Phonology of Modern Javanese

INFORMATION TO USERS
This material was produced from a microfilm copy of the original document. While
the most advanced technological means to photograph and reproduce this document
have been used, the quality is heavily dependent upon the quality of the original
submitted.
The following explanation of techniques is provided to help you understand

markings or patterns which may appear on this reproduction.
1.The sign or "target" for pages apparently lacking from the document
photographed is "Missing Page(s)". If it was possible to obtain the missing
page(s) or section, they are spliced into the film along with adjacent pages.
This may have necessitated cutting thru an image and duplicating adjacent
pages to insure you complete continuity.
2. When an image on the film is obliterated with a large round black mark, it
is an indication that the photographer suspected that the copy may have
moved during exposure and thus cause a blurred image. You will find a
good image of the page in the adjacent frame.
3. When a map, drawing or chart, etc., was part of the material being
photographed the photographer followed a definite method in
"sectioning" the material. It is customary to begin photoing at the upper
left hand corner of a large sheet and to continue photoing from left to
right in equal sections with a small overlap. If necessary, sectioning is
continued again — beginning below the first row and continuing on until
complete.
4. The majority of users indicate that the textual content is of greatest value,
however, a somewhat higher quality reproduction could be made from
"photographs" if essential to the understanding of the dissertation. Silver
prints of "photographs" may be ordered at additional charge by writing
the Order Department, giving the catalog number, title, author and
specific pages you wish reproduced.
5. PLEASE NOTE: Some pages may have indistinct print. Filmed as

received.
Xerox University Mcrofsllms

3 0 0 North Z eeb Road
Ann Arbor, M ichigan <18106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
76 - 1 6 ,1 2 6
DUDAS, Karen Marie, 1946-

THE PHONOLOGY AND MORPHOLOGY OF MODERN
JAVANESE.
University of Illinois at Urbana-Champaign,
Ph.D., 1976
Language, linguistics
Xerox University Microfilms, Ann Arbor, Michigan 48106
THE PHONOLOGY AND MORPHOLOGY
OF MODERN JAVANESE
BY
KAREN MARIE DUDAS
B.A., Cornell College, 1968
THESIS
Submitted in partial fulfillment of the requirements
for the degree of Doctor of Philosophy in Linguistics
in the Graduate College of the
University of Illinois at Urbana-Champaign, 1976
Urbana, Illinois
U N IV E R S IT Y OF IL L IN O IS A T U R B A N A -C H A M P A IG N
TH E GRADUATE COLLEGE
.December.,, 1975
W E H E R E B Y R E C O M M E N D T H A T T H E T H E S IS BY
_______________ KAREN, TIARJE..DUDAS______ ____________
E X T FT IT '.!)_._...... THE PHONOLOGY AND MORPHOLOGY OF MODERN JAVANESE
BE A C C E P T E D IN P A R T IA L EC I.F I 1.I..M E N T O F T H E R E Q U IR E M E N T S FOR
T H E DEG R EE DOCTOR OF PHILOSOPHY
Ifirecto r of T li/iis R ese arch
D epar t
C o m m itte e on F in a l E x a m in a tio n ^
Chatrmai
___
t S' y-0
'I 'iG.1 ACtC LS •;.0'uULC'v
t R e t i u i r e d f o r d o c t o r ’s d e g r e e b u t n o t f o r m a s t e r ’s.
iii
ACKNOWLEDGMENTS
The many people who have in one way or another bene-
fitted the author in the course of her studies at the
University of Illinois are far too numerous to list individ
ually; they include nearly all th<? faculty and staff and
many of my fellow students in the Department of Linguistics.
If a list were to be made, however, places near the top would
surely be reserved for my advisor, Prof. Charles Kisseberth,
as well as for other members of my thesis committee, Prof.
Hans Henrich Hock and Prof. Michael Kenstowicz, all of whom
have significantly influenced my development as a linguist.
In addition, I would wish to thank my informant and valued
friend, Suharto Prawlrokusumo, without whose help the writing
of this dissertation would have been impossible. Finally,
many thanks go to the fourth member of my thesis committee,
Prof. Margie O'Bryan, who has been a special friend through
out the course of my graduate career, helping me to get
through the worst times (Including the throes of writing this
dissertation) and sharing in some of the best times.
It would also seem appropriate to express my gratitude
to the various agencies and organizations whose monetary sup
port has made my graduate studies possible. These include the
National Science Foundation, the Institute for Research on"
Exceptional Children, and the University of Illinois Research
Board.
iv
PREFACE
As its title indicates, this dissertation will deal with
the phonology and the morphology of Modern Javanese. However,
this very general statement of subject matter may be some
what misleading, and, thus, it seems that a certain amount
of clarification is in order.
First of all, the language to be dealt with is indeed
spoken in modern-day Java; however, the general label
"Modern Javanese" covers a wide area and needs some qualifi
cation. Modern Javanese speakers make use of two major
speech levels: Ngoko is used for informal every-day conver
sation, while Krama is the level used in business or polite
conversation (there are also a number of sub-classes within
each of these major levels). The two major levels differ
from each other mainly in their inventories of lexical items,
but also, to some degree, in morphology. This dissertation
will deal exclusively with the Ngoko or informal speech.
Furthermore, discussion will be based on data collected in
the course of two year's work with a single informant,
Suharto Prawirokusumo, a graduate student at the University
of Illinois. He is a native speaker of Javanese who grew up
in the Surakarta area, where the dialect which is generally
accepted as Standard Javanese is spoken.
A second area in which clarification is necessary
involves the degrees to which the areas of phonology and

morphology will be treated. Chapters 2 and 3 of this
dissertation form an Integrated whole in which a relatively
detailed generative analysis of the phonology of the language
will be built up, step by step. However, the remainder of
the dissertation consists of four chapters which are only
loosely related, Insofar as all deal with some facet of
morphology; no attempt will be made to build up a complete
analysis of Javanese morphology. There are good reasons
for this departure from the course followed with the phono
logy of the language. Not the least of these is the fact
that, at this time, there is no satisfactory theoretical
framework readily available for setting up any sort of com
plete morphological analysis; generative linguists are still
at a fairly early stage in their investigation of the morpho
logical aspects of language. At the same time, while a
number of the Javanese morphological processes are so
straightforward that their analysis would cause no difficulty,
examination of each in turn would Involve a great deal of rep-
itition and would probably turn out to be of little value.
ft
Finally, it should be pointed out that I do not claim
to be presenting startling new data in this dissertation;
nearly all Javanese constructions are to be found In the
various Javanese grammars and dictionaries which are avail
able. Nor can it be claimed that this is the first time any
serious linguistic analysis of Javanese has been attempted;
to make such a claim would do shameful disservice to a great
deal of excellent work done in the past, most notable, that
Vi
of E. M. Uhlenbeck. What I would claim for the present

work, however, is that, by approaching the analysis of
Javanese from a generative point of view, I have been able
to produce a more insightful and natural analysis of the
phonology of the language than any which has been offerred
in the past. Furthermore, it is to be hoped that, in the
course of discussions contained in this dissertation, partic
ularly those dealing with morphology, some contribution has
been made toward furthering linguists’ knowledge of the ways
in which morphological processes may function in the grammars
of natural languages, and, in addition, how such processes
may interact with others contained in grammars.
v ii
TABLE OP CONTENTS
CHAPTER Page
I. INTRODUCTION................................. 1
II. VOWEL PHONOLOGY............................. 33
III. CONSONANT PHONOLOGY........................ 118
IV. CAUSATIVES AND LOCATIVES................... 1*15
V. ELATIVE FORMATION.......................... 180
VI. REDUPLICATION.............................. 202
VII. HABITUAL-REPETITIVE........................ 229
REFERENCES........................................ 265
VITA..............................................267
1
CHAPTER I
INTRODUCTION
1.0 In this first chapter, it is my aim to provide
the reader who is unfamiliar with the Javanese people and
their language with a certain amount of background infor
mation to aid him in his understanding and evaluation
of the linguistic analyses which will be presented in
later chapters. A very brief sketch of some of the major
events in the history of the Javanese people and language
will be included, as well as a description of some of the
more basic facts about the modern-day language.
1.1 Although there already existed an indigenous
population on the island of Java, migrations began as early
as 3000 B.C., when the Malay ancestors of today’s Javanese
people came to the island. Their agrarian culture
flourished on its own until near the beginning of the
Christian Era. At that time, the expansion of Asian trade
led to increased contact with Indian civilization. In the
course of the following centuries, the impact of Indian
civilization was strongly felt In all areas of Javanese
life, but especially in the area of religion, where
Hinduism and, later, Buddhism became widespread.
It was during this period that the written literature
of Java began to develop. Using first an Indian writing
system, and, later, a Javanese system based on the latter.
2
Javanese scholars translated Sanskrit works Into their

own language. Soon there arose a native literature using
the Old Javanese written language, or, more precisely*
the Kawi or "poetic” language.
It Is significant that, though the spoken language
must by this time have also contained many words borrowed
from Sanskrit, the written language was made up of perhaps
90J5 Sanskrit borrowings, used within a Javanese grammatical
framework. As the spoken language changed in the course
of time, Kawi remained more or less the same, so that even
today a version of it is still In use as the literary
poetic language of Java.
With the rise of Islam In the fifteenth and sixteenth
centuries, Java entered Into a period of cultural and
political unrest. When opportunities for trade brought
the Portuguese to the Island In 1511, they became engaged
In a power struggle with Islamic traders which eventually
developed into a religious war.
Beginning In 1596, the Dutch, also brought-to the
East Indies by the magnificent possibilities for trade,
through various circumstances managed to gain the upper
hand. They took charge of commercial activities on the
island and retained control for the next 350 years.
As would be expected, during this period of Dutch
supremacy a great many changes took place in Javanese
life and, consequently, a great many words of European
3
origin came Into Javanese. As the island developed

