[muiko]. The only
possible answ er is that the process is possible In the first
exam ple because affricates are single segm ents, as we also
concluded from distributional facts.
A ffricates must be characterized by the feature [-cont],
since they undergo the rule. They must also be characterized
by the feature [+cont], because they become fricatives by the
application of the rule. The phonetic order of the two features
appears as irrelevant: the trigger of the rule is on the right,
[+ 0 0 nt], edge of the affricate but it is its left [-cont] branch
that undergoes delinking (there are no edge effects). This is
thus a case of overapplication of a rule to affricates from the
point of view of the contour hypothesis. Under the hypothesis
th a t a ffric a te s a re c o n to u r s e g m e n ts w ith o rd e re d
a rtic u la tio n s , the process w ould s till seem to apply to
a ffric a te s in vio la tio n of lo c a lity even if a ffric a te s are
single segments, if feature adjacency is required (observe the
rep re se nta tio ns in (7)). Sagey's hypothesis that a ffrica te s
will universally show edge effects is based on the assumption
that feature adjacency is necessary. The explanation that the
facts exam ined here received in Archangeli and Pulleyblank
(1987) and Hualde (1987) was that in this case only segment
adjacency, and not feature adjacency, is required. In Hualde
 218

(1988), instead, it was argued that locality is respected in

the application to affricates of Stop Deletion, because the
tw o a rtic u la tio n s of an a ffric a te are p h o n o lo g ic a lly
unordered. In the follow ing subsections, other processes in
Basque phonology which affect affricates are studied. In no
case does the order of surface a rtic u la tio n s , o cclu sion -
fricative release, condition the behavior of affricates. The
phonetic order appears as irrelevant for the phonological
component. No rule of Basque needs to refer to the ordering of
a rticu la tio n s of a ffrica te s.
In Hualde (1987), I assumed that the feature [continuant]
hung from the supralaryngeal node. In the representations in
this dissertation, however, the feature [continuant] has been
situated directly under the root node. Both positions for the
location of this features have been taken in the literature. It
seems necessary to determ ine this issue, before continuing
our exploration of the behavior of affricates in Basque.

4. Excursus: Where is the feature [continuant]?

In the "hierarchical" literature there is a certain amount
of variation in the location of the continuant feature within
j th e h ie ra rc h y . C le m e n ts (1985) and A rc h a n g e li and
; P ulleyblank (1986) locate [cont] under the supralaryngeal
• node, w ith o u t offering evidence fo r this position. Sagey
j (1986), on the other hand, situates [cont] directly under the
 219

root node. Her m otivation com es from com plex segm ents
w here the [cont] specification must be J in k e d with specific
articulators. By situating [cont] under the root node she can
then establish the necessary links between this feature and a
given place node.
A convincing argum ent for placing [cont] directly under
the root node, in my opinion, is that the feature [+/- cont] can
be used to distinguish the glottal stop [? ] from [h], two
segm ents with no supralaryngeal features, as argued in Lass
(1976), and Steriade (1987) among others. As Lass (1976)
shows, voiceless stops are often reduced to glottal stops by
loss of the supralaryngeal gesture, whereas [h] often results
from the weakening of voiceless fricatives. The reduction of
v o ice le s s stops to the g lo tta l stop is a w ell atte ste d
phenomenon in several English dialects. The examples in (8)
are from Lass. The items in (8a) are from a New York City
variety, whereas those in (8b) are from Lowland Scots :
(8)a. ['sen?cis] 'sentence'
['wer)?r)] 'one can'
['ôœ ?w sn] 'that one'
b. [to ? ] 'to p '
[ka?] 'c a t'
[ba?] 'back'
[,te? 0 ,s1?] 'take a seat'
 220

The process can be represented as in (9):

(9)
R
[-cont]
L
[-voice]
SL
P

The w eakening of fricative s into [h] is a very well

. attested phenomenon in many languages. The examples in (10)
are from the Spanish dialect of Coria in Extremadura, Spain
(Cummins (1974)) . In this dialect /s/ and /0 / (orthographic q,
' and zj, in fact, all vo ice less fric a tiv e s , in a coda are
weakened to [h], and then lost under certain conditions (see
. Hualde, to appear, for details):

(10) [doho treh] dos o très 'two or three'

[m aharfba] mas arriba 'further up'
[em bohalta] en VÔZ alta 'out loud'
[abihpa] avispa 'wasp'
[e hkw é la] escuela 'school'
 221

This weakening process is captured as in (11):

(11)
R
[+cont]

[-voice]
SL
P

A fter the supralaryngeal node is pruned in the English

and Spanish reduction processes, the resulting representation
would be as in (12a) and (12b) respectively for [?] and [h]:

(1 2 )a .[? j b. [h]

R
[-cont] [+cont]

[-voice] [-voice]

Only by representing [cont] under the root node can we

capture the phonetic and phonological behavior of purely
laryngeal sounds. If [cont] were under the supralaryngeal node,
 222

i degree of occlusion sp e cifica tio n s w ould be lost in the

: weakening processes exemplified above and we would not be
able to capture the difference between / ? / and /h/ or their
! re la te d n e ss to v o ice le ss stops and v o ic e le s s fric a tiv e s ,
respectively. Instead of the results in (12), we would have an
am biguous representation, as in (13), fo r both a purely
laryngeal stop and a laryngeal fricative:

(13) X
R
L ^
[-voice]

We w ill now proceed with the study of phonological

processes in Basque that affect affricates.

5. Sibilant harmony
A ffricates and fricatives are grouped together in Basque
with respect to a morpheme structure constraint that requires
; all sibilants (affricate and fricative) in a morpheme to share the
same point of articulation (cf. M ichelena (1985), Salaburu
(1983)). W hether an affricate Is preceded or followed by another
affricate or fricative sibilant, agreem ent in place features is
equally required. Examples, in orthographic representation (s [^],
: z [s], X [S]) are given in (14) (from Salaburu (1983,80)):
 223

(14) asots, eltsuntse, urtxintx, samats, zuz’en, zezen, azazkal,

zim itz, sasoin, eskasi, osasun, zintzur, itseso, sasi, zozo,
SOS, zorrotz, izotz, zortzi, zize, zizeri, zizel, zurruzte,
za pe latz.

Salaburu claim s that no counterexam ples are found, at

least in the Baztan dialect, which he examines. This constraint
is re m iniscen t of the s ib ila n t harm ony processes found in
Chum ash (cf. Poser (1982)) and Navajo (cf. Yip (1988)). A
difference is that, in Basque, sibilant harmony is only observed
within the morpheme. Across m orphem e-boundaries, harmony is
quite freely violated, e.g.: s o r-tz e n 'come out (im p.).' Another
interesting difference between the Basque and the Chumash and
Navajo processes of sibilant harmony is that Basque possesses
three harmonizing sibilants instead of the two found in Chumash
and Navajo.
I will suggest that the m orphem e-internal sibilant harmony
of Basque must receive a treatm ent essentially identical to that
given by Pulleyblank (1988) to m orphem e-internal restrictions
in the cooccurrence of vowels in languages such as Tiv.
Based on distibutional gaps, Pulleyblank (1988) proposes
that in Tiv a morpheme can bear only one [+round] specification.
This autosegm ent will link to the first vowel in the .morpheme
by u niversal le ft-to -rig h t associa tion co nve ntio n s: .T h en the
feature spreads to the other vowels in the root: '
I 224

! (15) P o o
I I I
lab lab
I
[+rd]

' Adapting Pulleyblank’s analysis of Tiv to the Basque facts,

I w ill claim th a t the fe ature [d istrib u te d ] is a property of
morphemes and not of segments. A morpheme may bear a feature
[+distr] or be unspecified for that feature. If a m orphem e
contains a feature [+distr], it w ill be linked to the leftm ost
; s ib ila n t by u n iv e rs a l a s s o c ia tio n c o n v e n tio n s , since o nly
sibilants need or can be specified for this feature.
Given the fact that the only use of the feature [distr] in
Basque is to distinguish different sibilants, I w ill assum e that
[distr] docks on the feature [strident]. The latter can be taken as
a m onovalent fe a tu re c h a ra cte rizin g s ib ila n t fric a tiv e s and
a ffr ic a te s :

(16) z i n tz u r
X X X X X
R t ; 1 r. 1
1
0
f1
0
SL
/
o o o
?
! i i
[strid] [strid]
[+distr]

The feature [+distr] is then spread rightwards by rule: ^

 225

(17) Distributed Spreading

Operation: Spread
Argum ent: [+distr]
D irection: Rightwards
T rig ge r/T arge t C ondition: [strid]

(18) z i n tz
X X X X
1 / \
R I 0 0 0 0
1 1 1
1 1
SL 0 0 0
1
[strid] [strid]

[+distr]

The sibilants in words such as s o s will receive a default

[-d is tr] specification. Finally we need to consider m orphemes
containing palatal sibilants. To d istinguish among the three

3 From regular sound correspondences, we w ould e xp ect * fra n tze s

'F re n c h ,' with d isagreeing sibilants, corresponding to S panish trances
[fra n G é é l. In fact we find the harmonized fra n ts e s . even though the name of
the country is F rantzia France.' In the case of fra n ts e s it has been the [ distr]
rightmost sibilant that has retained and spread its feature. If more exam ples of
this type are found, our assumptions in the text about the direction and the
argument of the rule of sibilant harmony would have to be modified. But, since I
cannot think of any other exam ples that show the active application of sibilant
harm ony (Distributed Spreading) in the adaptation of borrow ings, I will
m aintain w hat seem s the unm arked option (at least with respect to initial
linking).
 226

pairs s /ts . z /tz . and x /tx we need at least two features. I have
suggested th a t one of these two fe atures is [strid]. I w ill
assume that the other one is the feature [anterior]. By means of
these two features, we can characterized the three sib ila n t
pairs of Basque as in (19):

(19) s /ts z /tz x /tx

[distr] - + +
[ant] + - +

The fourth possibility, i.e. [-distr, -ant] does not arise. If a

segm ent is [-ant] it w ill be [+distr]. Again, to capture this
dependency we can make the feature [-ant] depend hierarchically
on the fe a tu re [+ d is tr], at lea st fo r se gm en ts th a t are
c h a ra cte riz e d fo r this fe a tu re (cf. M ester (1987), fo r an
approach that makes extensive use of structural dependencies
between features). The rule of Distributed Spreading will apply
in an example such as u rtxin txa as follows:
 227

(20) u r tx i n tx
X X X X X X
1 1 /\ 1 1 / \
R 0 0 0 0 0 0
SL
1
0
1
0 0
/
0
1
0 0
/
1
[strid] [strid]

[+distr]

[-ant]

This Is thus an instance of a process that groups affricates

and fricative s together regardless of the relative position of
trigger and target of the rule. The surface contour articulation
of affricates is simply irrelevant for this process.

6. F rica tivizatio n
There is a rule that transform s a sibilan t plus affricate
sequence into a sibilant plus stop cluster: s t^ becomes The
effects of this rule are limited to verbal morphology, but some
of the most productive verbal suffixes start with an affricate
' and quite a few verb roots end in a sibilant. Verbs which take an
z i or - tu e nd in g in the p e rfe c tiv e form c o n s tru c t the
im perfective by means of a suffix - tz e n . This suffix becomes
- ten when preceded by a sibilant. Compare the examples in (21a)
 228

with those in (21b). In the examples in (21b) the verbal root ends
in a sibilant:

(21 )a. perfective im p e rfe c tiv e

ip in - i ip in tz e n 'p u t'
ib il- i ib iltz e n 'w a lk '
s a r - tu s a rtze n 'e n te r'
e s k a -tu eskatzen 'ask'
b. ik a s - i ik a s te n 'le a rn '
ik u s - i ik u s te n 'see'
ira b a z - i ira b a z te n 'earn, win'
h a s -i hasten 'begin'
h a z -i hazten 'g ro w '

If the verb root ends in an affricate, Stop Deletion will also

apply. The effects of both rules combined are to transform an
a ffric a te -a ffric a te sequence ts -ts into st:

(22) ir a k a ts - i ira k a s te n 'te a ch '

itx - i ix te n 'open'
h a u ts -i hausten 'break'

The rule of fric a tiv iz a tio n also a ffe c ts the re la te d

nom inalizer -tze , to which determiners and case endings can be
added. Corresponding to ib iltz e a ’the walking (abs.).' ib iltz e k o
L_
 229
i
I 'for w alking,' ib iltz e k e 'w ithout w alking,' etc., with the sibilant-
I final Ik a s -. we obtain ikastea 'the learning (abs.),' ik a s te k o 'for
. learning,' ik a s te k e 'w ithout learning,' etc.
I N on -re la te d ve rb al su ffixe s are affe cte d in the same
manner. The agentive -tz a ile gives a n to la tz a ile from a n to la -tu
'organize,' but m o z ta ile 'cutter' from m o z -tu 'cut.' The verbal
suffix -tz a p e n 'action or result' gives o a ltz a o e n 'loss' from a a l-
du 'lose;' but hastapen 'beginning' from has-i 'begin.'
T h is p ro c e s s show s again th a t the a ffric a te /ts /
participates of the nature of both /t/ and /s/. This should
probably be treated as a sort of disim ilatory process. As in the
cases above, the order [-cont], [+cont] does not seem to be
phonologically relevant.
In the follow ing sections, I examine two processes which
affect stops and where the trigger is to the left of the target;
that is, processes of the type in (23):

(23) [-cont] - > X / Y ____

If the order of articulations were the determ ining factor in

predicting w hether a ffrica te s undergo rules a ffe ctin g stops,
affricates should undergo both processes, since in both cases
the triggering environm ent is adjacent to the [-cont] side of the
affricate. The fact is that affricates fail to be affected by both
processes to be exam ined, showing again that the order of
 230

j a rticu la tio n s of a ffrica te s Is irre le va n t for the operation of

; phonological rules in Basque. A fter concluding th at surface
' linear order is not what determ ines whether affricates are going
, to be affected by rules a ffe cting stops or rules affe cting
fricatives, the question remains of whether there is any way to
, predict the behavior of affricates with respect to a given rule
that targets stops (or fricatives). It is suggested that there is a
p aram e ter w hich m ust be fixe d fo r each rule th a t could
potentially take affricates as inputs. The particular setting of
the param eter will indicate w hether the specification [a cont] in
the rule will be satisfied by segments that are both [a cont] and
[-a cont].

7. Voice Assim ilation

In most Basque dialects there is a process of stop voicing
after nasal or lateral. At some historical point this was a quite
g en era l process p ro b a b ly a ffe c tin g all such w o rd -in te rn a l
sequences. This can be established by examining old borrowings
where the source language had sequences of nasal or lateral and
voiceless stop:^

4 This change does not operate in the easternm ost dialects, Souletin and
Roncalese. In these dialects, Latin and Rom ance borrowings m aintain their
voiceless stops after nasal or lateral. Som e native items that present voiced
stops in this environment also have voiceless stops in Souletin/Roncalese, e.g.,
St. igande. Soul, igante 'Sunday'; St. g a ld e tu . Soul, galthatü 'to ask.'
 231

(24) Latin Basque

altare aldare 'a lta r '
tempera denbora 'tim e '
v o lu n ta te borondate 'w ill'
sanctu saindu 's a in t'

The rule of voicing ceased to affect morpheme internal

sequences at some given point. Nowadays morpheme internal
sequences of lateral or nasal plus voiceless stop are so frequent
that no voicing rule can be said to operate at this level. The rule
in fa ct p ro d u c tiv e ly affects only in fle c tio n a l su ffixes, both
verbal and declensional. ^
In the standard dialect, as in m ost dialects, both the
p e rfe c tiv e /-tu /, the fu tu re /-k o / and the p a rtic ip ia l /-ta /
undergo this rule:

(25) a. perfective
/neka-tu/ [nekatu] 'get tired'
/ar-tu/ [artu] 'ta k e '
/afal-tu/ [afaidu] 'have dinner'

5 Curiously teg! 'place' shows voicing in a few compounds such as arrandegi

'fish market.' In the coinage of recent compounds this idiosyncrasy has not
been respected, e.g. euskaltegi 'Basque language instructional center,' without
voicing of the coronal stop. A second curious case Is the suffix / - l a r / 'native
of w hich con sisten tly shows voicing a fte r/n / or /I/, e.g . baztandar
'B aztanese,' b rasild ar 'Brazilian.' This is most surprising since, as shown in
chapter 2, this suffix does not undergo stratum I rules such as Vowel
Assimilation and, in most dialects, Palatalization.
 232

/lan-tu/ [landu] 'la b o r'

/ken-tu/ [kendu] 'take away'
b. future
/neka-tu-ko/ [nekatuko] 'get tired'
/egin-ko/ [e fig g o ] 'do, make'
/eéan-ko/ [eéaqgo] 'say'
/il-ko/ [ilgo] 'die, kill'
c. p articipial
/n e k a -tu -ta / [nekatuta] 'tir e d '
/ e g in - t a / [eginda] 'done, made'
/eéan-ta/ [eéanda] 's a id '
/ i I - 1a / [ild a ] 'dead'

The im p e rfe c tiv e /-ts e n / w h ic h s ta rts w ith an

affricate, on the other hand, does not show voicing in any
d ia le c t:

(26) im p e rfe c tiv e

/neka-tsen/ [nekatsen] 'get tired'
/ar-tsen/ [artsen] 'take'
/afal-tsen/ [afaltsen] 'have dinner' *[afaldzen]
/lan-tsen/ [lantsen] 'labor' *[landzen]
/ken-tsen/ [kentsen] 'take away' *[kendzen]
 233

In the nominal declension, both suffixes starting with a

vo ice le ss stop, ablative /-tik / and g e n itive -lo ca tive /-k o /
undergo voicing after nasal or lateral, although only with
proper names and temporal and locative expressions (27a).
W ith com m on nouns ending in nasal or lateral there is
epenthesis of [e], as with all other consonant-final stems
(27b). Again, the usage in the standard language reflects the
most extended pattern. Suffixes that start with an affricate
never voice (27c):

(2 7 )a ./b a ^ o -tik / [badotik] 'from the forest'

/le rin -tik/ [lerindik] ' from Lerin'
/non-tik/ [nondik] 'from where'
/u ^ u rb il-tik / [u ^ u rb ild ik ] 'from U surbil'
/ e m e n - t ik / [em endik] 'from here'
b ./b a é o -ko / [ba^oko] 'of the forest'
/lerin-ko/ [leriogo] ' of Lerin'
/non-ko/ [noogo] ' of where'
/orain-ko/ [orayogo] ' of now'
/u ^ u rb il-k o / [u ^ u rh ilg o ] ' of Usurbil'
/bein-ko/ [beyogo] of once'
c ./b a ^ o -ts a t/ [ba^o tsa t] ‘for a forest* *[b a é o d za t]
/bein-tsat/ [beyntsat] 'at least (lit. 'for once')
*[b e y n d z a t]
 234

/ u ^ u r b il- t s a t / [u ^ u rb ilts a t] 'for Usurbil'

* [u ^ u rb ild z a t]
/m a r tin - ts a t/ [m a rtin ts a t] 'for Martin '
* [m a rtin d z a t]

The rule can be given a linear, informal formulation as in

(28):

(28) [-cont] —> [+voice] / [-cont, +son]

This is a rule for which the contour segm ent/ edge

effects hypothesis predicts that a ffrica te s should be lic it
targets. From their left side, which is the relevant side for
this rule, affricates are [-cont] segments.
An possible explanation for the fact that affricates are
not voiced by rule (28) is that a voiced affricate [dz] is not
part of the inventory of m ost Basque dialects. S tructure
preservation would then block the application of this rule to
affricates. There are however (Biscayan) Basque dialects that
have a voiced affricate, in words such as those in (29):

(29) [dzaqga] 'immersion'

[d z ird z a rtak] 'beams'
[dzartako] 'b lo w '
 235

In these dialects a ffricates also fail to voice in the

context of rule (28). That this hypothetical explanation is not
along the right lines is clearly shown by the Palatalization
facts to be discussed in the following subsection.
Another possible explanation is that the morphemes that
undergo the Voicing rule must be expressly marked in the
lexicon, given its restricted application. The fa ct th a t no
a ffric a te -in itia l s u ffix undergoes vo icin g w ould then be
accidental. This explanation is again not available for the
failure of affricates to undergo progressive Palatalization.

8. P alatalization
In chapters 2 and 4, we saw that in a number of Basque
dialects /t/( as well as /I/ and Inf), undergoes progressive
palatalization. As was mentioned when the domain of the rule
was studied, this palatalization process does not a ffect /ts/.
The examples in (30) are from Ondarroa:

(30 ) [itsu] 'blind'

[itsal] 'sad, huge'
[menditsu] 'mountainous'
[a rits u ] 'stony'
[b e ré in ts u ] 'sim ilar'
[ekintsa] 'action'
 236

W hat could prevent a progressive palatalization rule that

I

applies to /it/ from applying to /its /? This blockage clearly

does not have anything to do with the difficulty of producing a
palatal allophone of /ts/, since both [tS] and [S] are common in
Basque. I will suggest that the rule does not apply to affricates
because they bear the feature [+cont] in addition to [-cont] and
the input is required to have exclusively the 'minus' value for
this feature. In fact, I will assume the existence of a parameter
that needs to be fixed for particular rules with two options:
"exclusively [a f]," if segments which are both [af] and [-af] will
not undergo the rule, and "inclusively [af]," if [a f]/[-a f] segm ents
are allowed to undergo the process. If this param eter is set as
; "inclusively," a rule affecting [acont] segm ents w ill also affect
segm ents characterized as both [acont] and [-acont]. This is the
case of the Stop Deletion rule seen above. If the setting is
"exclusively," on the other hand, the specification [acont] of the
input w ill exclude affricates. Both P alatalization and Voicing
would have the "exclusively" setting.
 237

Chapter 6
Suprasegmentals

O. Introduction
Basque dialects differ widely in the prosodic systems that
they possess. W hereas eastern varieties are stress languages,
many central and western dialects are of the pitch-accent type.
W ithin each of these categories there is also quite a bit of
variation. From both a descriptive and a theoretical point of
view, the pitch-accent varieties are of the g re atest interest.
These are also the least well understood system s. In this
chapter I will examine representative pitch-accent varieties in
some detail. First I w ill briefly review the facts of the best
known stress varieties in order to show the range of variation
that obtains among Basque dialects.
There are both stress-accent varieties and pitch-accent
varieties where the distribution of suprasegm ental patterns is
m orphologically determ ined.
I w ill show th at w hereas in some stress-a ccen t types.
Stress A ssignm ent precedes a num ber of phonological rules
w hich reduce vow el sequences (S o u le tin ), in p itc h -a c c e n t
va rie tie s, p itch -p a tte rn s are determ ined w ith respect to the
number of tone-bearing units in the surface form.
 238

1. Stress-accent systems
Three m ajor stress-accent types are generally recognized
in the Basque-speaking area (cf. M ichelena (1972; 1985), for
whom the pitch-accent type co nstitutes the fourth accentual
system. Cf. also Txillardegi (1984)).

1.1. Soule (Zuberoa)

One of the major three stress-accent types is that of Soule
(Zuberoa). The Souletin stress type is relatively w ell-know n,
e spe cially is its north-eastern variety w hich is described in
Larrasquet (1928; 1934; 1939) (cf. also Michelena (1958; 1985)
Txillardegi (1984, 387-405)). In this dialect, penultim ate stress
c o n s titu te s the g e n e ra l p a tte rn . P e n u ltim a te s tre s s is
determined in a domain comprising the stem and the determiner.
O ften the position of the stress in indefinite form s is thus
shifted to the left in singular and plural form s, as in the
absolutive examples in (1):

(1) indef sg pi
[gisun] [gisuna] [gisûnak] 'man'
[m it^il] [mit^ ila] [m it^ilak] 'servant'
[büüsâgi] [büüsagia] [büüsagik] 'owner'

N otice that in a form such as /b ü ü sa g i-a k/ [büüsagik]

penultim ate stress presupposes that stress assignm ent must
 239

take place before the contraction of the two last vowels. The
derivation of this form is shown in (2) :

(2) /b ü ü s a g i- a k /
S u ffix a tio n : büüsagiak
Stress Assign.: büüsagfak
C o n tra ctio n : büüsagik

The reduction of a vowel sequence after penultim ate stress

assignm ent is also observed in the singular form s of stems
ending in a low vowel such as /alhaba-a/ [alhaba] ’the daughter
(abs),' w here Stem -Final Low Vowel Deletion applies ( of.
chapter 2, section 2.1.1, rule (8)), wich contrasts with [alhaba]
’daughter’.
In Souletin, number markers are quite clearly segm entable
from case endings. All plural nonlocative form s, with the
exception of the absolutive, carry an /-e -/ plural marker which
obligatorily carries the accent:

(3) /m ith il- a k / [m ithilak] ’the servants (abs)'

/m ith i l - e - k / [m ithilék] ’the servants (erg)’
/m ith i l - e - z / [mithilés] 'by means of the servants’
/ m it h i l- e - n / [mithilén] 'of the servants’
/m ith il- e - r / [mithilér] 'to the servants (dat)'
 240

M ichelena (1958, 215) argues that in these plural form s

oxytony is also due to a historical contraction. Synchronically,
however, plural /-e-/ must sim ply be listed as accented in the
lexicon.
Stress is assigned at the stratum where num ber markers
(and other determiners, cf. [gisumbat] 'a man') are attached and
before rules reducing certain vowel sequences apply at this
s tra tu m .
On the other hand, stress assig n m e nt m ust precede
com pounding. In com pounds w here the second m em ber is
m onosyllabic, stress is on the last syllable; e.g.: /bürdün-hur/
[bürdünhûr] 'ferric water.' This indicates that stress is assigned
to each of the members of the compound separately and the first
of the two stresses is then deleted or reduced to secondary
stress (a secondary stress is marked on the first m em ber of
some other com pounds such as [elisag isun] 'churchm an').
Consider the examples in (4):

(4) /bürdün-hur/ /e lis a - g is u n/

Stratum I:
Stress Assign.: bûrdün hur elisa g isun
Stratum II
Compounding: bürdünhûr e lis a g is u n
Stress Reduction: bürdünhûr e lis a g is u n
 241

Some derivational suffixes which we saw in chapter 2 are

attached at the same stratum as compounding in many dialects
also attract the stress; e.g.: /e lis a -tia r/ [ e lis a t y a r ] ’c h u rc h ­
g o e r.'T h e s e d e riv a tio n a l s u ffix e s thus are s tra tu m II s u ffix e s
and c o n s titu te a dom ain f o r s tre s s a ss ig n m e n t in s tra tu m I,
se p a ra te fro m the ste m . T h is , h ow e ver, does not apply to a ll
d e riv a tio n a l s u ffix e s , e.g.: /fr ü ta - d ü n / [frü ta d ü n ] "bearing fr u its '.
Suffixes like /dün/ are stratum I suffixes. The different effects
of being attached at each of the two strata are illustrated in (5):

(5) /gisun/ /gisun-a/ /elisa-tiar/ /früta-dün/

Stratum I
Morphology: gisuna frütadün
Stress Assign.: gisun gisuna elfsa tyâr frütadün
Stratum II
Morphology: elisatyar
Stress Reduction: elisatyar
[gisun] [gisuna] [elisatyar] [frütadün]

A subvariety within this stress type is constituted by the

now extinct Roncalese dialect, where the general rule is also
penultim ate stress; but the domain of stress assignm ent is the
bare stem. Thus instead of the Souletin pattern with stress shift
in [gisun], [gisuna], [gisûnak], in Roncalese there is [gison],
 242

[gisona], [gisonak], with stress fixed on the penultim ate syllable

of the stem.