politically and economically under the firm hand of the
Dutch, borrowings entered the language from the realms of
bureaucratic organization, the military and, later, modern
technology, transportation, etc. In the course of the
years, many of these originally foreign words have been
remodeled and Integrated into the native vocabulary to
such an extent that they are now indistinguishable from
other Javanese vocabulary.
The Dutch rule of Java lasted until World War II,
when the Japanese occupied the Island in 19^2. Through
out the preceding years, there had been a growing spirit
of nationalism throughout the islands of Indonesia, and
with the fall of the Japanese invaders in 19^5, the
opportunity to gain independence was seized and the
Republic of Indonesia was born.
Today, the island of Java supports over 60% of the
population of the Republic of Indonesia. The official
language of the nation is Indonesian, which is the descend
ant of the Malay dialect used for centuries as a lingua
franca among the islands. Even on Java, Indonesian is
the language used for business transactions and official
communications, as well as in the schools.
Thus, in the course of the 30 years since Indonesia’s
independence was declared, Javanese has been relegated
to a secondary position, today being spoken mainly In
informal settings. This factor, along with the breakdown
of old social structures which came with Indonesian
independence, has led to a great reduction in the impor
tance of formal speech levels within the language.
1.2 At least since the seventeenth or eighteenth
century, speakers of Javanese have made use of two main
social speech levels: Ngoko, the everyday informal
dialect used among close friends and for speaking to
one's social inferior; and Krama, literally "correct"
speech, which is used for speaking to one of a higher
social level, to one's elders, or to strangers. These
two levels are distinguished mainly by having different
lexical items to refer to certain basic objects and
concepts. Thus, for example, the Ngoko word for 'house'
is Omah, while the Krama word is grlJo. In general, In
order to speak Krama, one substitutes these different
lexical items into ordinary sentences; however, there are
also a few Ngoko affixes which have Krama counterparts.
In addition to the two major levels, there are
numerous other sub-parts to the vocabulary of Javanese
to be used In various more subtle social situations.
(Horne (197^:xxxll) Isolates ten separate speech levels.)
For example, Sumukti (1971:10) describes the level called
Krama Andap as the level "whose forms are used to honor
9
a person by lowering the status of another person who

does some action in regards to the person to be honored".
Other levels Include Krama Inggel, to be used to show
particular respect when speaking of a person who is not
present, as well as Krama Madja, to be used in situations
which socially require one to use Krama, but which are
too informal for ordinary Krama.
It is not difficult to see that such a complex system
of speech levels could only function well in a society
where social roles are clearly defined, so that each person
readily knows what his status is with regard to that of
anyone he must deal with. With the freedom and rapid
changes which characterize Indonesian life today, the more
subtle distinctions of this type have become impossible to
maintain. Furthermore, because Indonesian is the official
language, relegating Javanese to the home or to use among
close friends, very little opportunity for using the
vocabulary of the more formal speech levels arises. As a
case in point, my Javanese informant tells me he feels
very ill at ease using Krama, even though he learned it
as a child; and his own children will probably not ever
learn to use it at all.
1.3 The Javanese language makes use of ten distinct
surface vowel sounds which may be represented phonetically
as follows:
i u
1 V
a
e o
6
Throughout this dissertation, in citing Javanese forms,

I will refer to these sounds by means of the following
orthographic symbols:
I 0
i u
E fl O
e o
a
I.1! The surface consonant system of Javanese consists
of the following 27 sounds, arranged according to the
classifications given in Horne 196l:xxix:
Light Heavy Nasal Nasalized Others
Heavy
Labial P b m mb w
Dental t d n nd
Alveolar t• d• nd
u
0)
Palatal tj dj nj ndj
Velar k g ng ngg
Lateral 1 lh
Glottal h,q
Although, for simplicity, the distinction between the
series labeled "light" and "heavy" may be thought of as
being chiefly one of voiceless versus voiced consonants,
strictly speaking, this is not sufficient. Rather, the
difference in question is one more accurately described by
Horne (196l:xxix): "The light consonants are sharp and
clear, while the heavy consonants have a murmured, fuzzy
quality. In addition, the heavy consonants affect the
vowel after them by making it a bit lower in pitch and
and giving it a breathy sound."
In addition to the sounds listed here, Javanese makes
occasional use of a number of consonant sounds in words
borrowed from other languages, such as f , ch, etc.
Usually, such foreign sounds are used only by more sophis
ticated speakers, while the majority are replaced by the
nearest native Javanese sound. Thus, for example, the word
borrowed from English as interview may be pronounced
InterfiU or InterplU, with the substitution of the native
bilabial.1
Although I have followed Horne, (1961) and included it
on the above chart, the sound written there as lh and
classified as a heavy lateral appears to be extremely rare
in the modern language. This is illustrated by the fact
that in Horne 1972*, only two words are listed with lh in
word-initial'position (the exclamations lha and lhO) and
in both cases the words are cited as variant pronunciations
of la and 10. The sound never occurs at all in word-final
position.
Finally, in addition to the consonant sounds listed
above, two glides occur in certain intervocalic environments
(cf. 3-3.3.2.2.1). These sounds, which I will write as ^ and
w, are nearly identical in articulation to the consonants
and w, but are less forcibly pronounced.
8
1.5 Within morphemes, the following syllable types

may occur:
V VC
CV CVC
CCV CCVC
Notice that, although syllables may begin with more than

one consonant, they never end In more than a single consonant.
It follows naturally from this that Javanese primary mor
phemes may begin with two-consonant clusters, but may never
end in clusters.
Morphemes may consist of one syllable, but the over
whelming majority are bisyllabic. Consequently, discussions
in this dissertation will be geared mostly toward bisyllabic
stems; the treatment of stems with more or fewer syllables
can easily be extrapolated from treatment of bisyllabics.
Although all possible bisyllabic combinations of the above
syllable types are represented in the morphemes of the
language, by far the most frequently occurring shape is
CVCVC. This can be seen from the fact that, According to
Uhlenbeck 1949:96, out of 11,100 Javanese bisyllabic
morphemes (with a single internal consonant), 6,358 are
of the CVCVC type, while another 1,307 are CVCV, 1,240 are
VCVC, and 1,789 are of the CCVCVC type.
1.5.1 Morphemes may begin with any single consonant
except £ and y_, and they may end with any consonant except
t_, d, tJ 3 dj, w, y, or any nasalized heavy consonant.
9
Javanese primary morphemes begin with the following types

of consonant clusters:
(1) £» b, k, £, n£ plus
(2) All stops except t and d, affricates, nasals,
• •
as well as £ and w plus 1

(3) All stops, affricates and nasals, as well as
£ and w plus r
The first type occurs only in a few words, mostly onomato
poeic and expressive (e.g. pjur 'representing spraying,
sprinkling'; bjah 'representing suddenly breaking apart'),
while types (2) and (3) are quite commonly found in all
types of vocabulary. In certain derived forms, the heavy
nasalized consonants mb, nd, nd, nd.1 and ngg may also occur
word-initially and with 1 or r, and, rarely, with £.
1.5.2 Within words, non-initial syllables may begin
with approximately the same types of clusters, with a few
additions. Internal (I.e. non-initial) syllables may begin
with any of the following combinations of consonants:
(1) A consonant plus a liquid
(2) A consonant preceded by its homorganic nasal.
(Note that, for purposes of syllabification,
n plus d, t, tj_ or dj_ and ng_ plus s act
o o
as though the nasal and consonant are

homorganic.)
(3) A consonant preceded by Its homorganic nasal
and followed by a liquid.
10
1.6 Within words, stem-penultimate open syllables are

preferred. Syllabification takes place in such a way that
a single internal intervocalic consonant will be pronounced
as a syllabic unit with the following vowel. Thus, bapaq
'father' is pronounced ba/paq; IbU 'mother' is syllabified
as I/bU.
If two or more consonants of the types described above
in 1.5.2 occur together internally, they are similarly
treated by the phonological vowel realization rules of the
language (cf. chapt. 2) as a syllabic unit with the following
vowel, and thus the preceding syllable Is open, so that,
for example, dOmbo 'variety of large goat' Is syllabified
as dO/mbo, while dOmblO 'chubby In the cheeks' Is divided
as dO/mblO.2
If two or more consonants occur In an internal cluster
which are not of the type described In 1.5.2, the first is
pronounced with the preceding vowel, closing the penultimate
syllable, and the second (and third, if there Is one) is
pronounced as a syllabic unit with the following vowel:
darmo 'duty' Is divided as dar/mo; sastro 'writings, letters'
is syllabified as sas/tro.
1.7 It Is significant that. In native vocabulary,
only the vowels a and g? are usually found in penultimate
closed syllables. In most cases, when other vowels are
found in this position, according to my Informant, the word
is felt to be either "non-native" (i.e. borrowed) or "old"
11
(i.e. part of the archaic Kawi literary language), and

accordingly seems to fall outside the scope of the normal
phonological rules of the language.
Although many items like bensin 'gasoline* and dlktE
'dictation' are in fairly common use by Javanese speakers,
most are still felt, at least by my informant, to be part
of the non-native vocabulary. This feeling is undoubtedly
to some degree a result of the very fact that such words
have vowels other than a or £ occurring in closed penultimate
syllables. At the same time, however, there are frequently
other structural clues to the non-native status of a given
word. For example, many fairly recent borrowings such as
lnfeksl 'infection', dEreksI 'director' and dEtektlf are
trisyllabic. At the same time, borrowings are recognizable
by the fact that they contain sounds which do not occur in
native words, or by the presence of unusual combinations
of sounds, as in lnfeksl, for example, which has the .
non-native phoneme f.
Generally in pronouncing such words, regardless of the
specific way in which they may differ from native vocabulary,
in contrast to the strong intuitions he usually has for
the "correct” pronunciation of native words, my informant
seems to have little feeling for what is the correct or
incorrect pronunciation; he will rate nearly all possible
pronunciations as being equally acceptable to him.
12
There are, however, some logical exceptions to this