1.2. Baztan
A second stress type is that employed in an area of Navarre
com prising the Baztan V alley, where stress is also m ostly
penultim ate but stress placem ent cannot signal any difference
among words, grammatical or otherwise. My description is based
on Salaburu (1984, 307-27). Penultimate Stress Placement is in
this dialect a postlexical rule applying after all suffixation and
all obligatory and optional rules reducing vowel sequences and
creating glides:

(6) [etSe] house"

[étSya] /etSe-a/ "the house"

[gisona] "the man"

[glsonari] "to t h e man"

[gisonarékin] "with t h e man"

[glsonarendako] f o r t h e man"
[andrya] /andre-a/ t h e woman"
[ a nd r y a y n e ] /andre-a-ren-a/ the one of the

woman"

In bare stems used w ithout inflectional affixes, if the last

syllable is closed it is usually stressed; but this is only a
 243

tendency; e.g. : [gison] (more usual) or [gison] 'man', [lagun] or

[lagun] 'friend'.

1.3. Bortzerrieta (Cinco Villas)

Finally, in the Navarrese region known as Bortzerrieta in
Basque and Cinco Villas in Spanish and surrounding areas in
Navarre and Guipuscoa, there is a different stress type where
the predom inant pattern is stress on the second syllable from
the left (of. Holm er (1964), Txillardegi (1984, 342-349)), as in
[e ^ k u k ] 'the hands', [gisona] 'the man', [b a ^ é ria ] 'the farm ',
[b a é é rita ra ] 'the farmer', [m indfsale] 'm ountaineer', [s a k u re n ts a t]
'for the dog/ for the dogs', [ilis e ta tik ] 'from the churches'**.
Unstressed vowels are often reduced to [I, u, 0] or deleted; e.g.
gizonak eoin du [oisonak mdu] 'the man has done it.'
A system atic exception to this general pattern of stress on
the second syllable is constituted by w ords w hose stem is
m onosyllabic. In these items, the stress falls on the initial
syllable, i.e. on the stem. Thus from /lu r/ 'land' we obtain forms
such as [lu ra n ] 'in the land', [lu ra k u k ] 'the ones of the land',
[lu ra ta n ] 'in the lands', [lû ra tik], [lu rtik] 'from the land'; and from
/ a r / 'worm' [a re n tsa t] 'for the worm / for the worm s' all with
stress on the initial syllable. These facts show that stress is
assigned at the stem level, if the stem has two or more
sylla b le s , the in itia l sy lla b le is e x tra m e tric a l and stress.

lT h e exam ples in this section were all recorded from Lourdes T elletxea,
from the town of Labaien, whom I want to thank.
 244

assigned to the first metrical syllable from the left, will fall on
the second syllable. If the stem domain has only one syllable,
extram etricality is inapplicable, since it would make the whole
domain extram etrical, and the only syllable of the stem will
receive the stress.
Some exceptional bisyllabic stems also have initial stress,
e.g.: [bérde] ( sg [bérdia], pi [bérdik]) 'green' [ba^o] 'glass', [p o ru ]
'leak', [t^ tu] 'w histle'. I w ill assum e that these item s are
marked in the lexicon as constituting exceptions to the rule that
assigns extram etricality to the initial syllable.

2. Pitch-accent systems
In this section, I w ill exam ine four d ia lects representing
d iffere nt degrees of phonological restrictiveness. I w ill start
with the description and analysis of a full-fledged pitch-accent
system , that of G ernika in the Biscayan interior, where pitch
patterns allow for a number of phonological oppositions. Next I
w ill study the system of O ndarroa, on the Biscayan coast,
sim ilar in many important respects to that of Gernika, but where
surface pitch patterns are lim ited to a tw o-w ay opposition.
Then, I will present data from the still more limited variety of
Altzaga in the Guipuscoan Goiherri, which both phonetically and
phonologically represents a rather d ifferent system. I will end
with the dialect of Arbizu in the Sakana Valley of Navarre, akin
to th a t of A ltz a g a , but w h e re p ro s o d ic p a tte rn s are
 245

p ho no lo g ica lly devoid of content and w hich co nstitu te s the

p itch -a cce n t e qu ivale nt of a stress d ia le c t like Baztan. I
believe these four varieties to be representative of the range of
variation that can be found among pitch-accent varieties of
Basque.

2.1. Gernika
My description of the prosodic system of the Gernika dialect
is based fo r the m ost part on my own fie ld w ork with
in f o r m a n ts .2 My point of departure, however, has been the
e xcellent description by B asterrechea (1974; 1975), which I
have followed rather closely in many points in obtaining my own
data (cf. also Jacobsen (1972)).
In the unmarked case, nouns (adjectives, demonstratives and
n o n c o n ju g a te d v e rb a l fo rm s ),^ have a uniform high pitch
th ro u g h o u t, w ith the e xce p tio n of the firs t s y lla b le of
p olysylla b ic item s, which has a lower pitch. These are the
patterns found in the focus position of the sentence, which is
the position of maximal contrast. Elsewhere, neutralizations of

2| wish to thank A ran tza A praiz from A jan g iz, Izaskun Intxausti from
Gernika and Karmele Ikazuriaga from Kortezubi.
^ I will leave inflected verbal forms outside of the present study. The
surface patterns of these items is related to the assignment of focality in the
sentences in ways that are still totally mysterious. M ichelena (1976, 149 ),for
instance, rem arks the two different patterns of the conjugated auxiliary d u te
in gaur izan duté ' (yes) they had it today' and oaur izan dutè 'it is today that
they had it' in his dialect of Errenteria. These facts are also discussed in
M ichelena (1981 ).
 246

several types operate, as we will see below. Examples are given

in (7) of nouns in th eir uninflected and absolutive singular
fo rm s:^

(7) u n in fle c te d abs sg

ar ara ‘worm*

lu r lu re g ro u n d ,la n d '

base basoa fo re s t"

mendi mendiZe "mountain"

gison gisona "man"

andra andrfè "woman"

lagun lagune "friend"

guntsurun guntsùrünè "kidney"

4 In the exam ples, pitch patterns are represented for the most part in a
redundant manner both by means of lines above {high toned syllables) and under
the characters {low toned syllables) and also by acute accents on the vowels of
high toned syllables. The first system of transcription has been used by
B asterrechea (1 9 7 4 ; 1975) for G ernika Basque and is also the system
normally used in other languages of the pitch-accent type such as Japanese (cf.
for instance Haraguchi (1 9 7 7 ), Clark (1987 )). The second system has been
employed by Rotaetxe (1979) for Ondarroa Basque and is the usually utilized
system in tone languages, African and other. In underlying, phonological
; representations, on the other hand, it is low tones that must be marked. This is
' done by means of grave accents (as in Azkue (1931; 1932) M ichelena (1972;
I 1 9 7 6 ;1 9 8 5 )).
 247

I propose to account for the pitch patterns in (7) by means

of the tone assignment rules in (8):

(8) Tone Assignm ent

I. H-Assignment: Assign a high tone to all syllables in the
tonal unit.
II. H-Delinking: Disassociate a branching high tone from the
leftm ost syllable.

The application of these rules is exemplified in (9):

( 9 ) I. l u r lure guntsurune
I \I \_ \ I /
H H H

II. n.a. lure guntsurune

I \ 1/
H H

Extratonal syllables will receive a Low tone by default.

A m odification in this pattern is introduced by form s
bearing plural suffixes, which cause a final drop in pitch. Often,
the only difference between singular and plural form s for a
given case is prosodic. Singular and plural ergative, dative and
genitive forms are given in (10) for some examples:
 248

( 10 ) a . e r g s g e r g / a b s pi

one k o nek 'this/these'

arak arak ■wor m’

lurek lurek land'

andriek andrak "womar

gisonak gisonak ’m a n ’

basoak basoak ■f or e s t

tjgkûrék t Sa k u r e k dog'

quntsûrûnék Gunt s ur unek kidney

dat sg dat pi

arari arari worm'

luréri l ur e r i land"

t Sa k û r é r i tSakureri ■dog'

gisonari gisonari man'

ge n s g ge n pi
— " —

a ran aran "worm"

— ■— ___

l ur e n luren "land"
 249

Üakûrén t S a k ù ren dog'

gisônan g i s o nan 'man'

I w ill assum e th a t plural su ffixes such as a b so lu tive /

e rgative /-àk/, dative /-à ri/, genitive /-a n / /- e n /) ,® have an
underlyingly linked low tone (I mark the underlying low tone
with a grave accent); unlike the corresponding singular markers
w hich are not m arked for tone.® The underlying low tone is
associated with the first syllable of the suffix, if it has more
than one. I modify the first rule of Tone Assignment as in (11):

s In the genitive there is a certain tendency to also distinguish the singular

and plural m arkers segm entally: singular /-a n /, plural /-e n /; but there is
much fluctuation. The distinction is consistently m ade in other areas and
perhaps also by some Gernika speakers , e.g. [gisônân] 'of the man' vs.
[gisonen] 'of the men.' The reason for the fluctuation in the speech of at least
some speakers is clearly the neutralization that necessarily operates with
roots whose final vowel is high as a consequence of Vowel Assimilation, e.g.:
[lagûnén] 'of the man', [lagûnen] 'of the men'. In M arkina, on the other hand,
the tendency is for Vowel Assimilation to be blocked in the genitive (cf. Rollo
(1925)).
® To assume underlying low tones seems preferable to positing extratonality
for the relevant affixes, since their extratonality should disappear when other
affixes follow, appearing with a high tone. This is not what happens. Cf. ,for
in s tan ce, /la g u n -è n -a / [la g û n e n a ] 'the one of the frie n d s '. In the
extram etricality account plural /e n / should lose Its property of failing to
receive a high tone, following common assumptions.
 250

; (11) I. H-Assignment: From the left boundary, assign a high tone

' to all syllables in the tonal unit up to (and not including) the
' first syllable bearing an underlying tone.

Formally, (11) would include two operations: linking of the

high tone to the leftm ost Tone Bearing U nit by U niversal
Association Conventions (of. Pulleyblank (1986)), and rightward
spreading of this tone to all free Tone Bearing Units. Derivations
are given in (12):

(12 ) / a r - a r i / /a r - a r i/ / t ë a k u r - a r i/ / t ë a k u r - à r i/
To the w orm ' to the w orm s' to the dog' to the dogs'

1. a ra ri a ra ri të a k u re ri të a k u re ri
\ 1/ 1 1 \ 1 1/ \ / 1
H H L H H L

II. a ra ri (n. a.) tS aku re ri tS a ku re ri

\/ \ 1/ 11
H H H L

A fte r d e fa u lt low to ne s are a s s ig n e d to to ne -

syllab le s, the co rre ct surface p atterns are obtained. 1
assume that the assignment of default tones obeys the OOP:

(13 ) a ra ri a ra ri t& a kü ré ri t& a ku re ri

1 \/ 1 \ / 1 \ 1/ 1 1\ /
L H H L L H L H L
 251

G enitive m arkers can be follow ed by other Inflectional

suffixes. Combining singular and plural suffixes the patterns in
(14) obtain (absolutive form s):

( 14)
/on-en-a/ onéna the one of t h i s o n e ’

/on-èn-a/ onena the one of t h e s e o n e s ’

/on-en-èk/ on e n a k the ones of t h i s o n e ’

/on-èn-èk/ o ne nak the ones of t h e s e o n e s ’

/tëakur-an-a/ t ë a kùré na t h e one o f t h e d o g ’

/tëakur-an-a/ t ë a kùr ena t h e o n e of t h e dogs '

/ t ë a k u r - a n - à k / t ë a k ü r e nak the o n e s of the do g ’

/ t ë a k u r - à n - à k / t ë a kür enak the o n e s of the d o g s ’

As our rules of Tone Assignm ent predicts, the high toned

unit extends in each case only up to the leftm ost undelying low
tone. The tonal derivation of the first four exam ples In (14) is
given in (15):

(15) /on-en-a/ /o n - è n - a //o n -e n -a k //on-èn-ék/

I. on e n a onena on e n a k on e n a k
\ I / I I Ml I I I
H HL HL H L L
 252

II. onena (n.a.) onenak (n.a.)

\/ I I
H H L

su rfa ce fo rm s a fte r d e fa u lt in s e rtio n :

gnéna o nena o n é nak o nenak

Another illustration of the effect of a low toned suffix on

follow ing suffixes is given by the suffix /-tsa t/. This suffix
has to uses. It can be used directly attached to the noun stem
two produce prolative forms or it can be used after the genitive
to give benefactive forms. In its prolative use, It can be shown
that this suffix does not bear an underlying low tone; it does not
cause a drop in pitch (see 16a). It will be low, however, if it
follows a plural genitive (see 16b):

( 16 ) a . / g i s o n - t s a t / glsontsat f o r / a s a man"?