apparently unrestricted pronunciation of non-native vocab
ulary. In cases where my informant is familiar with the
borrowed word as it was originally pronounced in the
language from which it was borrowed, he is likely to feel
that the Javanese pronunciation most like the original
is more correct. On the other hand, when asked how a
villager or a less sophisticated person would pronounce
the same word, he usually reverts to the more liberal
position of accepting nearly all possible variations as
being equally correct. Other exceptions consist of borrowed
lexical items which have two unrelated meanings: for example,
the word written biskop may mean either ’bishop* or
'theater'. According to my informant, the pronunciation
biskop usually refers to 'bishop', while blskOp denotes
'theater'.
1.8 One of the most important areas to be covered in
this introductory chapter is the morphology of Javanese.
Unlike Indo-European languages, which are highly inflected,
Javanese, like most other Malayo-Polynesian languages, has
very little inflectional morphology. There are no required
markings for tense, person, or number, although plurals
may optionally be marked by doubling of the stem. Thus, for
example, mUrlt can mean 'student' or 'students', but the
latter may optionally be represented as mUrlt-mUrit. To
the best of my knowledge, this optional doubling for plurals
13
is the only morphological process in the language which

can unequivocally be called Inflectional.
At the same time, however, Javanese is by no means
lacking in morphological processes; in fact, the language
possesses a rich derivational morphology which will play
a crucial role in the analyses presented in later chapters.
At this point, however, before any detailed analysis or
discussion of Javanese phonology or morphology is attempted,
it will be helpful to give a very brief sketch of certain
facets of the morphology in order to facilitate understanding
of the data which will be referred to in subsequent chapters.
It must be stressed that the aim of this initial description
is not to give a general overview of Javanese morphology,
but merely to begin to familiarize the reader with a few
of the morphological processes which will play a large role
in the actual body of this dissertation. At the same time,
no claims will be made about the actual operation of these
processes; a brief description of the segmental relationships
between surface primary roots and their various derived
forms is the only immediate goal.
The morphology of Modern Javanese includes processes
of a wide variety of types generally recognized by linguists:
prefixation, infixation, suffixation, and reduplication.
In addition, one might also speak of "circumfixation" in
•
Javanese, the latter being the marking of a given category

by the surrounding of a stem with both a prefix and a suffix.
14
However, for the purposes of this dissertation, I will

treat circumfixation as merely the application of the two
separate processes of prefixation and suffixation to a
single stem. This approach is possible because all affixes
which occur in so-called circumfixed forms also occur
either independently or in conjunction with other affixes;
that is, each affix may at some time occur in a derived
form minus "the other end" of the circumfix. Thus, although
the question of whether circumfixation should properly
be treated as a unitary morphological process is an
intriguing one, it will not be dealt with in the present
work.
1.8.1 Of all Javanese morphological processes,
prefixation is the most frequently occurring, since most
verb forms, active or passive, are marked by the addition
of some type of prefix to the root, with or without accom
panying suffixation.
1.8.1.1 The most interesting type of prefixation for
the purposes of this dissertation is that used to form
Simple Active verbal forms. In general, the surface effect
of this process, which will henceforth be referred to as
ng-Preflxatlon, is the replacement of a voiceless stem-initial
consonant with its homorganic (or nearest to homorganic)
nasal and the addition of nasalization to a stem-initial
voiced consonant;
15
stem gloss Active

kel^m become submerged ngeljrfm
pangan food mangan
tjampur mix, mixed njampur
totoq back of the hand notoq
• •
tUlis write nUlis
baj ar pay,salary mbaj ar

dandan bamboo rope ndandan
• •
• •
djogo guard, watch ndj ogo

dUduq place, position ndUduq
• • « •
gOleq get, seek nggOleq

Root-Initial s- Is replaced by nj- :
sljung canine tooth nJIJung
Initial w- Is usually replaced by m-
wotjo read mot Jo
When the nasal prefix Is added to a stem beginning with a
vowel, or 1- or r-, the surface form of the active verb
consists usually of ng- or, optionally, ngjg- plus the root:3
adj ar learn ngadJ ar

irfndok egg ng^ndok
o •
IbU mother nglbU

OmbE drink ngOmbE
llwafc go past, by way of ngllwat

rantjak qu1ck ngrant j ak
16
1.8.1.2 A second very commonly occurring prefix is

the dl- which marks the Simple Passive form of transitive
verbs. This affix appears as dl-, regardless of the shape
of the stem with which it occurs, and, at the same time,
the presence of the prefix does not in any way affect the
pronunciation of the stem:
stem gloss Passive
bakar roast dl-bakar
godok boil dl-godok
•
OmbE drink dl-OmbE

pangan food dl-pangan
tUlis write dl-tUlis
Although the morphological system of Javanese makes
use of fairly large numbers of other prefixes, nearly all
of these behave like dl- in that they do not vary in their
surface forms, nor is their presence accompanied by any
change in the primary stem to which they are affixed.
Because such processes are, from the point of view of the
present study, quite uninteresting and will play no part
in later discussions, I will not bother to list all those
made use of by the language, but will instead refer the
interested reader to the excellent sources listed in the
bibliography, especially Horne 1961.
1.8.2 In general, the infixation processes of Javanese
are also straightforward, consisting mainly of the Insertion
of -1- or -r-, or a vowel followed by 1_ or r, after the
17
first consonant of the stem. There is in fact little doubt

that the use of such infixes is no longer even productive
in Modern Javanese. At any rate, they will play little role
in ensuing discussions, and the interested reader is again
referred to a source such as Horne 1961 for a description of
their role in the language.
There are, however, certain other morphological processes
in the language which may loosely be considered to involve
infixation, to the extent that they involve a change of
some segment internal to the stem, rather than the addition
of an affix to the stem. Certain processes of this type
are clearly productive in the modern language and will be
discussed in detail in later chapters. At this point,
however, they may be generally characterized as involving
the change of one or more of the stem vowels to the char
acteristic marker vowel(s) of the derived category in question.
1.8.2.1 The first such process is one which I will refer
to as Elative Formation, following Uhlenbeck (19^9).
Elatives are intensified forms of primary adjectives, and
their formation may roughly be characterized as a change in
the last vowel of the primary adjective stem to tense I or U,
the choice between these marker vowels being determined by
whether the vowel to be changed is itself front or back. Thus,
for example, one finds the following primary adjectlve-Elative
pairs t
18
primary gloss Elative

abot heavy,, hard abUt
alus refined, smooth alUs
angel hard, difficult angll
IdJO green IdJU
ramE noisy rami
rlndiq slow rlndiq
It should be noted that, In addition to the vowel change,
Elatives are also characterized by rising intonation and a
drawing out of the final syllable.
1.8.2.2 A second Javanese morphological process
involving change of stem vowels is one which I will refer to
as the Habitual-Repetitive Formation, or Hab-Rep, for con
venience. As the name indicates, this process applies
in the formation of derivatives denoting continual repetition
of whatever act or attitude is implied by the primary stem.
Because the formation of Hab-Reps Is characterized not only
by certain vowel changes but also by root doubling (a type of
reduplication in which the entire root is copied (cf. 1.8.4.2
below)), It is an extremely complex and interesting process
which will receive detailed attention later on, In chapter 7.
At this time, however, In an attempt to give the reader
enough familiarity with Hab-Reps to enable him to evaluate
data presented in the following chapters, I have found it
most convenient to describe the process in terms of a number
of ’'instructions” to a hypothetical speaker. By the use of
19
this unorthodox approach, I am of course making no claims

that Javanese speakers haveinternalized anything like my
four instructions:
(1) Double the primary stem to get Stema- Sten^.
(2) If the final vowel of Stema is not a already,
change it to a.
(3) If the penultimate vowel of Stema is a, change
it to 0.
(4) If the final vowel of Stem_a is already *•
a, change
the final vowel of Stem^ to e.
The above instructions are all that are necessary to
understand the superficial relationship between the great
majority of Javanese primary stems and their Hab-Rep deriv
atives, as represented by the following examples:
primary gloss Habitual-Repetitive
adus take a bath Odas-adus
ball return bOla-ball
bUruh wage earner bUrah-bUruh
dEdE sun oneself dEda-dEdE
® 9 • • • •
dOjan like, enjoy dOj an-dOJ en

Eling remember Elang-Eling
IbU mother Iba-IbU
ramE noisy rOma-ramE
rEwang servant rEwang-reweng
skOlah school skOlah-skOleh
20
1.8.3 There are numerous processes of suffixatlon used

in Modern Javanese, and many of these provide interesting
data for the purposes of this dissertation. In general, it
is possible to divide the suffixes of the language into
three main groups on the basis of the behavior of their
initial segments. These groups consist of: (1). suffixes
which are invariably consonant-initial, regardless of the
shape of the stem they occur with; (2) suffixes which are
consonant-initial whenever they occur with a vowel-final
stem, but which do not have this initial consonant after
a consonant-final stem; and (3) suffixes which are invariably
vowel-initial. In the following sections I will discuss
each of these groups In turn, concentrating on describing
the major differences between stems bearing each different
type of suffix and their primary counterparts.
1.8.3.1 Group 1 consists of only two suffixes: the
First and Second Person Possessive pronouns -kU and -mU.
These suffixes may be added to almost any type of stem
without any accompanying change in the stem itself. However,
there are certain Instances in which the suffixed stem and
its primary counterpart do differ, as In the following:
stem gloss Poss. 1 Poss. 2
dOngo prayer dOnga-kU dOnga-mU
mEdJo table mEdja-kU mEdja-mU
rUpo color rUpa-kU rUpa-mU
As is indicated by these examples, when the non-suffixed
21
primary stem ends in -o, the suffixed form of the word

exhibits -a in this position.
1.8.3.2 Group 2 consists of suffixes which may be
termed nvariable" in that their specific realization depends
on the shape of the stem to which they are affixed. In
general, the variable suffixes may be divided into three
separate groups on the basis of the changes occurring in
stems to which they have been added.
1.8.3.2.1 The first group consists of suffixes such as
-(n)E, which characterizes both the Third Person Possessive
and Demonstrative forms, the Passive Imperative suffix -(n)^n,
as well as a number of other suffixes which will play no
role in subsequent discussions. Because the stem changes
associated with all the affixes in this group are identical,
I will, for simplicity, discuss only -(n)E in detail here.
Suffixes like -(n)E, although they themselves exhibit
various forms, in general may be added to stems without any
accompanying change in the stem. There are, however, two
exceptions to this. The first involves stems in final -o,
where the same change of -o to -a which was noted as occurring
with -kU and -mU takes place. Thus, alongside the First and
Second Person Possessives given in 1.8.3.1 above, one finds
dOnga-nE ’his prayer’, mEd Ja-nE ’his table’, and rUpa-nE
'his color’, from dOngo, mEdJo and rUpo, respectively.
The second change accompanying ~(n)E suffixation involves
stems ending In -iC or -uC, where C stands for any permissible
22
stem-final consonant. In such cases, the derived stem