/a r-ts a t/ a rts a t fo r /a s a w orm '

b. /o n - e n - t s a t / a n e n tsa t fo r (in b eh alf o f) th is one

/o n - e n - t s a t / on e n tsa t fo r these ones'

/tSakur-an-tsat/ tSakürèntsàt f o r t h e dog'

/tSakur-an-tsat/ t S a kur e n t s a t ' f o r t h e dogs'

7 As in "I took him for a man (but he was an extraterrestrial creature).”

 253
I

I Besides plural suffixes, other suffixes also bear a low

I tone. One of them is the com itative marker. C om itative forms
' m ight seem to bear a suffix /-ag a s/ in the singular and a
' different suffix /-akas/ in the plural. A more abstract analysis,
accounting for the apparent allom orphy, w ould postulate the
existence of a single com itative suffix /-gas/ which is attached
to form s bearing the singular dete rm in er /-a / or the plural
/àk/. To the plural sequence /-àkgàs/ the rules of Devoicing and
Stop D eletion w ould o b lig a to rily apply, giving [-akas] (cf.
Chapter 2). The tonal facts do indeed support this view. In the
comitative case, both singular and plural form s have a final drop
in pitch, but in the plural the drop to low pitch is retracted one
syllable with respect to the singular:

( 17 )
V gison-a-gas/ g iso n a gas ' w i t h t h e man'

/ g i s o n - a k - g a s / g i s o nakas ' w i t h t h e men'

/ar-a-gas/ aragas "with t h e w o r m '

/ar-àk-gàs/ arakas 'with the worm s'

Among the locative suffixes, the inesive /-an/ and adlative

/-ra/ are toneless. The ablative /-tik/, on the other hand bears a
low tone; that is, its underlying form is /-tik/: ®

® As B asterrechea points out, the behavior of the other basic locative

suffix, the genitive /-k o /, is harder to establish. W hen word final it is low,
but there is a tendency to distinguish betw een singular and plural when a
 254

(18) basoan ‘in t h e f o r e s t *

basera ‘to t h e f o r e s t *

basotik from the f o r e s t ’

me n df Ze n ‘in t h e mountai n*

m e n d Iré to the mountain'

m e n d f t ’ik ■from t h e mountain"

g e r ni ken in Gernika*

g e r nf ker a t o Gernika*

g e r nfkètik ' f rom Gernika*

The l o c a t i v e p l u r a l i z e r t s l o w - t o n e d : / - è t a /

( 19 )
/lur-èta-n/ luretan in t h e l and s '

/lur-èta-ra/ luretara t o t h e l a nd s '

/lu r-èta-tik / lùretatik f r o m t h e l a nd s '

/m endi-èta-n/ m e n d i ^etan in t h e m o u n t a i n s '

/m endi-èta-ra/ rnend i Ze t a r a to the m o u n t a in s

/m endi-èta-tlk/ mendiZetatik from the m ou n ta in s'

determ iner follows the locative genitive, e.g. [basôkôâ J'the one of the forest
vs. [basokoak] 'the ones of the forest'.
1 255
f

j The a dje ctival su ffixes /-e n / 'm ost', /-è g l/ 'too ', and
■ /-a g o / 'm ore' also bear a low tone, as indicated. A few
i
derivational affixes, such as the numeral ordinal /-g à re n / (also
an exception to Vowel A ssim ilation in m ost dialects) and
/-(t)à r/ are also marked lexically with a low tone (for more
examples see Basterrechea (1974:1975):

(2 0 )
ana T h e good o n e ’

o nena Th e b e s t one'

baltsa The b l a c k one*

baltsagoa t h e b l a c k e r one'

baltsena ' the b l a c k e s t one'

b i garena 'the s e c o n d one'

s a s D igarena t h e s e v e n t h one'

S a m p l e t o n a l d e r i v a t i o n s are g i v e n in (21):

(21) /on-a/ /on-èn

1. H - a s s i g n m e n t ona onena
\/ 1 1
H H L

II. H - D e l i n k i n g ona (n.a.)

11
H
 256

D e fa u lt L ona onena
II I \/
LH H L

The assignm ent of tones to nonconjugated verbal form s

follows the same rules as nouns, adjectives and dem onstratives.
M ost p e rfe c tiv e ve rb a l form s fo llo w the p a tte rn in (7).
Im perfective form s are created either by adding the suffix /-
tsèn/ to the verbal root or the suffix /-te n / to the perfective
form. Both suffixes have a lexical low tone. The related suffixes
/-tsè/, /tè / are also marked with a low tone. The future marker
/-ko/ is also low toned. Examples are given in (22) and a couple
of sample derivations in (23):

(22 ) / g a l - t u / g a ldù to lose (p e rf.)’

/a r-tu / a rtu to take (p e rf.)

/g a l- ts e n / g a ltsen , / g a l- t u - t e n / g a lduten lo lose (im p .)’

/a r - t s e n / a ris e n to take (im p.)'

/ g a l- t s e - a / g a lts ie . / g a l- t u - t e - a / g a ld u tie 'the lo s in g '

/ g a l- t u - k o / g a lduko to lose (fu t.)'

(23) /g a l- ts e n / / g a l- t u - t e n /

I. g altse n galduten
I I \ / I
H L H L
 257

II. (n.a.) galduten

I I
H L

D e fa u lt L (n.a.) galduten
I I I
L H L

A small number ofnouns and adjectives presenta different

tonal pattern from that seen in (1). Theses are mostly bisyllabic
stem s with a high-low pattern; although, at least for some
speakers, there are also a few trisyllabic stems with a final low
syllable in this group. Both uninflected and absolutive singular
form s of some of these words are given in (24a). A great
percentage of the items with this exceptional tonal pattern are
borrow ings; but not all of them are. Also some, very few,
p e rfe c tiv e v e rb a l fo rm s p re s e n t th is irre g u la r p a tte rn .
Basterrechea (1975) gives only six examples, a couple of them
also borrowings, obtained after revising the entire verb-list of a
dictionary. I found another example, nevertheless, which I offer
at the end of the list. These are given in (24b)

® Basterrechea (1974 ) gives a long list of exceptional nouns. There is only

one item listed there which is not exceptional for my informants. This is the
word /d e i-a / 'the call', which my informants pronounce with a regular pattern
: [dey^éj. On the other hand, Basterrechea does not list any trisyllabic items;
but, as mentioned, at least for some speakers of the Gernika area there is a
few trisyllabic stems that fall in this group, such as [basé ri] [b aséri^ e]
'farm', [bask ari], [baskâri^e] 'dinner' and [ond âru] 'Ondarroa'.
 258

( 2 4 ) 8 . un i nf l abs s g

bànku b a n kue ( S p . b a n c o ) bank"

barku b a r kue ( S p . b a r c o ) boal'

bwélia b wé l t j j e (Sp. b u é l l a ) Turn'

d r ô ga drogie (Sp. dr ôga) drug'

xénie xénlle (Sp. g é n i e ) ' peopl e'

kafe kafie ( bu l Sp. c a f é ) 'coffee'

may/Cu m a u /Cue ' hamme r'

léku l é kue place'

laide laldie ■group'

gam 9arm Top'

éwfê éw ile fronl'

aise alsie 'back'

biTbû ( b u l Sp. B i l b a o ) Bi l bao'

b. v e r b s

laga lo leave'

bôla lo Ihrow'

xawM lo fall'

xauo l o be born'

éroan l o carry'
 259

è r ai n "to c a u s e ’

bè yt ' u "to l o o k at'

There is even a couple of minimal pairs where one item has

a regular pitch pattern and the other one is exceptional:

(25) basoa the forest" / basoa t h e dr i nk i n g g l a s s ' (Sp.

vaso)

t S i s tüé ' the s p i t ' / t S i s tue ' the f l u t e '

W hen a tta ch e d to e xce p tio n a l item s, th e d iffe re n c e

between toneless and low-tone affixes is lost. In (26) I give
exam ples to show that singular and plural form s of marked
items have an identical pitch pattern (a) and that this also
applies to all locative forms (b):

( 2 6 ) a . marked: t S i s tuek f l u t e , erg s g & a b s / e r g pi'; c f .

unmarked: t S i s t ü é k s p i t , e r g sg' vs. t S i s tuek 'spit

a b s / e r g pT

b. marked: biTbon b i l bora b i l b o t i k 'in, t o , f r o m Bi l bao' ; cf.

unmarked: g e r ni ken g e r nikèra g e r n i k è t i k 'in. t o . f r o m

Gernika'

I will propose to treat the exceptional items with a marked

tonal pattern as having an underlyingly linked low tone on their
 260

second and final syllable (e.i /tS is tù / 'flu te ', /b ilb o / 'B ilbao').
The presence of this low tone w ill prevent the assignm ent of
high tone by the general rules of Tone Assignm ent beyond the
firs t syllab le . These form s w ill th e re fo re never p re sen t a
context for the application of the rule of H-Delinking. Sample
derivations are given in (27):

(27) /tëistù-a/ /tëistù-àk/ /bilbô-ra/ /bilbô-tik/

the f l u t e ’ the f l u t e s ' t o Bi l bao' f r o m Bi l bao'

I. H - A s s i g n m e n t

tSistue tëistuek bilbora bilbotik

I I I I I I I I I
H L H L H L H L L

II. H - D e l i n k i n g

(n.a.) (n.a.) (n.a.) (n.a)

D e f a u l t L (& ÜCP)

tSistue tSistuek bilbora bilbotik

I \/ I \/ I \ / I \ /
H L H L H L H L

2.1. 1. Raising and lowering

We must now determ ine if the rule of H -D elinking that
produces an initial low pitch in the unmarked cases is a word-
level rule, as we have been assuming so far or, instead, is a
phrase-level rule. I w ill argue that H-Delinking must apply at
 261

the word level in spite of the existence of evidence that seems

to indicate that the domain of the rule is the phrase. I will start
by presenting this evidence.
Basterrechea points out that the initial low of words such
as the examples in (28a) is lost when the word is not phrase
in itia l, in norm al rapid speech, as the exam ples in (28b)
illu s tr a te :

(28)a. e t§ ié the ho u se , a b s ’

giSona t h e m a n , abs'

b. pure é t ë i é 'our h o u s e , abs'

m a rkihèko giSona 'the man fro m M arkina, abs'

Facts like these would seem to indicate that the rule of H-

Delinking must apply at the phrase level. Consider however the
effects of another rule of lowering before pause.
C haracteristically, if a word has more than one high
syllable, the final high syllable is pronounced with a falling
contour before pause (generally a high-m id contour). This final
fall is quite typical, but not obligatory, and it disappears when
the word precedes a verb (i.e. in focus position). For this reason I
believe that the creation of this final contour should be treated
! as a low level phenomenon:
 262

(2 9 ) tS èkùrê ’the dog'

guntsurune 'the kidney'

The process creating the final falling contours in (29) is a

rule that links a low tone associated with a phrase boundary to
the last syllable linked to a branching high tone, w itho u t
delinking its tone. The restriction to branching high tone input
has the effect of not affecting words with a single high toned
syllable such as [a ra l 'th e w o rm ' o r [të à k ü r] 'dog', w h ic h are
never pronounced w ith a fa llin g tone (i.e. * [è râ l, *[të à k û r]):

(30) Final Falling Contour Formation

V V H
\ /
H L

e.g. tëakürê <5

I
L H L

Now, if the high tone assignm ent and initial high-tone

delinking rules had the phrase as its dom ain, Final Falling
Contour Formation could not distinguish between the phrases in
(31a) and those in (31b):
 263

(31)a. W i t h o u t Final F a l l i n g C o n t o u r / Wi th Final F a l l i n g Co n t o u r

guntsürünè guntsùrùnê \ ' the ki dney'

guntsùrùn guntsürûn \ 'kidney*

tSaküré t ë a kùrêX ' the dog'

b. irù t S a k ù r * irù tSakûr \ t h r e e dogs'

g u re ara * g u rè a r â \ 'our w o r m '

How the incorrect result in the right column of (31b) are

produced by applying all tone rules at the phrase level is shown
in (32):

(32)
I. H - A s s i g n m e n t

guntsurune iru t S a k u r
\ 1 1 / \ \ I /
H H

II. I n i t i a l H - D e l i n k i n g

guntsurune iru t ë a k u r
\ 1/ \ I /
H H

III. Fi nal F a l l i n g C o n t o u r F o r m a t i o n

guntsurune i ru t S a k u r
\ I /\ \ I /\
H L H L
 264

Final F o r m s ( a f t e r D e f a u l t L):

guntsurune *iru tëakûr

The application of the rule creating a fa llin g contour

before pause is restricted to phrases ending in a word of at
least three syllab le s, like in the exam ples in (31 a). This
restriction is captured by stating the output of Final Falling
Contour Formation as a branching high only if H-Assignment and
Initial H -D elinking apply at the w ord level, p hrase -in te rn a l
w ord-initial lows being then leveled to high by a later rule.
Applying the rules in this fashion we will be able to capture the
restrictions in the form ation of a prepausal contour:

(33)
I. H - A s s i g n m e n t

guntsurune iru t S a k u r
\ I I / \/ \ /
H H H

II. I n i t i a l H - D e l i n k i n g

guntsurune iru t S a k u r
\ I / I I
H H H

III. Final Fall C o n t o u r F o r m a t i o n

guntsurune (n.a.)
\ I /\
H L
 265

The rule of Final Falling Contour Formation cannot apply to

the second example in (33) because it does not bear a branching
high tone. An analysis in which both rules of H-Assignment and
Initial H-delinking apply at the word level is thus supported by
the prepausal falling contour data. There is more evidence to
s u b s ta n tia te th is c o n c lu s io n , w h ich w ill be im m e d ia te ly
considered.
Especially in longer phrases, the phenomenon of prepausal
lowering may have the effect of lowering the pitch of the last
syllable or last two syllables. W hat this means, in term s of
rules, is that there can be an optional leftward spreading of the
phrase-final boundary low tone. Basterrechea is quite explicit in
d istin g u is h in g this o ptio na l p h ra se -fin a l low ering from the
obligatory lowering induced by certain suffixes. He states that
the optional phrase-final lowering of high tones produces a less
marked drop in pitch than the obligatory lowering that plural
suffixes cause and claim s that there is a contrast between
segm entally identical singular and plural form s even if optional
final lowering applies to the singular:

(34) t ^ i s t u l a u^ek the f l a u t i s t (erg) '

t S i s tulayZek t he f l a u t i s t s ( a b s / e r g ) '

From my own observations it is not clear at all that the

distinction between singular and plural is alw ays m aintained
 266

before pause, especially at the end of a phrase of five or more

syllables; although it is indeed often the case that singular
form s present a final prepausal mid tone which is perceptually
d istin g uish ab le from the obligatory final low tone of plural
forms. What seems apparent, on the other hand, is that this is an
optional rule of gradient application.
Since in rapid speech there is only one initial low tone per
phrase, the pitch pattern of a long noun phrase can be quite
altered with respect to the patterns that each word would show
in isolation. I have recorded the follow ing exam ples (sim ilar
examples are given by Basterrechea)i"*®

(35) amak arosi aban tSapela

mother-erg buy-perf aux-rel cap-abs

a ma k e r o s i e ban t S a p e l a t h e cap t h a t m o t h e r bought'

gur e etsle
our house-abs

g u re e t s i ê gur e e t s i e 'our h o u s e '

“•o In fact, I used the Spanish translation of many of the phrasal examples that
B asterrechea provides in my elicitation. A difference with Basterrechea's
description is that he claims that the lowering of the final syllable in the
singular is not possible in trisyllabic words , although it is in longer words.
That is, there can be no lowering in words such as [gisônâ] or [tSapélâ]; but
lowering can apply to [bentânié] or [gastélùé] to give [bentânie], [gastélue]. I
have recorded cases of lowering with trisyllabic words at the end of a longer
noun phrase, however.
 267