differs from the primary stem by exhibiting tense I or U in
its final syllable:
stem gloss Poss. 3
anggit thought,feeling angglt-E
dUwit money dUwIt-E
• •
klUwung rainbow klUwUng-E

rambut hair of the head rambUt-E
1.8.3.2.2 The second type of variable suffixes presents
some of the more interesting data in the language. These
are the suffixes which are used to form Causatives: -(q)akE,
-(q )no, and -(q )nE, characterizing the Simple Causative,
Causative Imperative, and Causative Subjunctive, respectively.
All of these suffixes can occur with consonant-final roots
without any accompanying change in the root. This can be
seen in the following examples, where I have, for simplicity,
given only the Simple Causative forms. The corresponding
Causative Imperatives and Causative Subjunctives can in all
cases be formed simply by substituting the appropriate affix
for the -(q)akE which appears on the derived forms cited.
It must also be noted that the suffixes in question all occur
In conjunction with the nasal prefix discussed in 1.8.1.1
above (i.e. Causatives may be thought of as being derived by
circumfixatlon). Thus, the following Simple Causative forms
exhibit appropriate nasal initials:
23
stem gloss Causative

adang steam ngadang-akE
apal know by heart ngapal-akE
bakar roast mbakar-akE
dj aluq request ndj aluq-akE
gOleq get, seek nggOleq-akE
godok boil nggodok-akE
• •
lUngguh sit nglUngguh-akE

panas hot manas-akE
tUlis write nUlis-akE
When Causatives are formed from roots which end in
vowels, certain changes take place in the final vowel of
the root:
anggO use, wear nganggo-qakE
dj/*r0 deep ndJ£ro-qakE
gawE Job, task nggawe-qakE
g^ntl change ngg^nte-qakE
klro thought nglra-qakE
lakU walk nglako-qakE
iungo go (away) nglUnga-qakE
sUwE long (time) njUwe-qakE
On the basis of such examples, one can characterise
the relationship between the last vowel in the unsuffixed
primary root and the stem-final vowel of Causatives by the
following diagram:
24
primary root vowel
1.8.3.2.3 Locatives represent the third and final

group of variable suffixes in the language. The same diagram
just given for Causatives can also be used to summarize
the relationship between the last vowel of primary vowel-
final roots and the stem-final vowel of corresponding
Locative forms. In fact, in nearly all respects, the addition
of the Locative suffixes -(n)I, -(n)ono, and -(n)anE, forming
Simple Locatives, Locative Imperatives and Locative Sub
junctives, respectively, is accompanied by the same root
vowel changes as the Causative suffixes just discussed. This
may be seen by comparing the Simple Causatives given above
and the Simple Locative forms of the following roots (again,
Locative Imperatives and Locative Subjunctives may be formed
by substituting the appropriate ending for the ~(n)I of the
cited forms):
stem gloss Locative
apal know byheart ngapal-I
bakar roast mbakar-I
gOleq get, seek nggOleq-I
25
djjrfrO deep ndj^ro-nl

gawE job, task nggawe-nl
g£ntl change ngg^nte-nl
klro thought nglra-nl
lakU walk nglako-nl
Note, however, that there is one way in which the
suffixed stems of Causatives and Locatives differ. The
addition of the Causative suffix -(q)akE to stems in final
-iC or -uC is not accompanied by any change in the last
vowel of the stem; however, when the Locative suffix -(n)I
is added to such stems, the derived form has tense I or 0
as the final vowel of the stem:
stem gl03S Locative

apiq good, nice ngaplq-l
dJ aluq request ndjalUq-I
glUndung a roll ngglUndUng-I
k^pj ur emerge ng^pJUr-I
pllpir edge, border mllplr-l
tUlis write nUUs-I
1.8.3.3 The last type of suffixes to be discussed
are those listed as Group 3 in 1.8.3 above. This group
consists of those suffixes which are invariably vowel-initial,
regardless of the character of the stem to which they are
affixed. There are two suffixes of this type, which will be
discussed individually, beginning with -o, which is used to
form Simple Imperatives.
26
1.8.3.3.1 The suffix -o may be added to vowel-final

stems with no accompanying change In the stem, Including
those In final -o, as may be seen in Imp. nglUngo-o , from
lUngo *go (away)'. (Note that -o occurs in conjunction with
ng-Preflxatlon.) The same may be said for consonant-final
roots, except those in final -1C or -uC. In such cases, as
was seen to occur with various other types of suffixes, the
derived stem exhibits tense I and U:
stem gloss Imperative
apiq good, nice ngaplq-o
dJUpuq go .get ndJUpUq-o
tlliq visit nlllq-o
tUlis write nUUs-o
1.8.3.3.2 The latter changes are also visible in
stems to which the second Group 3 suffix to be discussed,
the substantivizer -an, has beenadded:
stem gloss Substantive
dJUpuq go get dJUpUq-an
Entuq get, obtain EntUq-an
tUlis write tUHs-an
wlwit begin wlwlt-an
In addition, -an suffixation may result in certain
changes with vowel-final stems. As a result of an old
process of contraction which Is probably no longer productive,
suffixed forms sometimes exhibit the vowel changes summarized
above in 1.8.3.2.2:
27

anggO use, wear anggon
ball return balen
OmbE drink Omben
tj£ko come t^kan
tUrU sleep tUron
However, the more productive manner of forming substantives
involves adding the -an suffix directly to the unaltered
stem in all cases except those in which the stem ends in
-o. For stems in final o, the change exemplified by t^ko;
Subst. t^kan above is still productive. With all other
stem-final vowels, forms such as the following occur:
bantu help, assistance bantU-an^
dadl become dadl-an
dJj£rO deep dj grO-an
gawE Job, task gawE-an
IdJO green IdJ O-an
1.8.4 The last morphological process to be discus:
in this chapter is that of reduplication. Reduplication
traditionally involves copying all or part of the primary
stem; and in Javanese morphology, both forms of this process
play a role. Thus, one finds grammatical use being made
in Javanese of: (1) copying part of the primary root or
stem, with or without accompanying vowel changes; and
(2) copying the entire primary root or stem, again, with or
28
without accompanying vowel changes. For clarity I will

distinguish between these two types by referring to the
former simply as "reduplication" and the latter as "doubling".
1.8.4.1 The most frequent use of reduplication is in
a kind of randomizing or generalizing derivation, which is
exemplified by forms like those which follow. Note that
reduplication may occur with or without an accompanying
nasal prefix:
stem gloss reduplicated
mangan eat m^mangan
nggUJU Joke ngg^gUJU
s^pl quiet,lonely nj^nj^pl
tltl careful n^nltl
Note that the reduplication syllable consists of a copy of
the initial consonant of the stem followed by £. In a few
cases, certain derived forms have the original vowel copied
as part of the reduplication syllable, as in sUsUtJI from
sUtJI 'pure", but in all cases, an alternate form with £
exists (e.g. s^sUtJI) also, with the former reserved for
formal or literary style. Notice that when ng-Preflxation
occurs along with reduplication, the nasal is present both
in the stem and the reduplicated syllable, just in case
It has replaced the Initial consonant of the stem, as In
tltl; njgnltl; otherwise, the nasal Is present only word-
initial ly, as In nggUJU; ngg^gUJU.
1.8.4.2 Doubling Is a very versatile and useful
29
morphological process in Modern Javanese. By the use of

doubling, either alone or in conjunction with vowel change
and affixation, speakers may form plurals, reciprocals, and
a great variety of other derivative categories. Because my
concern is mainly with the segmental relationship between
primary and derived forms, and not with the various possible
semantic relationships, I will not describe in detail the
semantic functions of doubling. Rather, I will simply give
a few different examples to illustrate the diverse uses of
the process (cf. also the discussion of Hab-Reps in 1.8.2.2):
stem gloss doubled gloss
bUkU book bUkU-bUkU books
llrih low (sound) llrih-llrih too low
slnaU study slnaU-sinaU-an pretend to study
wong person wong-wong-an dummy
Without a doubt, some of the most interesting data for
an analysis such as that attempted in this dissertation is
produced by forms which are the product of various combin
ations of doubling with suffixation and prefixation. Doubling
can co-occur with nearly any other morphological process, and
in the course of some of these combinations, certain inter
esting things occur which will play a significant role in
later discussions and can profitably be described in this
preliminary chapter.
1.8.2.1 When a word is formed which consists of a
I
30
doubled stem and a Causative or Locative suffix, If the

primary root Is consonant-final, the derived word consists
simply of the doubled stem followed by the appropriate
suffix:
stem gloss Caus. or Loc. doubled
bakar roast mbakar-I mbakar-mbakar-I
baliq turn mbaliq-akE mbaliq-mbaliq-akE
tUlis write nUlis-akE nUlis-nUlis-akE
Note that in these examples, the nasal prefix occurs In both
parts of the doubled stem. This is the most common way of
pronouncing such forms; however, strictly speaking, the
second nasal prefix Is optional if the nasal is added to the
stem (cf. 1.8.1.1 above). On the other hand, the second
nasal Is obligatory if the primary loot begins with a
voiceless consonant or s- or *^-, so that the nasal replaces
the Initial consonant. Thus, mbaliq-mbaliq-akE could
alternatively be pronounced mballq-ballq-akE, but
nUlls-nUlls-akE is the only permissible pronunciation for
this word.
1.8.4.2.2 When a word consisting of a doubled vowel-
final stem and a Causative or Locative suffix Is formed, two
peculiarities may be observed. Any change which takes place
In the final vowel of the second member of the doubled form
also takes place In the final vowel of the first member.
Furthermore, the n of the Locative suffix and the q of the
Causative suffix occur not only as part of the affix, but also
31
appear after the final vowel of the first member of the