W hat is interesting about the rule of phrase-final lowering

is that it never has the effect of retracting the first high tone
that the word would have in isolation. That is, unmarked words,
w hose first syllable is low phrase-initially, can never surface
with a high tone on the first syllable immediately followed by a
drop in pitch. A word like [bèntânfé] 'window' can be pronounced
[bèntânîè] (or [béntânîè] if not phrase-initial) before pause, as in
the exam ple in (36), from Basterrechea, with lowering of its
last two syllables; but never *[béntànîè]. A word like [gisona], on
the other hand could appear as [gisona], [gisônà] (or [gisônâ],
[gisônà]); but never as *[gisônà]. In an intuitive sense it is as if
these two words (and all words except for exceptional words
and bisyllabic plurals and others with an underlying low on the
second syllable) were prim arily characterized by a high pitch on
the second syllable from the left: [bentânie], [g is o n a ], any
surface pitch pattern must respect this high tone on the second
syllable:

(36) bigèren Pisuko bentanie

second floor-of window-abs

b i g a r e n m'suko b e n t a n i e the w i n d o w of the s e c o n d floor'

b i g a r e n pI suko g i Sona ' t he man o f t h e s e c o n d f l o o r '

 268

Sim ilarly lowering of the final syllable is possible in the

example in (36a) , but in (36b) it is not:

( 3 6 ) 8 . i r u l l b u r u , i r u l i buru T h r e e b o o k s ( l i t . t h r e e book, Indef)'

b. i r u t S a kur. * i r u t ë a kur t h r e e d o g s ( l i t . t h r e e dog, I n d e f )

If the rule of H-Delinking were to apply at the phrase level,

th e re is no w ay in w hich P h ra se -F in a l Low ering could
distinguish between these examples:

( 3 7 ) I. H - A s s l g n m e n t : Iru l i buru iru t Sa k u r

\ \ I / / W I /
H H

II. H - De l l n k i n g : iru l i buru Iru t S a k u r

\ I // \ I
H H

III. Final Lowe ri ng: iru l i buru *iru tSakur

(& d e f a u l t L) I \ I / I I \ / I
L H L L H L

If we treat the phrase as a unit for tone assignm ent,

Phrase-Final Lowering will be able to apply regardless of the
number of syllables of the last word in the phrase, producing
incorrect results as in (36b). Only by applying H-Delinking at the
 269

; w o rd -le v e l w ill we be a ble to in tro d u c e the re le v a n t

distinctions between different exam ples:

(38) W o rd -le ve l tone

I. H -A sslgnm ent

iru tS akur iru lib u ru gure bentanie gure tSapela

\/ \ / \ / \ 1/ \/ \ I // \/ \ I/
H H H H H H H H

I. H -D elinking

iru të a ku r iru lib u ru gure bentanie gure tëapela

I I I \ / I \ 1/ I \/
H H H H H H H H

Phrase Level tone-rules

Phrase-Final Lowering (maximal spreading)

(n.a.) iru lib u ru gure bentanie gure tSapela

I I I I I \/ I II
H H L H H L H H L

From the final representations in (38) it is clear why final

lowering can never produce phrases such as *[iru tS a k u r] , *[iru
libu ru ], *[guré béntanie], *[guré tSàpela]; a high tone is not
linked to the first syllable of the last word of these examples at
the stage where final lowering applies. All high tones in the
final word could be neutralized as a consequence of phrase-final
lowering in a long phrase. But what final lowering cannot do is
 270

to place a high tone on a syllable that would not bear this tone
were the word in isolation.
If Initial H -D elinking m ust apply at the w ord-level, how
are the patterns in examples such as those in (36) produced,
w h e re n o n -p h ra s e -in itia l w o rd -in itia l low s have been
suppressed? It is evident that this leveling is caused by a
phrase-level process of neutralization which applies even later
than phrase-final lowering.
As Basterrechea points out, short up and downs in pitch are
n eu tralized p h ra se -in te rn a lly. This leve llin g process affects
both high and low syllables. This postlexical leveling of tones is,
I believe, what has been responsible for the once w idely held
opinion that prosodic facts at the lexical level are very much
irrelevant in Basque.^ ^ Postlexical tone leveling has indeed the
effect of obscuring the lexical pitch contours of words to a
great degree.
Consider the example in (39). In (a) the underlying tones
are given; the lexical pitch contour of each word in the phrase
are shown in (b). The phrasal pitch patterns in (c), (d) and (e) are
from Basterrechea (1974: 393). The pattern in (f) was recorded
by myself from a native informant:

11 Of. Rotaetxe (1978, 1979; 142-47) for discussion of this view , which is
associated with Altube (1923).
 271

(39)a. / b i - g è r e n / / p i s u - k ô / / b e n t a n a - a /
second floor-of window-sg

The w i n d o w of the s e c o n d floor*

b. b i g a r e n pisuko bentanie

c. b l c a r e n p i s u k o b e n t a n i e

d. b i g a r e n p i s u k o b e n t a n i e X

e. b i g a r e n p i s u ko b e n t a n i e

f. b i g a r e n p i s u k o b e n t a n i e

The lexical assignment of tones to the example in (39) is

shown in (40);

(40) /bi-garen/ /pisu-ko/ /bentana-a/

I. H - A s s i g n m e n t : bigaren pisuko bentanie

II \/ I \ I//
H L H L H

II. I n i t i a l H- De l i n k i n g : (n.a.) pisuko bentanie

I I \ 1/
HL H

If no o the r rule applies p ostle xica lly, o the r than the

assignm ent of d efa u lt low tones to toneless sylla lb le s the
result w ill be:

(41) b i g a r e n p i s u ko b e n t a n i e
 272

If P hrase-F inal Lowering applies then, we w ill obtain

: (39e). All other possible surface pitch patterns are obtained by
le ve lin g ups and dow ns p h ra s e -in te rn a lly . Low e rin g the
sandwiched high tone in [pisuko], produces (39f) from (39e). In
(39c) and (39d) it is instead the low toned syllables between the
highs in the last two words that are raised to reduce phrase-
internal ups and downs.

2.1.2. A note on historical change

There is one main difference between the data presented

here and Basterrechea's description which is clearly factual and
not purely of interpretation. In the variety that Basterrechea
uses as basic for his description, trisyllabic plural forms have a
drop in pitch that affects the last two syllables, instead of only
the last one, as in our examples. Basterrechea is only explicit
about plural absolutive forms; but from his examples elsewhere
it is clear that this difference also extends to the genitive and
dative and even to plural locatives.
Basterrechea in fact points out the existence of competing
form s and indicates that there is a correlation between the
usage of a set of forms and the age of the speaker.
The forms preferred by Basterrechea correspond to an older
generation. Examples of these forms are given in the left column
in (42); the corresponding form s of the younger generation.
 273

which I have used In my description are given in the right

colum ns:

(42) Ol de r g e n e r a t i o n Younger g e n e r a t i o n

g i sonak g i s o nak the men ( a b s / e r g ) '

g i sonarl g i s o narl 'to t h e m e n (dat)'

g isonen g i s o nen of t h e m e n (gen)'

m endl^etara rnend i Z e t a r a to the mountain s

(adiy

B asterrechea reports th at form s like those in the left

column were more frequent in the 1930's and before; but admits
that the form s in the right colum n were clearly the dom inant
form s in G ernika at the tim e he was w ritin g his tw o-part
article. My inform ants (born in the 1950's and early 60's) use
exclusively the form s in the right colum n; although they are
aware of form s such as those in the left colum n, which they
have heard (from their grandparents).
I propose to account for the old generation forms by means
of a rule that spreads an underlyingly linked low tone leftwards,
delinking a branching high tone. The rule m ust apply after H-
insertion but before Initial H -delinking, since a/it does not
apply to bisyllabic words, which do not bear a branching high and
b/it has the effect of retracting a high tone one syllable with
respect to the singular. The rule is given in (43) and sample
derivations where this rule applies are shown in (44):
 274

(43 ) L-Spread

V V V
\ I
H 1
(44) /lur-ak/ /gison-èk/ /m endi-èta-ra/

H-lnsertion l ur ek gisonak mendl2elara

I I \ / I \ I I
H L H L H L

L-Spreading (n.a.) gisonak r n e ndl Ze t a r a

I \ / I \ /
H L H L

Ini t i al H - D e l i n k i n g (n.a.) (n.a.) (n.a.)

l ur ek g i sonak mendlZetara

The rule of L-Spreading is m orphologically restricted in its

application. It is triggered by plural suffixes but not by any
: other suffixes with an underling low tone. It can not apply for
' instance to /m e n d i-tik/ to give *[m éndîtîk]. Both generations
present [m ènditîk], where the underlying low tone of /-tik / does
not spread to the left. Basterrechea mentions the fact that the
distinction between [ônénàk] 'the ones of this one (abs/erg)' (e.i.
. /on-en-àk/ and [onènàk] 'the ones of these ones (abs/erg)' (i.e.
/on-èn-àk/) is not obliterated in favor of the second form for
I both meanings, contrary to what one would expect. He concludes
 275

that the lowering of the two last syllables in trisylla bic (and
lo n g e r) p lu ra ls (i.e . L -S p re a d in g ) does not a p p ly to
pronom inal/dem onstrative form s. The younger generation has
sim ply lost this rule altogether.

2. 2. Ondarroa
In the dialect of Ondarroa, as in G ernika, the prosodic
pattern of plural forms differs from the unmarked singular and
indefinite pattern in that there is a final drop in pitch. Examples
of singular/plural contrasts are given in (45):^ ^

(45) a. a b s o l u t i v e

I gi Sona ra] "It's t h e m a n ’

[ gi Sona t 1 s ] / g l S o n - a k d i s / ' l t ' s t h e me n ( t h e m e n are)'

[ kat u re] 'It's t h e c a t '

[katu t i s ] it's the c a t s ’

[ etSi re] it's t h e ho u s e '

[ét^l t i s ] / e t S e - a k d i s / 'It's t h e h o u s e s '

[m e n d ié e ra] /m e n d 1 -a d a/'1t's th e m o u n ta in '

[ m e n d i ^a t i s ] / m e n d l - a k d i s / ' l t ' s t h e m o u n t a i n s '

b. e r g a t i v e

[ gi Sonak e yh j a w ] the man h a s done (It)'

[giSonak eyh Jabe] the m en have done (It)

[andri k e yh j a w ] t h e w o m a n h a s done (It)'

12 All exam ples are given in preverbal position to avoid the effects of
prepausal lowering.
 276

[andrak eyh Jabe] t h e w o m e n h a v e done (i t)'

[katuk xan da w] the c a t has e a t e n (it)'

[ katuk xan dabe] t h e c a t s have e a t e n ( i t )'

c. d a t i v e

[ g i So n a y e m o t s a t ] I h a v e g i v e n ( i t ) t o t h e man'

[giëonay e m o t s e t ] I h a v e g i v e n ( i t ) t o t h e men'

[aray e m o t s a t ] I have gi ven (it) to the worm'

[aray e m o t s e t ] I have given (it) to th e w o r m s'

d. g e n i t i v e

[ l agunan kat u re] It's t h e c a t of t h e f r i e n d '

[ l agunan kat u re] It's t h e c a t o f t h e f r i e n d s '

[ katun e t ë i re] It's t h e h o u s e of t h e c a t '

[ katun e t ë i re] It's t h e h o u s e o f t h e c a t s '

Also as in Gernika, there is no singular/plural prosodic

contrast in the case of some exceptional items:

(46) [kafi re] / k a f e - a d a / it's the coffee'

[ kafi t i s ] / k a f è - è k d i s / it's the c o ffe es '

The facts are also as in Gernika as per the distinction

between low-toned and toneless locative markers: the ablative
/-tik / has a low tone, causing a final drop of pitch, and the
adlative /-ra / and inesive /-an/ are toneless:
 277

(47) a./-tik/

[ m e n d i t S i nat or ] I am c o m i n g f r o m t h e m o u n t a i n '

[luréti tator] he i s c o m i n g f r o m t h e land*

[eliSati tator] "he i s c o m i n g f r o m t h e c hur c h

b. / - r a /

[ m e n d i r é nu] ■| am g o i n g t o t h e montain"

[ ba s û r é ru] ‘he i s g o i n g t o t h e f o r e s t *

[ e l f S a ru] "he i s g o i n g t o t h e church"

c. / - a n /

[ me n d i Sa n da] he i s in t h e mountain"

[Turin da] ■(he) i s in t h e land"

[ el f San da] ■(he) i s in t h e church"

G e n itiv e -lo c a tiv e form s w ith /-k o / show the s in g u la r/

plural d istin ctio n co nsistently:

(47) [ me n d i k u re] / m e n d i - k o - a d a / "it's t h e o n e o f t h e mountain"

( men df k u t i s ] / m e n d i - k o - a k d i s / t h e y a r e t h e o n e s o f t h e

mountain"

So far, the facts exam ined are identical in all relevant

respects to those of Gernika. There Is however an im portant
difference between the two dialects: in every case where in the
Gernika dialect the presence of an underlyingly low-toned affix
causes a drop in pitch in a syllable other than the last, in
 278

O ndarroa the drop in pitch affects only the last syllable. An

exam ple is offered by form s containing the plural low-toned
locative /-èta-/. As the examples in (48) show, in Gernika there
is a drop in pitch from the first syllable containing this suffix.
In Ondarroa, on the other hand, only the last syllable of the word
has a low tone:

(48) Gerni ka Ondarroa

me n d i l e t a n m e n d i S è t a n ' in t h e m o u n t a i n s ’

m e n d ile ta iik m endi^etatik from the mountains

me n d i Ze t ar a me n d IS e t a to the mountains'

The underlying low tone of the genitive plural suffix is

realized not on the syllable containing this affix but on the last
syllable of the word if another suffix follow s. The effects of
this displacement of the low tone can be observed by comparing
form s ending in the genitive and form s where another suffix
fo llo w s :

(49) [aran e t S i re] it is the hou se of the worm'

[aran e t S i re] it is the house of the w o r m s '

[àrana ra] ‘i t i s t h e one of t h e w o r m '

[arana ra] i t i s t h e one o f t h e w o r m s ' ; cf.