doubled form. These peculiarities of doubled Causatives
and Locatives may be seen clearly in the following examples:
stem gloss Caus. and Loc. doubled
ambU odor ngambo-qakE ngamboq-ngambo-qakE
ngambo-nl ngambon-ngambo-nl
.ball return mbale-qakE mbaleq-mbale-qakE
mbale-nl mbalen-mbale-nl
dJ^rO deep ndj£ro-qakE ndj£roq-ndjj£ro-qakE
ndj£ro-nI ndj £ron-ndj j#ro-nI
1.9 As a conclusion to this section on morphology,
I would like to stress again the fact that no attempt has
/
been made here to provide a general description of Javanese

morphology. This introduction is meant to serve only to
acquaint the reader with the various types of derived forms
which will be cited as data in later discussions and, also,
to provide him with an easily accessible summary of the
segmental changes which occur in conjunction with the various
morphological processes of the language.
32
FOOTNOTES
■Ûnless otherwise noted, all glosses for Javanese
words given in this dissertation will be from Horne 1961
or Horne 1972*- Glosses will be basically those given by
Horne, although they may occur here occasionally in a
slightly shortened form. In cases where a given item
has a number of different glosses, the one used here is the
first one cited by Horne.
2
Strictly, speaking, clusters composed of a nasal
followed by its homorganic light consonant (with or
without a following liquid) are actually pronounced so that
the nasal seems to belong more with the preceding vowel,
while the light consonant is pronounced with the following
vowel. However, it is not at all clear whether the nasal
actually closes the penultimate syllable in such a case;
indeed, the vowel realization rules of the language treat
such clusters as if they do not close the preceding syllable.
3
In a few cases, the Active form of a vowel- or
resonant-initial root is formed by prefixation of m-, as
in Act. mlakU, from lakU 'walk*. This type of nasal prefix
ation is not productive in the modern language and has to
be treated as an idiosyncratic fact about Individual roots.
i)
The vowel change from E to e in the first syllable
of the second member of forms like this will be discussed
in chapter 2.
5
În actual pronunciation, a glide is inserted between
the final (tense) vowel of the stem and the first vowel of
the suffix in forms such as these. This is a purely phono
logical process which will be discussed in chapter 3-
33
CHAPTER II
VOWEL PHONOLOGY
2.0 The surface vowel Inventory of Modern Javanese
consists of the following ten distinct sounds:
i u
X u
e ® o
t o
a
As was noted above in 1.3* in citing Javanese forms, I will
refer to these sounds by the following orthographic symbols:
I U
i u
E fS 0
e o
a
These sounds may be characterized as follows:
I i U u E e O o £ a
high + + + +
low
front
back +
Note that two sounds which differ from each other only
with respect to the tenseness feature value are represented
in my orthography by the same letter of the alphabet, with
the tense member of the pair symbolized by an upper case
letter and the lax member by a lower case letter (e.g. I 9i;
E,e; etc.).^"
34
For convenience, and for reasons which will become

clear in the course of following discussions, sounds which
occur in such tense/lax pairs will be referred to as
"alternating vowels", while those which do not (i.e. a and £)
will be termed "non-alternating". In speaking of a given
alternating vowel, when no specific reference to either
member of the tense/lax pair is intended, I will use symbols
consisting of the tense and lax members of the pair separated
by a slash: I/i, E/e, etc. Thus, reference may be made,
for example, to "the front mid vowel E/e".
In this chapter on vowel phonology, I will consider
the distribution of the ten Javanese surface vowel sounds
as they occur in primary as well as derived words of the
language. The ultimate goal of this examination will be
to determine the basic vowel inventory of the language and
the nature of the phonological rules which must be responsible
for converting underlying vowels to their surface forms.
The discussion will be divided into four main parts,
the first dealing with the non-alternating vowels a and
the second dealing with the high vowels 1/1 and U/u, and
the third devoted to consideration of the mid vowels E/e. and
O/o. In the fourth part, an attempt will be made to tie
together observations from the previous three discussions, and
I will deal with the relevant phonological rules as they
interact with each other and with other processes of the
grammar.
35
It is crucial to note that in nearly all of this

chapter attention will be focussed wholly on the predict
ability of vowels occurring in those morphemes which may
serve either as words in their own right or as stems on
which derived forms may be built. Not until Sect. 2.9 will
attention be devoted to the predictability of vowels occurring
in bound morphemes such as the various derivational suffixes
which will figure in the data. When examples are cited, the
pronunciation of suffixes will be indicated, but not until
2.9 will be actual pronunciation of the bound morphemes
themselves be directly relevant to discussion.
2.1 In order to facilitate certain discussions in this
chapter, the various permissible shapes for Javanese bisyllabic
non-derived words are summarized below in Chart I. In this
chart, Initial C stands for any consonant or cluster which
is permissible in that position (cf. 1.5.1 above); Internal
C represents any single consonant or any cluster which does
not close the initial syllable (cf. 1.5.2); and final C
represents any consonant which Is permissible word-finally
(1.5.1).
The horizontal and vertical indices of Chart I refer
to penultimate and final syllable vowels respectively, so that,
for example, the Intersection of the column headed by a with
the horizontal line beginning with E/e contains morphemes
of the shape CaCeC and CaCE.^ Note that, for each vowel
combination, two forms are listed, the first having two open
36
syllables, and the second consisting of an open Initial

syllable followed by a closed final syllable.
It should also be noted that, although all of the
different vowel combinations represented on this chart do
actually occur in independent morphemes of the language,
certain combinations are extremely rare, while others occur
with great frequency. The relative frequency of occurrence,
however, will play no part in this particular discussion,
but will figure prominently in chapter 5 ‘
a I/i U/u E/e O/o
a CoCo CICo CUCo CECo COCo C&Co
CaCaC CICaC CUCaC CECaC COCaC C0CaC
1/1 CaCI CICI CUCI CeCI CoCI CfSCI
CaCIC CICiC CUCiC CECiC COCiC CfiCIC
U/u CaCU CiCU CUCu CeCU CoCU C£CU
CaCuC CICuC CUCuC CECuC COCuC Cj*CuC
E/e CaCE CICE CUCE CECE COCE CdCE
CaCeC CICeC CUCeC CeCeC COCeC C&CeC
O/o CaCO CICO CUCO CECO COCO C£CO
CaCoC CICoC CUCoC CECoC CoCoC CjrfCoC
e *CaC£ *CI/iCs* *CU/uC<2 *CE/eCg acO/oCjg *c<tc&
CaCf*C CICj^C CUC0C CeC&C CoC^C c&cgc
Chart I
2.2 The lax vowels a and £ occupy a somewhat unique
position In the system of Javanese vowels by virtue of the
fact that neither Is a member of a tense/lax pair. That Is,
while all other vowels of the language may be realized in
a tense variant or a (lax) variant differing from the first
only with respect to the feature of tenseness, depending on
37
morphological function, syllable shape or phonological

environment, a and £ have only a single unvarying lax
realization, regardless of context (but cf. 2.3.1.2).
This difference can clearly be seen by examination of
Chart I. Comparing, for example, the column indexed E/e
with that indexed a, one finds that in the former there
appear to be "alternations" between tense E and lax e in
penultimate syllables, as the vowels and shapes of final
syllables vary down the column. Thus, for example, one
finds a morpheme of the shape CeCI with a lax penultimate
e, but also CECiC, with a tense E. Note also that there
is a tense/lax alternation in the final syllables of such
forms. On the other hand, in the a column, no such alter
nation occurs; the penultimate vowel remains a down the
column.
These alternations between the tense and lax variants

of not only E/e. but also O/o, U/u and I/i, may be seen
not only in the various hypothetical primary stems included
on Chart I, but also in the various derived forms of the
language. For this reason, I will refer to E/e, O/o, 1/1, and
U/u as the "alternating vowels" of the language, in order to
distinguish between them, on the one hand, and a and
which do not vary, and which I will call "non-alternating".
2.3 Although the non-alternating vowels a. and do not
have tense variants and can be seen to occur in nearly all
environments given on Chart I— even those in which the tense
38
variants of alternating vowels occur— their distribution Is

not totally free and unrestricted: as can be seen from
Chart I, no surface morphemes exist with a or £ in a final
open syllable (i.e. in absolute word-final position).^
For 0, the story ends here; £_ occurs in all positions

in Javanese morphemes, except word-finally. Examination of
primary and derived forms of the language reveals that there
is no reason to suspect that this distribution does not
reflect the situation at the most basic level.**
2.3.1 The case is quite different for the other
non-alternating vowel, a. Although in most cases there
is no reason to suspect that surface a reflects anything
other than an identical underlying segment, there is evidence
to indicate that in certain Instances, basic /a/ is not
realized as a at the surface level.
2.3 .1.1 It has already been noted that there are no
Javanese morphemes ending in word-final a. In view of the
fact that this vowel may occur in all other positions in
Javanese words, both primary and derived, this gap is quite
striking. This absence of surface word-final a becomes
even more suggestive when one notes that there are examples
of another segment occurring in this environment which would
not normally be expected to occur there, considering the
usual character of word-final segments In the language. '
As may be seen on Chart I, although In nearly all cases a
word-final alternating vowel occurs in its tense alternant
39
(i.e. one finds morphemes of the shape CVCI, CVCU, CVCE,