Gernika: [arana ba]

 279

[ t S a k û r è n e t S i re] i t i s t h e h o u s e of t h e dog'

[ t Sakûr an e t S i re] i t i s t h e h o u s e of t h e dogs '

[ t Sak ûr ana ra] i t i s t h e one o f t h e dog'

[ t Sak ûr ana ra] 'it i s t h e one of t h e dogs' ; c f

Gernika: [ t S a k u r e n a da]

As the examples show, the low tone of the genitive plural

will be moved to the right to be realized on the last syllable of
the word. This rightward m ovement of the low tone produces
different patterns form those in Gernika where the drop in pitch
coincides with the first syllable containing an underlying low
tone.
The three-way tonal contrast that obtains in Gernika from
the com bination of singular and plural genitive and absolutive
suffixes is thus reduced in O ndarroa to a tw o-w ay contrast:
either there is no final dro p,if both genitive and absolutive
affixes are singular, or there is a drop on the last syllable if
either affix is plural:

Underlying Gernika Ondarroa

/on-en-a/ oné na onéna t h e one of t h i s one'

/on-èn-a/ one na onéna t h e one o f t h e s e o n e s '

/on-en-àk/ onénak on é n a k t h e o n e s of t h i s one'

/on-èn-àk/ ô ne nak onénak t he o n e s o f t h e s e o n e s '

 280

Even if the underlying low tone belongs to the stem itself,

it is nevertheless realized on the last syllable of the word
regardless of its m orphological structure. Thus the ablative of
the lexically marked word [bilbo] 'Bilbao' is [bilbotik] and not
[bilbotik] as in Gernika. Also compare the contrast between the
benefactive form s of unmarked /jo n/ 'Jon' and marked /jo ne/
'Jone' in Ondarroa and in Gernika. In Ondarroa the underlying low
tone of /jonè/ is transposed to the last syllable of the affix:

( 5 1 ) Gerni ka Ondarroa

/jon/ J onè n t s a t da J onè n t s a k ô ra "it i s f o r Jon'

/ j o n è / j o n e n t s a t da J o nè n t s è k o ra "it i s f o r J o n e '

All this evidence indicates that Ondarroa Basque has a rule

that links underlying low tones to the last syllable of the word:

(52) Underlying Low Tone Association:

Associate an underlying (low) tone with the last syllable
of the word.

We may assume that the difference between Gernika and

Ondarroa is that In Gernika underlying low tones are linked In
the lexicon, whereas in Ondarroa low tones in affixes and other
lexical entries are part of the morpheme but are not associated
I 281

I w ith p articu la r syllables. The association of underlying low

I tones takes place after suffixation by the Underlying Low Tone
I Association rule. Once this rule has applied, H-Assignment and
■ H-Delinking take place exactly as in Gernika. Sample derivations
are given in (53):

( 5 3 ) Unde rl yi ng: /tSakur-an-a/ /jon-ntsako/ /jonè-ntsako/

L L

Low T o n e A s s o c : t Sa k u r a n a (n.a.) jonentsako

H-Assignment: tëakurana jonentsako jonentsako

\ 1/ I \ \ / / \ I / I
H L H H L

H - De l i n k i n g : tSakurana jonentsako jonentsako

\/ I \ I / \ / I
H L H H L

[ t Sakurana] [jonentsako] [jonentsako]

t h e one o f t h e dogs ' ' for Jon' fo r Jone'

To determ ine if a form ulation like (52) is enough to

account for all surface patterns, we must consider now cases of
multiple underlying low tones, like the example in (54):

(54) /bilbo-tik/ [ b i l b o t i k] f r o m Bilbao"

L L
 282

If (52) is interpreted as requiring one-to-one association

of tones with Tone Bearing Units from the right edge, we will
obtain the incorrect form in (55):

(55) bilbotik
I I
L L *[bilbotik] ( a f t e r H - A s s i g n m e n t )

To obtain the correct surface form [b ilb o tik] we must

ensure that only one underlying low tone is linked. We will
obtain this result if we associate tones left-to-right with Tone
Bearing Units starting from the right, as in (56):

(56) bilbotik
/
L L [ b i l b o t i k]

We may then restate (52) as (57):

(57) Underlying Low Tone Association

A ssociate underlying (low) tones le ft-to -rig h t with Tone
Bearing Units starting from the right edge of the dom ain.
Unlinked low tones would be deleted at this point.

A prosodic system identical in all im portant respects to

this which we have described for Ondarroa is also the basis of
Azkue's (1930, 1931) description. The prosodic facts that Azkue
 283

describes are with all probability those of his native dialect of

Lekeitio; although, follow ing a pattern that has created much
confusion in the understanding of Basque prosody, he suggests
that the facts of his own dialect correspond to the most general
situation in Basque, going as far as to give the tonality of forms
and affixes only used in eastern dialects w here the prosodic
facts are certainly very different. Perhaps for this reason the
w ork of Azkue has been largely ignored in later studies of
Basque prosody. When Azkue's study is read taking the necessary
p recautions, however, it is evident th a t the kernel of his
d e s c rip tio n c o rre s p o n d s to a re a l p ro s o d ic s y s te m ,
fundam entally identical to that of Ondarroa. As in Ondarroa, in
the system described by Azkue an underlying low tone carried by
a stem or affix is realized on the last syllable of the word. I
believe it is worthwhile to quote Azkue's own words. In Azkue's
notation a grave accent indicates that the syllable bearing it and
all syllables to its right have a low tone:

[S]ufijos tonicos, como los declinativos en y ra, al unirse

a vocables ditonos como M a r k o s y B ilb o, reciben ellos
mismos la atonia de ko s y bo; pues en vez de Markôsen
e t x e a n y Biibàra e t o r i decim os M a r k o s è n e tx e a n y
Bilborà etori. (1931,12)

[tonie suffixes, such as the declensional en and ra, when

joined to ditonal words such as M a r k o s and Bilbd, receive
the atony of k o s and b o them selves; since instead of
M arkôsen e txe a n and Bilbdra etoriv/e say M arkosè n
etxean and Bilborà etori.] (my translation, J H)
 284

In Azkue's term inology a ditonal word is a stem which

' carries an underlying (low) tone (such as /m arkos/ and /bilbo/).
Tonal suffixes are those which do not cause a drop in pitch (such
as the indefinite genitive /-e n / and the adlative /-ra /). The
prosodic facts that Azkue describes in the quoted paragraph only
differ from those of Ondarroa in the minor detail that because of
the obligatory deletion of the /r/ of /-ra/ after nonhigh vowels,
in Ondarroa /bilbo-ra/ 'to Bilbao' is bisyllabic and thus does not
show displacem ent of the low tone: [bil.boa] (but cf. [bilbotik],
where the high syllable is the second and not the first as in the
base form).
Azkue (1930, 295) indicates that the benefactive form s
corresponding to the names of the m em bers of the Basque
Academy Campion and Altube (accent marks provided by Azkue)
are C a m p io n e n tz a t , without final drop, and A ltu b e re n tz à t , with
a drop on the final syllable of the word. These examples are
comparable to the Ondarroa examples [jonèntsakô] 'for Jon' and
[jonentsako] 'for Jone.'
Azkue is aware of the existence of other prosodic systems.
In p articula r he notices that in other Biscayan dialects an
underlying low tone is realized on the morpheme that carries it,
as in our description of the G ernika prosodic facts; but he
believes that the most general system is the one where only the
last syllable of the word may receive a phonological low tone.
! Directly after the sentences quoted above he adds:
 285

Creo (y espero hacer pruebas de ello para convencerme) que

en la region de E lorio y B eriz se oyen esos Markôsen y
Bilbôra, pero la tonalidad general que aqui se analiza
réserva siempre la atonia para la ultima silaba. (Azkue
1 9 3 1 , 12 )

[I believe (and I hope to test it in order to convince myself)

that in the region of Elorio and Beriz those forms Markôsen
and Bilbôra are heard, but the general tonality analyzed
here reserves the atony always for the last syllable.] (my
translation, J H)

Azkue does not make any reference to the first syllable

being lower in pitch than the rest in the unmarked case. For him
Basque words are either monotones, which bear a sustained high
tone, or ditones, w hich have a final drop in pitch. It is
conceivable that the dialect that he described lacked the rule of
Initial High Delinking. It could also be the case that the initial
lowering, which is not phonologically significant escaped his
a t t e n t io n . 3

13 The description of the Ondarroa facts presented here, on the other hand,
differs in some respects from the data found in Rotaetxe (1979) (and Hualde
(1987), who relies entirely on Rotaetxe's data). The most important difference
is precisely in those cases w here displacem ent of a low tone from its
underlying position takes place, such as in plural benefactive and locative
forms. Rotaetxe gives in these cases forms essentially identical to those given
here for the Gernika dialect. W hatever the reason for this disagreement in the
facts, my informants rejected forms given by Rotaetxe such as [mendiCletatik]
as utterly foreign to Ondarroa.
 286

2.3. Tone rules and Lexical Phonology

The rules of Tone A ssignm ent m ust follow all rules
affecting vowel sequences, since the surface tonal patterns take
into account the number of syllables found in the surface, which
can be different from the underlying syllables before reduction
of vowel sequences by various rules. Those observe the contrast
betw een /etS e-a/ [etSf] 'the house (abs)' and /etSe-àk/ [étS ik]
'the houses (abs/erg)'. If tonal rules applied before the reduction
of the vowel sequence we would obtain a plural tonal pattern
[etùîk], where the first syllable has a low tone and the syllable
resulting from the contraction of the underlying second and third
has both a high and a low. This is essencially the pattern that we
find in G ernika where the sequence is not reduced to one
syllable: [etsfek]. Interestingly, some speakers of the Gernika
area optionally delete the vowel of the genitive marker after [u]
and the tonal patterns are accordingly accom m odated to the
surface num ber of syllables. There is thus variation between
singular form s [katûén], [katun] form /katu-en/ 'of the cat' and
plural form s [katuen], [katun] from /k a tu -e n / 'of the cats'.
W hether the syllable /ka/ will appear with a high tone in the
plural depends entirely on whether the word is pronounced with
two or three syllables.
Tone Assignment, on the other hand precedes cliticization.
This can be observed quite clearly in Ondarroa, where only the
last syllable of the word can be low it there is an underlying low
 287
tone. To determ ine this dom ain c litic s are excluded from
consideration. In a form such as /lagun-àn-a da/ [lagunana ra]
'it is of the friends', the copula /da/ shows its status as a clitic
in that its 161 is realized as a sonorant and however does not
count for low tone assignment, the low tone is assigned to the
rightmost syllable in a domain without the clitic: /lagunana/.

2.4. Altzaga (G o ih e rri)i^

In this dialect words in isolation have a high tone on either
their first or second syllable and a second high tone on their last
syllable. This second high tone is actually assigned at the phrase
level, as the examples in (54) show:

(54) [lu re ta n ] in the lan d s’

[lû re ta n dawdé] they are in the lands'
[lu r e ta tik ] 'fro m the lands'
[lu r e ta ti ta to s ] they are coming fro m the lands'

Phonologically significant is only the place of the first

high tone. In the singular this high tone is generally placed on
the second syllable from the left:

(55) [etëéa] 'the house (abs)'

[baéôé] 'the fo re s t (abs)'
[gisona] the man (abs)'

14 I am grateful to José Antonio Sarasola for the data in this section.

 288

[mendisaha] t h e m o u n t a i n e e r (abs ) '

[katuk] t he c a t ( e r g ) ’

[lurén] in t h e land'

[ l u r é t i k dat ôr ] i t c o m e s f r o m t h e land'

[badora doa] he g o e s t o t he f o r e s t '

[badon d a w ] he i s in t h e f o r e s t '

[ b a é ô t i k datôr] he c o m e s f r o m t h e f o r e s t '

[ me n d i a noa] '1 go t o t h e mountain"

[ me n d f t i nat or ] '1 c o m e f r o m t h e m o u n t a i n '

[ me n d f k o a da] It i s t h e one o f t h e mo u n t a i n '

There is, however, as In other dialects, a number of nouns

that exceptionally have a high tone on their first syllable in the
singular; among them /libu ru/ 'book', /bàéo/ 'glass' and /b à é e ri/
'fa rm ':

(56) [l iburua] t h e book ( a b s ) ‘

I b hé oh] the g l a s s ( a b s ) ’

[ b é é e r i n daw] ‘he i s on t h e f arm'

[ b a é e r i a noa] I am g o i n g t o t h e f a r m ’

[ ba ^ e r i t i nat or] "I am c o m i n g f r o m t h e f a r m

In all nouns, unmarked and exceptional, the high tone is

placed on the first syllable in the plural:
 289

(57) [étSeék] the h o u s e s ( a b s /e r g ) '

[gisonak] t h e me n ( a b s / e r g ) '

[ba^oak] 'the f o r e s t s ( a b s / e r g ) ' or t he

glasses'

[ m é n d i e t a r a noa] 'I am g o i n g t o t h e m o u n t a i n s '

[méndietan] in t h e m o u n t a i n s '

[méndietatik] from the mountains'

[ m é n d i e t a k o a da] 'It i s t he one f r o m t h e m o u n t a i n s '

[ l û r e t a k o a da] i t i s t he one o f t h e l and s '

Exceptional nouns have the same pattern in the singular and

in the plural.
In the case of the unmarked singular pattern, the following
rules will give us a second syllable high tone:

(58) I. Make leftmost syllable extratonal

II. Assign a high tone to leftmost tonal syllable

The presence of a plural m arker makes the extratonality

rule inapplicable, and the high tone will be assigned to the first
syllable.
At the phrase level, another rule applies to assign a high
tone to the rightmost syllable.
 290
Interestingly, items that are exceptional in the singular in
that they receive the tonal pattern that normally corresponds to
p lu ral form s, are not d iffe re n t from o th e r item s in the
indefinite. This fact is also reported by Michelena (1981, 60) for
the also Guipuscoan dialect of Errenteria, which nevertheless
seems to possess a general system more similar to that seen for
O n d a r r o a M i c h e l e n a states that even though in this dialect,
as seemingly in all dialects of the pitch-accent type (and also in
the B otzerrieta stress variety) there are some w ords like
a a ld e ra 'the question' which have an anomalous pitch pattern in
the singular (which is identical in pattern to the plural for these
words), these same words are not exceptional in the indefinite,
as in lau aaldera 'four question(s).''* ® S im ilarly /to k i/ 'place'
and /leku/ 'place' present an anomalous pattern in the singular
form s t o k i v à n , l e k u à n 'in the place'; but not in the indefinite
form s t o k i t a n . lekutan 'in place'. This is in contrast with the
situation in Gernika and O ndarroa where exceptional words in
the singular are also exceptional in the indefinite number. The
m inim al pair offered by the w ords for 'fo re st' and 'glass'
contrasts in all three num bers in G ernika, but only in the
singular in Altzaga:

Unfortunately I do not have any direct knowledge of this system . My

Errenteria informants, of a younger generation than M ichelena, had lost the
plural/singular tonal opposition. Here I must thank lhaki Alberdi and Jose Mari
Carrere. The loss of the pitch-accent system of oppositions is something that I
have observed in young speakers of other localities as well.
16 Michelena only marks if the last syllable of the word is high (with an acute
accent)- or low (with a grave accent)
 291

( 5 9 ) a. Gerni ka

singular pl ural Indefinite

baéôâ baéôak la w baéo f o r e s t ' ( un ma r k e d )

baéoa badoak l a w bèéo glass' (marked)

b. A l t z a g a

baëoa baëoak l a w baéo 'forest'

baéoâ bâéoâk l a w baëô 'glass'

In A ltza g a e xcep tion a l item s m ust be m arked as

e xce p tio n a ly not undergoing the e x tra to n a lity rule in the
sin g u la r.