CVCO), there does exist one exception to this rule: namely,
the lax vowel o, which occurs word-finally.
2.3.1.2 Considering both the striking gap in the
distribution of a and the anomalous occurrence of o word-
finally, one wonders whether the two facts might be related.
This possibility becomes a certainty upon examination of
certain derived forms in which a suffix is added to primary
morphemes ending in -o, thereby causing this segment to
no longer be in word-final position. Consider, for example,
the following list of o-final stems and some of their suffixed
forms:
stem gloss derived
djlwo soul,spirit dJIwa-kU
djIwa-mU
djIwa-nE
IJo yes nglja-nl
k£no can, may ng<£na-qakE
ng^na-nl
xnEdjo table mEdJa-kU
mEdja-mU
mEd ja-nE
2.3.1.3 The alternation between surface word-final o
and non-final a exemplified by the above forms is so regular
throughout the language that there can be no doubt about the
existence of a significant relationship between the ttso
40
segments. In keeping with this, it seems reasonable to

suggest that perhaps a does indeed occur in word-final
position in Javanese morphemes— but only in their underlying
representations. Before such morphemes reach the surface,
a phonological rule could apply to change word-final /a/ to
o. Such a rule might be formulated as
/a/ -* o / _ J
For convenience, I will refer to the analysis employing such
a rule as the "Basic /a/ Approach".
2.3.1.4 It seems appropriate to digress here for a
moment to note that, to one reading actual Javanese written
words, the relationship of word-final o to internal a is
really quite a bit more obvious than it appears in the
phonetic transcription used here. This is because standard
Javanese orthography does not distinguish between the word-
final segment o and its alternant a, representing both sounds
as <a^« Thus, the word transcribed here as mEdJo would be
written in Javanese orthography as ^medj^ , with the Poss.l
^jnedjaku^. However, because this dissertation deals not with
written, but with spoken words, I have chosen to represent
the final vowel sound of Javanese ^ m e d j a s o, this choice
being influenced mainly by the fact that the pronunciation of
the final vowel of mEdJo coincides exactly with the pronun
ciation of the lax alternant of O/o.
2.3.1.5 It is also appropriate to note here that in
Uhlenbeck 1949 the notation a Is used for the ordinary lax a
41
sound, while a Is used for the o pronunciation, as In mEdJo.

This notation parallels 1,1; u,u; 0 ,0 ; and e,e, which
Uhlenbeck uses for the lax and tense variants of the
alternating vowels. There are some ways in which Uhlenbeck’s
notation is superior to that used here: notably, the way
in which it seems to capture the fact that the sounds he
// / / /
cited as i,u»£,o, and a all occur in final open syllables,
> \ \ \
while in final closed syllables, the lax variants i,u,e,o,
and a all occur.
At the same time, however, setting up such a parallel
has drawbacks, since there are other respects in which the
relationship between a and a differs significantly from
that between the members of other tense/lax vowel pairs.
First of all, the (tense) a*variant of a occurs only
word-finally or in a syllable preceding word-final a (2.3.3),
while the occurrence of the tense variants of other vowels
is relatively unrestricted internally. Thus, while a* occurs
✓ Z' /■
internally only before another a, tense i or u , for example,
can occur before any other tense or lax vowel in the language;
/ /
the internal occurrence of e and o, while not quite so free,
is nevertheless not nearly as restricted as that of af. On
the other hand, while the internal occurrence of a is
virtually unrestricted, the lax variants of the other vowels
do occur only in certain environments.
Secondly, the phonetic difference between a and a is
much greater than that between the tense and lax variants of
k2
> / \
othervowels, since thedifference between 1 and 1 or o and
cf, forexample, isreally
only one oftenseness; on the other
\ s
hand, the difference between a and a Is phonetically one of
backness and roundness.
In general, I feel that the orthography I have chosen
Is the best for the purposes of this dissertation; not only
does It avoid the problems inherent in Uhlenbeck’s approach,
but it also captures the overlap in pronunciation between
Uhlenbeck’s tT and o by using a single symbol to represent
both.
2.3.2 Leaving orthography now and returning to
discussion of the possible relationship between a and o
within the grammar of Javanese, recall that the possibility
of taking the Basic /a/ Approach has already been suggested.
At this point, it seems equally possible that the alternation
between internal a and final o could alternatively be the
result of a phonological rule which changes underlying /o/
to a in non-word-final environments which are created by
the addition of a suffix to stems in /a./#. Such a rule might
tentatively be formalized as
/o/ a / +suffix
For convenience, I will refer to this analysis as the "Basic
/o/ Approach" to distinguish it from its counterpart which
was formulated earlier (2.3.1.35.
2.3 .2.1 Already, at this early point in the discussion,
certain things are evident about the difference between
adopting the Basic /a/ Approach versus the Basic /o/ Approach.
Even if one were to posit a basic segment /o/ to account
for alternations like those illustrated in 2.3.1.3 above, it
would still probably be necessary to include in the under
lying vowel inventory a segment /a/ to account for the
numerous other non-alternating surface occurrences of a in
the language. That is, because surface a has a very free
distribution within the morphemes of the language, and thus
can occur in any environment except word-finally, it would
clearly not be feasible to derive all these occurrences
exclusively from basic /o/; the latter course would involve
the undesirable claim that all the diverse occurrences of
surface a are derived, while only the o which occurs strictly
word-finally (as in mEdjo, for example,) reflects the basic
segment. Thus, the Basic /o/ Approach would necessitate
positing /a/ as well as /o/ as a basic segment. On the other
hand, if the Basic /a/ Approach is chosen, the various
non-alternating occurrences of surface a may be thought of
as reflecting the basic segment, while the alternating o
which occurs only in word-final position would be derived.
As a result of taking this approach, it thus appears that
one can avoid positing both /a/ and /o/ and instead derive
surface a and o from basic /a/.
2.3.2.2 The importance of this saving can be seen
when it is noted that there is no need within the rest of
the grammar to posit basic /o/. There is no reason at all
44
to connect the word-final vowel of mEdJo, for example, which

alternates with a, as in mEdja-nE, with any other occurrence
of surface lax o (except in stems like kontjo ’friend', which
will be discussed immediately below) on Chart I, as in forms
of the shape CoCoC, CaCoC, CoCI^, CoCgfC, etc.In forms of
these types, surface o alternates exclusively with tense 0
(and not a); and in all environments, the occurrence of lax
o or tense 0 parallels the occurrence of the tense or lax
variant of the other alternating mid vowel E/e (cf. 2.5.1).
As will be seen later in this chapter, all occurrences of
surface lax o and e which alternate with tense 0 and E are
derivable from basic tense /0/ and /E/.
Therefore, there is no need within the grammar to
consider including a segment /o/ in the basic vowel inventory
except to account for surface occurrences of o which alters
nate with a, as in mEdjo; mEdja-nE. Thus, because, as was
noted earlier, the Basic /o/ Approach does not eliminate the
need for positing basic /a/, while the Basic /a/ Approach
does eliminate the need for basic /o/, it appears that tak
ing the latter saves using an extra underlying segment /0/
which would not otherwise be needed in the grammar at all.
2.3.3 In spite of the fact that the Basic /o/ Approach
necessitates positing an extra basic segment, it should not
necessarily be eliminated entirely from further consideration.
In general, on the basis only of the data considered so far,
it would seem that there is no compelling reason for choosing
4$
either approach; however, consideration of further evidence

reveals crucial inadequacies in one which do not exist in
the other. Consider the following forms which exhibit, in
addition to the o/va alternation already described for
stem-final syllables, an accompanying alternation in the
vowels of non-final syllahles:
stem gloss derived
djogo guard, watch ndjaga-nl
gowo bring nggawa-nl
nggawa-qakE
kontjo friend kantJa-kU
kantJ a-nE
loro ill* painful lara-nE
nglara-qakE
In primary stems like these, where the first syllable as
well as the final syllable contains lax o, the addition
of a suffix results in a surface form which not only has
a in the stem-final syllable, but also has a in the
penultimate syllable of the stem.
2.3.3.1 Prom examination of relevant forms throughout
the language, it is clear that the occurrence of an < w a
alternation in penultimate syllables is totally dependent
on an accompanying o-'va alternation in the final syllable
of the stem. Although it is possible to find numerous
stems which have penultimate o, and also many with penultimate
a, these vowels never show an O'va alternation unless there
46
is an alternating o or a in the final syllable of the stem.

Thus, one finds on the one hand forms such as mEd.1o ; mEd Ja-nE,
etc., where only the stem-final vowel alternates, and on the
other, those like kontjo; kantJa-nE, where both stem syllables
have the alternation; however, there are no cases where only
the penultimate vowel alternates.
2.3.3.2 It seems that, whichever of the two suggested
approaches to the final-syllable o/va alternation is ultimately
found to be preferable— Basic /a/ or Basic /o/— two means
of capturing the dependency of the penultimate OA/a alternation
on the final-syllable o v a alternation are available: !the
relevant rules could be extended to affect penultimate as
well as final-syllable vowels; or a separate rule could be
included in the grammar to match up a penultimate vowel to
the final vowel of the stem.
Because, whichever of these two courses is chosen, the
details of the problem remain basically the same, I will for
now assume that the internal o/va alternation is the result
of a process which is separate from, but in some way dependent
on, the phonological rule responsible for the stem-final
alternation; and I will proceed to consider this process in
terms of the two approaches to the stem-final o/v a alternation
suggested above,
2.3°3°3 Beginning with the Basic /a/ Approach then, one
would assume that a word like kontJo has an underlying shape
/kantja/. The simplest and most straightforward way to
**7
capture the dependency of the internal alternation on the
stem-final alternation would be to order the phonological
rule responsible for producing internal o after the rule
taking final /a/ to o has already applied, and then mentioning
the latter derived segment in the r u l e ’s environment. This
ordering would then account for why the internal vowel of a
stem does not alternate unless the final vowel shares the
alternation.
The input to the rule which accomplishes the internal
vowel change would then be /kantja/.^ This rule would
presumably match the internal /a/ to the final derived segment
and might be something like the following, which I will
tentatively call a-Vowel Harmony:
/a/ o / _ C flo#
2.3.3.** If one takes the alternative tack and assumes

that /o/ is basic in words like kontJo, the underlying
representation of this word would be identical to the surface
form; it would only be in suffixed forms like kantja-kU that
the phonological rules would come into play. In such a case,
the basic representation of kantJ a-kU would presumably be
/kontjo-kU/, with the rule that takes /o/ to a stem-finally
applying, followed again by a vowel harmony rule very similar
to that suggested above, differing mainly in the direction of
the change:
/o/ — 7>a / C a-frsuffis
'■ ni Q
48
2.3*4 Although, at this point, it is still the case
that neither the Basic /a/ Approach nor the Basic /o/ Approach
is clearly preferable over the other, there is additional
data to be considered which makes such a choice possible.
Consider the following roots and their various derived forms:
stem gloss derived
atmo soul, spirit atma-nE
brasto wipe out mbra3ta-nl
dJ almo human being dj alma-nE
sarto and, with njarta-nl
swargo heaven swarga-nE
Notice that in these examples, even though the vowels of

the suffixed forms are identical to those given in 2.3*3
above (i.e. a in both stem syllables), the non-suffixed
forms only show the o<va alternation in the word-final vowel,
while the first syllable still has a.
2.3.4.1 Aside from the difference in vowel alternations,
it seems that the one difference between the examples Just
given and those like kontJo in 2.3.3 is that the former have
an internal consonant combination which closes the first
syllable, while the latter do not. (For discussion of what
consonant combinations close a non-final syllable, see 1.6
above.) This difference and its effect on the vowels of the
words can be seen especially clearly by comparing the roots
and suffixed forms of pairs like the following:
stem gloss derived
dj akso public prosecutor dJaksa-nE
djoko young man dJaka-nE
dJ arwo meaning djarwa-nE

djoro drill dJ ara-nE
karjo work karj a-nE

koro climbing vine kara-nE
warno sort, variety warna-nE

woro say, speak mara-qakE
2.3.2 The above data does not pose any serious
problem for the Basic /a/ Approach. This approach can be
made to successfully account for the alternations In these
examples by simply adjusting the environment of a-Vowel
Harmony to ensure that the rule operates only when the
initial syllable is open. Thus, forms like swargo would
come from underlying /swarga/, with the rule that takes final
/a/ to o applying, but with a-Vowel Harmony failing to apply
because of the syllable-closing cluster -rg-„ Note that this
analysis makes the implicit claim that all vowels in forms
like swargo„ kontJo, and, for example, those like dJawap
'reply*, with closed final syllables where there is no vowel
alternation, are underlyingly /a/; it is the operation of
the phonological rules which accounts for the various surface
realisations of these vowels. At the same time, this analysis
50
predicts that, should one of the internal consonants of

swargo, for example, be lost, thereby opening the internal
syllable, the word would then be pronounced sworo* or
swogo*.^ Similarly, should forms of the shape CoCo or CaCa
gain or lose a final consonant, corresponding alternations
in pronunciation would be expected.
2.3.4.3 Problems arise when one tries to handle data
of this type with the Basic /o/ Approach. Because the first
vowel of a form like swargo never occurs as anything but a,
there is no reason to assume that it is any different from all
other non-alternating a vowels in the language (i.e. it is
Just like the non-alternating a vowels which occur in
morphemes such as dJawap ’reply1, gawat ’serious’, or tabah
'strong'). Thus, taking the Basic /o/ Approach, swargo would
presumably have an underlying form identical to its surface
shape, contrasting with, for example, kontjo at the most basic
level: /swargo/ versus /kontjo/. Suffixed form3 such as
swarga-nE would then be the result of application to the
basic form of the rule which takes stem-final /o/ to a before
a suffix.
In addition to the fact that this analysis necessitates
including basic /o/ in the underlying vowel inventory, while
the Basic /a/ Approach does not, the Basic /o/ Approach also
places an ad hoc restriction on the distribution of vowels
In possible underlying forms of the language. It implicitly
51
makes the claim that speakers must learn that in forms with
underlying stem-final /o/, penultimate /a/ may occur only in
closed syllables, while /o/ occurs in the penultimate position
only if the syllable is open. Besides the ad hoc nature of
this claim, it is in contradiction to the otherwise free
occurrence of /a/ in closed or open penultimate syllables
before final syllables containing all other vowels.
2.3.5 The most clearly convincing evidence against
the Basic /o/ Approach may be seen by consideration of
alternative pronunciations of certain Javanese words. Such
variant pronunciations occur quite frequently in normal"
usage, with only slight differences between the variants, and
with both pronunciations being quite acceptable to native
speakers. One such word is bapaq ’father’, which may either
be pronounced as cited, or, alternatively, bopo. The latter
has two a vowels showing up when a suffix is added, as in
bapa-nE. Note that the presence (bapaq) or absence (bopo)
of a final consonant is apparently what determines the pronun
ciation of the vowels in the two variants.
2.3.5.1 Taking the Basic /a/ Approach allows a straight
forward explanation for the differences in pronunciation of
the vowels in the two variants. It may simply be assumed that
speakers have for ’father’ a single lexical representation,
which might be formalized as something like /bapa(q)/, where
the final consonant is optional. This allows either of two
closely related alternative representations to serve as the
52
input to the phonological rules: /bapaq/ or /bapa/. By

the operation of the rule which realizes word-final /a/ as
o and subsequent application of a-Vowel Harmony to /bapa/,
the correct surface variant bopo is produced, while /bapaq/,
of course, does not undergo either rule and is realized as
bapaq. It is important that, in addition to producing the
correct surface variants, this approach also sets up a
basic relationship between the two variants which seems to
capture the fact that they are felt to be simply different
ways of saying the same word.
2.3.5.2 The Basic /o/ Approach does not allow such
simple accounting for pronunciation differences in dialectal
variants of this sort. Because the vowels in bapaq are in
no way distinguishable from the two a vowels in numerous
other roots of the language with the shape (C)aCaC, there
is no reason to assign bapaq any basic representation other
than /bapaq/. However, in order to account for the pronun
ciation of the variant bopo; Poss.3 bapa-nE, the latter
would presumably have to be assigned the basic shape /bopo/.
Thus the Basic /o/ Approach forces one to claim that speakers
feel the two variant pronunciations of the word to be unre
lated to the extent that each must have a distinct lexical
entry. Compare this to the Basic /a/ Approach which claims
that the variant pronunciations result mainly from the appli
cation of phonological rules to a single lexical item which
may or may not be pronounced with a final consonant.
53
2.3.6 More evidence against the Basic /o/ Approach aris

when one tries to apply it to the analysis of variants like
bakdo 'end, conclusion' and bodo, which, although it Is,
according to my informant, most frequently used to mean 'end of
Moslem fast', can also be used in place of bakdo, and In the
latter case is felt to be merely a variant pronunciation.
While bodo shows alternations of both vowels, as for example
In Dem. bada-nE, bakdo behaves like the examples in 2.3.** and
only shows alternation of the final vowel: bakda-nE.
2.3.6.1 Again, the Basic /a/ Approach can handle these
variant pronunciations with no difficulty by allowing the
existence of a single lexical item /ba(k)da/, which may serve
as Input to the phonological rules in either the form /bakda/
or /bada/. In the first variant, /a/ becomes o word-finally,
but a-Vowel Harmony does not apply because the internal
consonant cluster closes the penultimate syllable and there
fore blocks the application of the rule; In the second,
there is no internal syllable-closing cluster, so both rules
can apply.
2.3.6.2 When one attempts to use the Basic /o/ Approach
to produce the correct surface variants bakdo and bodo, there
arise not only the difficulties that were met In the discussion
of forms like swargo (cf. 2.3.4.3), but. In addition, the same
problem arises that was met in dealing with pronunciation
variants like bapaq and bopo. Not only is It Impossible to
capture the apparent dependence of the c w a alternation in
54
penultimate syllables on the occurrence of the same alternation

in stem-final syllables, but it is also impossible to capture
the intuitive relationship between the variant pronunciations
of bakdo and bodo, since each would presumably require a
separate lexical entry.
2.3.7 Thus, it seems that it may be quite reasonably
concluded that in Javanese the gap in the surface distribution
of a does not reflect a similar gap in possible underlying
representations, but rather is the result of a phonological
rule which applies to underlying word-final occurrences of
/a/ and converts them to surface o. It may also be concluded
that the grammar of the language contains some process to
change underlying penultimate /a/ to o in an open syllable
before a word-final o (from /a/ ) J Although neither of these
processes has been dealt with in much detail so far, they will
be discussed again later in this chapter, where their relation-
ship to each other and to various other processes of the
language will be discussed.
2.4 As was noted earlier, this second part of chapter 2
will be devoted to discussion of the high alternating vowels
1/1 and U/u and their place within the Javanese vowel system.
Just as the non-alternating vowels a and £_ may be seen to
be, in a way, set apart from the other vowel sounds of the
language by virtue of their non-alternating status, the high
alternating vowels I/i and U/u seem also to exhibit certain
distinctive behavior— In this case the difference being the
55
limited frequency of occurrence of their lax alternants. That

is, as can be seen from Chart I (2.1), while both the tense
and lax variants of the mid alternating vowels E/e and O/o
seem to occur fairly often, this is not the case for 1/1 and
U/u; for these vowels, the tense variants I and U occur in
nearly all environments.
2.4.1 In fact, from consideration of the forms containing
1/1 and U/u on Chart I, it becomes evident that, for bisyllabic
primary morphemes, the only environment in which the lax
variants i or u occur is in word-final closed syllables. The
great prevalence of the tense variants I and U, along with the
clear-cut environment for the occurrence of lax i_ and u,
suggests that it may be possible to exclude lax high vowels
from the basic inventory of Javanese and derive them from
basic tense vowels by means of a phonological rule.
The feasibility of such an analysis is maintained when
certain derived forms are brought into consideration:
stem gloss derived
apiq good, nice apIq-E
dJUpuq go get ndjUpUq-o
klUwung rainbow klUwUng-E
tUlis write nUlIs-o
wlwit beginning wlwlt-an
bUrl back, rear bUrl-nE