2.5. Arbizu
This prosodic system is acoustically very sim ilar to the
one just described. Phonologically, however, these two systems
are radically different. As in Altzaga, words in isolation present
both an initial and a final high tone. The location of these tones
does not, however, allow any phonological opposition. The tonal
pattern of plural nouns is identical to that of singular nouns. In
both sin g u la r and plural form s the firs t and last syllable
normally receive a high tone:

( 6 0 ) [ i t Sal [ i t § é k ] 'house' ( a b s s g , pi )

[agwa] [agwek] 'mouth'

[buuba] [büubékl 'head'

 292

[ ké ri ké] [ kéri kék] street"

[ kf Suë ki xa] [ki SuSki xék] "mason"

[ i t Saa] "to t h e h o u s e ( a d l ) ‘

[ mé n d i r e h ô ] "up t o t h e mountain"

[ménditaaho] "up t o t h e m o u n t a i n s '

There is no possible suprasegm ental distinction between

segm entally identical items such as the words for 'glass' and
'forest' in other dialects, which are hom ophonous in Arbizu:
[b àéuà].
O ptionally, in words of four syllables or more, the first
high tone can be realized on the second syllable from the left:
[ a r& is u a ra ] or [arfeisuara] 'native of Arbizu', with no difference in
meaning.
We can account for tone placement in Arbizu with the same
rules given for Altzaga; with the difference that here there is no
o bligatory extra to na lity of the firs t syllable in the singular.
Extratonality, which places the first high tone on the second
syllable is, rather, an optional rule for all words of more than
two syllables, singular or plural.

3. Discussion
After this survey of Basque prosodic types, a number of
questions come immediately to mind. One of them is what the
historical origin of the present diversity could have been and, in
 293

particular, w hat the historical relation between stress-accent

and pitch-accent system s might have been. W hich prosodic
system derived from the other, and how? In terms of change in a
phonological system , how can pitch-accent become stress or
viceversa? To try to provide an answer for these questions
would be well beyond the scope of this chapter and of this
disse rtatio n . Aspects of the issue of the historical relation
among Basque prosodic types are addressed in Michelena (1985),
Jacobsen (ms.) and Hualde (1986).
A related question is what are the physical correlates of
pitch-accent and stress-accent in different Basque varieties.
From the point of view of the trained listener, the difference
between pitch-accent varieties and stress-accent varieties is,
generally speaking, quite clear. In this respect, M ichelena's
im p re ssio n istic d escriptio ns of d iffe re n t prosodic types are
rather telling. For M ichelena, him self a speaker of a pitch-
accent variety, those stress varieties where stress is assigned
from the right edge, sound foreign, non-Basque. In fact, for him
they recall the way some Romance languages sound. With respect
to the Souletin-Roncalese stress type he says: "This type strikes
the Western hearer as being kindred to the Romance stress type;
more similar, perhaps to the (Occitan) Gascon type than to the
C astilian Spanish type (M ichelena (1972, 112)." He makes
sim ilar remarks about the High Navarrese type which we have
described for Baztan:
 294

ft is often said that Basque in the mouths of High

Navarrese people "sounds like Spanish, " and, if the inquiry
proceeds further, the statem ent is substantiated on the
grounds th at the d iffe re n ce betw een stressed and
unstressed syllables is very sim ilar to w hat can be
observed among speakers of Castilian Spanish. (Michelena
(1972, 113)

In his (1976) article he adds in reference to this High

Navarrese type: "cuando oimos hablar asi, nos parece que la letra
vasca Neva musica romance [when we hear somebody speak that
way, it sounds to us like Basque lyrics are being sung to
Romance music] (151)."
Equally interesting is the account that Basterrechea (1974)
gives of what makes the prosodic system of a variety such as
G ernika Basque e sse n tia lly d iffe re n t from th at of stress-
languages such as Spanish, English, German or Russian.
Basterrechea observes that whereas we can answer the question
of which syllable bears primary stress in a Spanish word such as
p a ia rra co or an English word such as c o n trib u tio n , that question
is senseless with respect to the Basque word e zp a ta d a n tza ri . In
the case of the Basque word (in the G ernika dialect), he
observes, we would have to answer that all syllables after the
first are equally stressed; because what we have is a rise in
pitch between the first and the second syllable and that level of
pitch is maintained from the second to the last syllable. If we
compare a pitch-accent variety such as Gernika with the stress-
accent dialects of Baztan or Soule, the difference between one
 295

and the other prosodic system would seem very great indeed.
Other pitch-accent varieties, like those of Altzaga or Arbizu do
not seem, however, that far removed from stress dialects. In
Arbizu or Altzaga, a word like e z p a ta d a n tz a ri is not pronounced
with the sustained high tone from the second syllable that is
found in G ernika. Rather, only two prom inent, high-toned,
syllables can be identified. Instead of the Gernikan pronunciation
e z p a t a d a n t z a r i ■ in Arbizu we obtain ezpatadantzav or
é zp a ta d a n tz â v . And yet, perceptually, there is a clear difference
between the tonal prominence that syllables receive in Arbizu
and stress. The instrumental analysis of the different prosodic
systems found in Basque can potentially shed much light on the
issue of the relation between stress and pitch-accent. My
hypothesis, very much along the lines of Beckman' s (1986)
suggestions and results, would be that it is the exclusive role
given to pitch which makes dialects such as Gernika, Ondarroa,
Altzaga, and Arbizu phonetically akin, in spite of profound
phonological differences, and which differentiates them from
stress varieties.
One must wonder, however, whether physical realization is
the only aspect that must be taken into account in determining
the m ost adequate analysis for a given prosodic system. In
current practice, there is a radical difference in the treatm ent
of stress and tone: stress is assigned by means of metrical
 296

rules, whereas tone is treated autosegm entally. A num ber of

Basque varieties, however, seem equally amenable to either
treatm ent. In footnote 15, it was indicated that in many areas
where older speakers possess pitch-accent systems of the types
described for Gernika and Ondarroa, younger speakers have lost a
number of phonological distinctions given by pitch patterns. In
some cases, distinctions are maintained only when there is one
syllable that can be defined as being more prominent than the
others in the word. Thus, some younger O ndarroa speakers
m aintain the difference between [katuk] 'the cat (erg)' and
[katuk] 'the cats (erg)' or between [basu] 'the forest (abs)' and
[basu] 'the glass (abs);' but do not distinguish between [tS aku rak]
'the dog (erg)' and [tS a kû ra k] 'the dogs (erg).' For these young
speakers, whether the last syllable is realized on a high pitch or
not in the two last examples seems to depend on intonation and
cannot be used to distinguish the singular from the plural. The
prosodic system of a dialect of this type would seem readily
characterizable by means of the metrical mechanisms generally
utilized in the description of stress systems. This is also true
for the Altzaga dialect, described in section 2.4. Nevertheless,
perceptually, the variety of the young Ondarroa speakers does
not greatly differ from that of older speakers of the same town.
A sim ilar state of affairs is described by Michelena (1975) for
the Biscayan dialect of Ohati (O hate),^^ which, as he puts it.

17 Ofiati is geographically situated in an area of Gipuscoa where the variety

spoken is traditionally considered a Biscayan subdialect
 297

never sounded strange to him. In spite of phonetic sim ilarities

with other western dialects of the pitch-accent type, in Ohati
singular b e l a u n a 'the knee' has the same prosodic pattern as
plu ral belaunak 'the knees.' But, b a s u a 'the fo re st' is
distinguished from b a su a 'the glass,' as in Gernika or Ondarroa.
This dialect, that of Altzaga and that of the young Ondarroa
speakers are cle a rly rem iniscent of the p e n -in itia l stress
system found in B ortzerrieta. In fact, in Hualde (1986), I
suggested that the Bortzerrieta system arose historically as a
consequence of a radical sim p lifica tio n in the system of
phonological rules of a pitch-accent system of the Ondarroa or
G ernika type. Nevertheless, there is a quite obvious phonetic
d iffe re n c e . The B o rtz e rrie ta d ia le c t p re s e n ts p h o n e tic
characteristics which are typical of stress languages. As we
saw in section 1.3, in this dialect there are processes of
unstressed vowel reduction and deletion which clearly mark it
as a stress system.
In this chapter, I have established a distinction between
stress-accent and pitch-accent varieties based on perceptual
differences which should be instrum entally verifiable. If we
compare the Soule stress facts, on the one hand, and the Gernika
pitch-accent facts, on the other, it seems sensible to conclude
that the rules resposible for the assignment of stress in Soule
are of a different nature from those rules which assign tonal
features in Gernika. The existence of fundamental differences in
 298

I the phonology seems rather less obvious if we compare, say, the

' B o rtz e rrie ta stress v a rie ty w ith the A ltza g a p itc h -a c c e n t
variety. One could conceivably have two essentially identical
rule-system s in two different dialects. In one dialect the rules
would directly assign tonal features; in the other they would
only assign prosodic prominence, to be realized in the surface
not only in terms of pitch but also by duration, vowel quality and
intensity. In one case, we would have a pitch-accent language
and in the other a stress language; but in terms of phonological
rules the languages could be rather sim ilar. The establishm ent
of a radical distinction between pitch-accent and stress-accent
must thus be reevaluated.

Hala bada, ha l a biz;

e t a e z bada, ora pro nobi s.

 299

References

Alonso, D. and V. Garcfa Yebra (1961) "El gallego-leonés de

Ancares y su interés para la dialectologia portuguesa, "
Cuadernos de Estudios Galleôôs 16.

Altube, S. (1923) El acento vasco. Bermeo.

Angulo, M. et al. (1983) "Bizkaiera," in Txillardegi, ed. EuskaI

dialektoloaiaren hastapenak. Udako EuskaI Unibertsitatea,
Irunea, 5-116.

Archangel!, D. (1984) Undersoecification in Yawelmani

Phonology and M orphology. Doctoral dissertation, MIT,
Cambridge, Massachusetts.

Archangel!, D. (1985) "An Overview of the Theory of Lexical

Phonology and Morphology," MIT Working Papers in
L in g u is tic s 7.

Archangel!, D. (1987) "Implications of the Maximal/Minimal

Parameter." Univ. of Arizona, ms.

Archangeli, D. and D. Pulleyblank (1986) The Content and

Structure of Phonological Representations. Univ. of Arizona
and u s e , ms.

Archangeli, D. and D. Pulleyblank (1987) "Maximal and Minimal

Rules: Effects of Tier Scansion," Proceedings of NELS 17.

Archangeli, D. and D. Pulleyblank (to appear) "Yoruba Vowel

Harmony, " Linguistic Inquiry.

Azkue, R. M. (1923) Morfolooia vasca. Bilbao.

Azkue, R. M. (1931) "Del acento tonico vasco en algunos de sus

dialectes," E u ske ra 4, 282-318 (part I).
 300

Azkue, R. M. (1932) "Del acento tonico vasco en algunos de sus

dialectes," E uskera 6, 3-50 (part II).

Basterrechea, J. (1974) "Estudios sobre entonaciôn vasca segùn

el habla de Guernica," Fontes Linguae Vasconum 18, 353-
393 (part I).

Basterrechea, J. (1975) "Estudios sobre entonaciôn vasca segùn

el habla de Guernica," Fontes Linouae Vasconum 21, 289-
338 (part II).

Beckman, M. (1986) Stress and Non-Stress Accent. Foris,

Dordrecht.

Bonaparte, L. L. (1864) Le verbe basque en tableaux, présentant

les formes des dialectes guipuzcoan. biscaïen. labourdin et
souletin. comparées entre elles: ainsi que les principales
variantes des autres dialectes, sous-dialectes et variétés
de l'euscara. Strangeways and Walden, London.

Bonaparte, L.L. (1873) "A Description of the Maps of the

Geographical Distribution of the Dialects, Sub-dialects, and
Varieties of the Basque Language," Athenaeum 61.

Carballo Calero, R. (1979) Gramatica elemental del aalleoo

co m u n . Galaxia, Vigo.

Chomsky, N. and M. Halle (1968) The Sound Pattern of English.

Harper and Row, New York.

Clark, M. (1987) "Japanese as a Tone Language," in Imai, T. and

M. Saito, eds., IssuBS in Japanese Linguistics. Foris,
Dordrecht, 53-103.

Clements, G. N. (1985) "The Geometry of Phonological

Features," Phonoloov Yearbook 2, 225-252.
 301

Cressey, W. (1978) Spanish Phonology and Morphology: A

Generative View. Georgetown Univ. Press, Washington, D.C.

Cummins, J. (1974) El habla de Coria. Tamesis, London.

DeCesaris, J. (1981) "On the Formulation and Interaction of

Certain Phonological Rules in Catalan," Cressey, W. and D. J.
Napoli, eds.. Linguistic Symposium on Romance Languages 9.
Georgetown Univ. Press, Washington, D.C., 24-36.

DeRijk, R, (1970) "Vowel Interaction in Biscayan Basque,"

Fontes Linguae Vasconum 5, 149-168.

Echaide A.M. (1976) "Distribucion de las variantes

palatalizadas y no palatalizadas de T y 'n' debidas al
contexte fonico en los dialectos vascos," Fontes Linguae
Vasconum 23, 163-169.

Eguzkitza, A. (1986) Topics in the Syntax of Basque and

R o m a n ce . Doctoral dissertation, UCLA, distributed by
Indiana University Linguistics Club.

Etxebarria, J. M. (1983) Euskal Dialektologiarako Teste eta

A rik e ta k . El Carmen, Vitoria-Gasteiz.

Euskaltzaindia (1985) Euskal Gramatika: Lehen Urratsak.

Institucion Principe de Viana, Iruhea.

Ewen, C. (1982) "The Internal Structure of Complex Segments,"

in van der Hulst, H. and N. Smith, eds.. The Structure of
Phonological Representations. Foris, Dordrecht, pp. 27-68.

Goenaga, P. (1982) Gramatika Bideetan. Erein, Donostia.

Halle, M. and K. P. Mohanan (1985) "Segmental Phonology of

Modern English," Linguistic Inquiry 16, 57-116.
 302

Haraguchi, S. (1977) The Tone Pattern of Japanese: An

Autosegmental Theory of Tonolocv. Kaitakusha, Tokyo.
. c,
Harris, J. (1969) Spanish Phonology. MIT Press, Cambridge,
Mass.

Harris, J. (1984) "Theories of Phonological Representation and

Nasal Consonants in Spanish," in P. Baldi, ed.. Papers from
the Xllth Linguistic Symposium on the Romance Languages.
Benjamins, Amsterdam, 153-168.

Holmer, N. (1964) El idioma vasco hablado. Diputacion de

Guipûzcoa, San Sebastian.

Holmer, N. and V. Holmer (1968) "Apuntes vizcamos," Anuario

del Seminario Julio de Urquiio 2, 87-141 (part I).

Holmer, N. and V. Holmer (1969) "Apuntes vizcamos," Anuario

del Seminario Julio de Urquiio 3, 171-227 (part II).

Hualde, J. (1986) "Tone and Stress in Basque: A Preliminary

Study," Anuario del Seminario Julio de Urquijo 20, 867-986.

Hualde, J. (1987a) "On Basque Affricates," Proceedings of the

West Coast Conference on Formal Linguistics 6. 77-89.

Hualde, J. (1987b) "Proto-Basque Yod," Fontes Linguae

Vasconum 49, 29-40.

Hualde, J. (1988) "Affricates are not Contour Segments," West

Coast Conference on Formal Linguistics 7.

Hualde, J. (to appear) "Delinking Processes in Romance," in

Kirschner, C. and J. DeCesaris, eds.. Proceedings of LSRL 17.
Benjamins, Amsterdam.
 303

Hualde, J. (in prep, a) "Compensatory Lengthening and Long

Vowels in Basque," USC ms.

Hualde, J. (in prep, b) "Metaphony in Asturian and Cantabrian

Romance," USC ms.

Hualde J. and J. Ortiz de Urbina (to appear) "Restructuring With

â l i Anuario del Seminario Julio de Urquijo.

Hayes, B. (1986) "Inalterability in CV Phonology," Language 62.

3 2 1 -5 1 .

Hayes, B. (1988) "Diphthongization and Co-indexing," paper

presented at The Arizona Phonology Conference, Tucson,
Arizona, April 1988.

Hyman, Larry (1975) Phonology: Theory and Analysis. Holt,

Rinehart and Winston, New York.

Hyman, L. (1984) A Theory of Phonological W eight. Foris,

Dordrecht.

Ito and Mester (1986) "The Phonology of Voicing in Japanese:

Theoretical Consequences for Morphological Accessibility,
Linguistic Inquiry 17. 49-74.

Iverson and M L. Ohederra (1985) "On Basque Palatalization,"

Folia Linouistica 19, 51-61.

Jacobsen, W. (1972) "Nominative-Ergative Syncretism in

Basque," Anuario del Seminario Julio de Urquijo 6, 67-109.

Jacobsen, W. (no date) "Historical Implications of the Western

Basque Tonal Accent," University of Nevada, Reno, ms.

Kaisse, E. and P. Shaw (1985) "On the Theory of Lexical

Phonology," Phonology Yearbook 2. 1-30.
 304

Kamprath, C. (1986) Suprasegmental Structures in a Raeto-

Romansh Dialect: A Case Study in Metrical and Lexical
P h o n o lo g y . Doctoral dissertation, Univ. of Texas at Austin.

Keating, P. (1987) "Tongue Body Features for Consonants:

Articulatory Evidence," presented at the Workshop on
Distinctive Features, USC, February 1987.

Kintana, X. et al. (1980) Hiztegia 80. Ekiten Taldea, Bilbo.

Kiparsky, P. (1973) "Elsewhere in Phonology," in Anderson, S.

and P. Kiparsky, eds., A Festschrift for Morris Halle. Holt
Rinehart and Winston, New York.

Kiparsky, P. (1982a) "From Cyclic to Lexical Phonology," in H.

van der Hulst and N. Smith, eds.. The Structure of
Phonological Representations (Part I), Foris, Dordrecht,
1 31- 176.

Kiparsky, P. (1982b) "Lexical Morphology and Phonology," in

Yang, l-S, ed.. Linguistics in the Morning Calm. Hanshin,
Seoul.

Kiparsky, P. (1984) "On the Lexical Phonology of Icelandic, " in

Elert, C., I. Johansson and E. Strangert, eds., Nordic Prosodv
3: Papers from a Symposium. Univ. of Umea.

Kiparsky, P. (1985) "Some Consequences of Lexical Phonology,"

Phonology Yearbook 2, 85-138.

Klavans, J. (1985) "The Independence of Syntax and Phonology

in Cliticization," L a n g u a g e 61, 95-120.

Ladefoged, P. (1982) A Course in Phonetics. 2nd. ed., Harcourt

Brace Jovanovich, New York.
 305

Lafitte, P. (1979) Grammaire Basque: Navarro-Labourdin

L itté ra ire . Elkar, Donostia.

Lafon, R. (1952) Études basques et caucasiques. Acta

Salmanticensia 5.2, Universidad de Salamanca, Salamanca.

Lafon, R. (1980) Le système du verbe basque au XVI& siècle.

2nd. éd., Elkar, Donostia

Larrasquet, J. (1928) Action de l'accent dans l'évolution des

consonnes étudiée dans le basque souletin. Vrin, Paris.

Larrasquet, J. (1934) Le basque souletin nord-oriental Paris.

Larrasquet, J. (1939) Le basque de la Basse-Soule orientale.

Paris.

Lass, R. (1976) English Phonological Theorv. Svnchronic and

Diachronie Studies. Cambridge Univ. Press, Cambridge.

Lozano, M.C. (1979) Stop and Soirant Alternations: Fortition

and Spirantization Processes in Spanish Phonology. Doctoral
Dissertation, Ohio State Univ., Indiana University
Linguistics Club, Bloomington, Indiana.

Mackay, I. (1987) Phonetics: The Science of Speech Production.

Little Brown, Boston.

Malberg, B. (1955) Estudios de fonética hispânica. Institute

Miguel de Cervantes, Madrid.

Mascaro, J. (1978) Catalan Phonoloov and the Phonological

C ycM , Doctoral dissertation, MIT, distributed by Indiana
University Linguistics Club, Bloomington, Indiana.
 306

Mascaro, J. (1984) "Continuant Spreading in Basque, Catalan,

and Spanish," in M. Aronoff and R. Oehrie, eds.. Language
Sound Structure. MIT Press, Cambridge, Mass.

Mester, R.A. (1986) Studies in Tier Structure. Doctoral

dissertation, Univ. of Massachusetts, Amherst.

Michelena, L (1958) "À propos de l'accent Basque." Bulletin de

la Société de Linguistique de Paris 53, 204-233.

Michelena, L. (1972) "A Note on Old Labourdin Accentuation,"

Anuario del Seminario Julio de Urquijo 6, 110-120.

Michelena, L. (1976) "Acentuaciôn alto-navarra," Fontes

Linguae Vasconum 23, 147-162.

Michelena, L. (1981) [Mitxelena, K.] "Galdegaia eta mintzagaia

euskaraz," Euskal Linouistika eta Literatura: Bide berriak.
Univ. de Deusto, Bilbao, 57-81.

Michelena, L. (1985) Fonética histôrica vasca. 3rd. éd.,

Diputacion de Guipûzcoa, San Sebastian.

Mohanan, K.P. (1986) The Theorv of Lexical Phonoloov. Reidel,

Dordrecht.

Moutard, N. (1975a) "Étude phonologique sur les dialectes

basques, I," Fontes Linouae Vasconum 19, 5-42.

Moutard, N. (1975b) "Étude phonologique sur les dialectes

basques. II," Fontes Linouae Vasconum 20, 141-190.

Moutard, N (1976) "Étude phonologique sur les dialectes

basques. III," Fontes Linouae Vasconum 22, 9-54.

Mujika, L. M. (1978) Orioen v desarrollo de la sufiiaciôn

euskérica. Ediciones Vascas, Bilbao.
 307

Mujika L. M. (1982) Latinaren eta erromanikoaren eraaina

euskaran. Sendoa, Donostia.

Navarro Tomâs, T. (1922) "Observaciones fonéticas sobre el

vascuence de Guernica, III Conareso de Estudios Vascos. 57-
64.

Navarro Tomâs, T. (1925) "Pronunciacion guipuzcoana:

contribuciôn al estudio de la fonética vasca," Homenaje a
Menéndez Pidal. Ill, 593-653.

N'Diaye, G. (1970) Structure du dialecte Basque de Mava.

Mouton, The Hague.

Nûhez-Cedeho, R. (1987) "Intervocalic /d/ Rhotacism in

Dominican Spanish: A Non-Linear Analysis," H ispania 70.
3 6 3 -3 6 8 .

Ortiz de Urbina, J. (1986a) Some Parameters in the Grammar of

Basque. Doctoral dissertation. Univ. of Illinois at
Champaign-Urbana. Revised version to be published by Foris,
Dordrecht.

Ortiz de Urbina, J. (1986b) "Aspects of Basque Lexical

Phonology," Fontes Linouae Vasconum 47.

Poser, W. (1982) "Phonological Representations and Action-at-

a-Distance," in van der Huls, H. and N. Smith, eds.. The
Structure of Phonological Representations (part II), Foris,
Dordrecht, 121-158.

Poser, W. (1986) "Blocking out of the Lexicon," Stanford Univ.,

ms.

Pranka, P. (1983) Syntax and Word Formation. Doctoral

dissertation, MIT, Cambridge, Massachusetts.
 308

Pulleyblank, D. (1986) Tone in Lexical Phonology. Reidel,

D ordrecht.

Pulleyblank, D. (1988) "Underspecification, the Feature

Hierarchy, and Tiv Vowels," presented at The Arizona
Phonology Conference, Tucson, Arizona. Revised version to
appear in Phonologv Yearbook.

Pulleyblank, D. (1988) "Vocalic Underspecification in Yoruba,"

Linguistic Inquiry 19, 233-270.

Rollo, W. (1925) The Basque dialect of Marquina. H.J. Paris,

Am sterdam .

Rotaetxe, K. (1978) "L'accent basque: observations et

hypotheses," La Linguistique 14, 56-77.

Rotaetxe, K. (1979) Estudio estructural del euskara de

Ondarroa. L. Zugaza, Durango, Vizcaya.

Sagey, E. (1986) The Representation of Features and Relations

In Non-Linear Phonology. Doctoral dissertation, MIT,
Cambridge, Massachusetts.

Salaburu, P. (1983) Arau fonoloaikoak. Univ. del Pais

Vasco, Bilbao.

Salaburu, P. and X. Kintana (1984) Euskara hobean hobe. Univ.

del Pafs Vasco, Bilbao.

Saltarelli, M. (1988) Basque. Croom Helm, Beckenham.

Schein, B. and D. Steriade (1986) "On Geminates," Linguistic

In q u iry 17, 691-744.

Selkirk, E. (1982) The Svntax of Words. MIT Press, Cambridge,

Mass.
 309

St. Clair, R. (1972) "Compound Phonological Segments," Lingua

29, 120-127.

St. Clair, R., (1973) "Affricates as Compound Phonological

Segments," Linouistische Berichte 26, 21-24.

Steriade, D. (1987) "Redundant Values," Chicago Linguistic

S ociety 23.2, 339-362.

Steriade, D. (1988) "Locality Conditions and Feature Geometry,

Proceedings of NELS 17. Univ. of Massachusetts, Amherst.

Tato, P. (1981) "Romance Phonological Evidence for the

Noncontinuant Status of /I/," in W. Cressey and D. J. Napoli,
eds.. Linguistic Symposium on Romance Languages 9.
Georgetown Univ. Press, Washington, D.C.

Txillardegi (1978) Euskal G ram atika. Ediciones Vascas, Bilbao.

Txillardegi (1980) Euskal Fonolooia. Ediciones Vascas, Bilbao.

Txillardegi (1985) Euskal Azentuaz. Elkar, Donostia.

Venneman, T. (1977) "Rule Inversion," Lingua 29, 209-242.

Villasante, L. (1976) Palabras vascas compuestas y derivadas,

Aranzazu, Ohate, Guipûzcoa.

Williams, E. (1981) "On the Notions 'Lexically Related' and

'Head of a Word,' " Linguistic Inquiry 12, 245-74.

Yrizar, P. (1981) Contribuciôn a la dialectoloofa de la lengua

vasca. CAP, Donostia.
 310

iH

Î
5® d
rm
o>

o\

V .