1
g^ntX change nggjrfntl-o1
56
IbU mother IbU-nE

tUkU buy nUkU-o
Notice that in these examples, while stems with word-final
tense high vowels remain unchanged in the cited derived
forms, the addition of the same suffixes to stems with lax
high vowels in stem-final closed syllables such as aplq
results in derived forms with the tense variant of the high
vowel in question: aplq; Dem. apIq-E .
The existence of such alternations of course supports
the hypothesis that the lax i and u which occur in stem-final
closed syllables are derived from basic /I/ and /U/. That is,
the realization of tense high vowels which occur in stems with
final open syllables (e.g. the I in ggntl) may be thought of
as reflecting the basic shape of the high vowel, and accord
ingly, there is no change even when the relevant suffixes
are added (e.g. Dem. g^ntl-nE). On the other hand, the lax
variants of high vowels which appear in primary stems with
final closed syllables (e.g. the i in aplq) may be thought
of as arising because of a specific environment which exists
only in the primary stem; when a suffix is added, altering
this environment, the vowel is realized in its tense alternant
(e.g. the I in Dem. apIq-E).
2.4.2 From the evidence presented by primary stems and
the suffixed forms given above, it might easily be concluded
that the environment for the operation of a high vowel laxing
rule would be word-final closed syllables, the tense vowel in
57
forms like apIq-E thus occurring because the suffix -(n)E

removes the stem-final syllable from word-final position.
2.4.2.1 The word-final closed syllable environment
for laxing cannot, however, be wholly maintained when certain
other suffixed forms are considered. Note that in the
following examples consisting of stems and their First and
Second Person Possessive derived forms, the lax vowel which
occurs in stem-final closed syllables does not occur in its
tense variant when suffixation removes the stem-final syllable
from word-final position
stem gloss derived
apiq good, nice apiq-kU
apiq-mU
bUrl back, rear bUrl-kU
bUrl-mU
djamur mushroom dj amur-kU
dJ amur-mU
IbU mother IbU-kU
IbU-mU
mUrit student mUrit-kU
mUrit-mU
tanduq actions tanduq-kU
tanduq-mU
2.4.2.2 Comparison of derived forms like apIq-E (2.4.1)
with those like apiq-kU and apiq-mU above reveals that all
suffixes which condition the occurrence of tense I and U in the
58
final syllable of stems ending in a consonant are either

invariant vowel-initial, such as -an, or variable suffixes,
which occur without their initial consonants after consonant-
final stems (cf- 1.8.3.2 and 1.8.3.2.3 above). At the same
time, note that stems with word-final tense high vowels such
as IbU retain their tense vowels, regardless of which suffix
is added: IbU; Poss.3 IbU-nE andPoss.l IbU-kU.
2.4.2.3 It is not difficult to arrive at a plausible
explanation for these facts about high vowels in stem-final
syllables when the various suffixed forms are considered in
terms of Javanese syllabification. As was mentioned in 1.6,
the standard Javanese non-final syllable has the shape (C)V,
while final syllables are (C)V(C). Thus, a word of the shape
CVCV(C) would automatically be pronounced CV/CV(C), with the
internal consonant being pronounced as a syllabic unit with
the following vowel. When a vowel-initial suffix is added
to a stem with a finalclosed syllable, such as aplq,the
final consonant of the stem forms a syllable with the initial
vowel of the suffix, so that the last vowel of the stem is no
longer in a closed syllable. That is, while apiq is pro
nounced in two syllables as a/piq, the derived form with, for
example, the ~(n)E suffix is syllabified as a/pI/q-E, with
the last consonant of the stem forming a syllable with the
vowel of the suffix and the last vowel of the original stem
now occurring in an open syllable. On the other hand, addi
tion of a consonant-initial suffix does not have the effect of '
59
rearranging syllabification, since the last vowel of the

stem remains in a closed syllable (i.e. a/piq and a/pI/q-E,
but a/piq/-kU).
2.4.2.4 Note also that, with stems like IbU which have
an open final syllable, no change ever takes place in the
stem-final vowel, regardless of which suffix is added. It
would seem reasonable that this Is because no suffix can
ever be added which closes the final syllable of such stems
with a consonant; variable suffixes like -(n)E which occur
in their consonant-initial form after stems in final vowels
and those like -kU, which are invariably consonant-initial,
can have no effect on the syllabification of vowel-final
stems: I/bU/-nE, I/bU/-kU.
Prom the above discussion, it seems that whatever rule
is responsible for deriving the lax variants of high vowels
needs to be sensitive to whether the vowel occurs In an open
or closed syllable, and not, as might have been thought from
%
consideration of only the forms in 2.4.1, to whether the

closed syllable In question Is In word-final position.
2.4.2.5 On the basis of evidence so far, a rule to lax
high vowels In the proper environments might thus be formulated
quite simply as:
The reader will have noticed that such a rule, which I will
refer to as High Vowel Laxing or HVL, applying freely
60
in the grammar of a language, would not only produce lax

vowels in the proper stem-final syllables, but would also
have the effect of ruling out the occurrence of tense high
vowels in final syllables ending in more than one consonant,
as well as in closed penultimate syllables.1^ Stems of the
former type never occur, since Javanese absolutely does not
tolerate stem-final consonant clusters; even borrowings are
remodeled to eliminate such clusters. Nor is there any
problem with penultimate closed syllables, since words
containing an alternating vowel in that position are usually
felt to be non-native and thus only undergo the native rules
optionally (cf. 1.7). I do know, however, of at least one
word of this type which my informant considers to be a native
word: gurdl 'drill'. And, as is predicted by the formulation
of High Vowel Laxing, this form is consistently pronounced
with a lax penultimate high vowel.
Thus, for the data considered so far, it appears that
High Vowel Laxing may be accepted as formulated. As it
stands, the presence of this rule in the grammar enables one
to predict in all cases whether a high vowel will surface
in its tense or lax variant. At the same time, it becomes
possible to reduce the underlying vowel system of the language
to the two tense high vowels /I/ and /U/, along with the two
basic lax vowels /a/ and /£/ which were discussed earlier.
Only the mid vowels E/e and O/o remain to be considered.
61
2.5 In contrast to the two non-alternating vowels a

and £, which are always lax, and the high alternating vowels
1/1 and U/u, which are nearly always tense, as can be seen
from Chart I (2.1), the mid alternating vowels E/e and O/o
have a more varied distribution of their tense and lax var
iants. Because of this varied distribution, the linguist is
faced with a much greater task when attempting to determine
the basic shape of the mid vowels; it is by no means obvious
whether the underlying inventory of Javanese contains lax /e/
and /o/ or tense /E/ and /0/, or whether both tense and lax
variants of the mid vowels must be present in underlying
forms. In order to determine which situation actually obtains
in Javanese and what phonological processes may be responsible
for the existing surface distribution of tense and lax vari
ants of the mid vowels, consideration must be given to the
behavior of these segments in a number of environments In
primary and derived forms throughout the language.
2.5.1 Beginning with the primary words given on Chart I,
it can be seen that the lax variants of the mid vowels occur
in three different surface environments: 11 (1) In final
closed syllables (e.g. akeh ’much, many’; d.1Enggot ’beard’);
(2) In Initial syllables before (word-final) tense high
vowels (e.g. merl ’envious, envy’; wolU ’eight’); (3) In Ini
tial syllables before final (closed) syllables containing £
(e.g. bos^n ’tired of’; emp^r ’resemblance’). The reader
should be aware that not all- of these environments occur with
62
equal frequency; while (1) and (3) are very commonly found,
words meeting environment (2) are In fact quite rare, with
probably not more than fifteen or twenty in the entire vocab
ulary. Furthermore, most of these are of the type represented
by klerU 'mistaken1, with the frontness or backness of the
first vowel differing from the value of that feature in the
word-final vowel, while a smaller number are like merl
'envious, envy’, with two front vowels, and even fewer have
two back vowels, like wolU ’eight’.
In addition to these three environments shared by both
E/e and O/o, there is a fourth which, for the time being,
will be stated for the front and back mid vowels individually.
The lax variant of the front mid alternating vowel E/e occurs
in one environment which is not shared by O/o: (4a) In an
initial syllable followed by a closed stem-final syllable

containing (lax) e (e.g. leren 'stop, rest’, compared with
djEnggot 'beard'). Similarly, the lax variant of the back
mid alternating vowel O/o occurs in one environment which is
not shared by E/e: (4b) In an Initial syllable followed by
a closed stem-final syllable with (lax) o (e.g. kodoq 'frog',

compared with Oleh ’get'). In all other environments In
native words, the tense variants of the mid alternating vowels
occur. 12
2.5.2 When one examines not just primary bisyllabics
but also various types of derived words, It becomes evident
that the predictability of these environments is maintained
63
quite regularly throughout the language. Here the reader is

again reminded that, even though derived forms are being
considered, attention is centered on the pronunciation of
stem vowels. The. pronunciation of suffix vowels will be
discussed in 2.9.1.
2.5.2.1 Beginning with Environment (1) (i.e. lax vowels
in word-final closed syllables), one can see at first glance
a striking similarity between this environment for mid vowel
laxness and that discussed earlier (2.4.1) for high vowels.
Despite this initial similarity, there is significant dif
ference in the behavior of mid versus high vowels when derived
forms are brought into the picture. Consider, for example,
the following primary and derived forms:
stem gloss derived
dJEnggot beard dJEnggot-E
dOngeng story dOnggeng-E
IdJ en alone ngldj en-I
katon appear ngaton-I
kodoq frog kodoq-E
Compare the above with stems in îC or -uC:
apiq good, nice apIq-E
dJUpuq go get ndjUpUq-I
Notice that in these examples, while high vowels in stem-final
closed syllables occur in their tense alternants before var
iable suffixes such as -(n)E (Dem. apIq-E, but Poss.l apiq-kU),
stem-final mid vowels are lax, regardless of whether or not
64
such a suffix is present.

2.5.2.2 In Environment (2) (i.e. in initial syllables
before final syllables containing high (tense) vowels), the
laxness of mid vowels is similarly unchanged in derivation,
as long as the conditioning stem-final vowel remains:
stem gloss derived
klerU mistaken klerU-nE
kopl coffee kopI-nE
merl envious, envy merl-nE
However, in derived forms where the stem-final vowel is
morphologically altered, thereby destroying the environment
for mid vowel laxness, this change is accompanied by tensing
of the preceding mid vowel. This can be seen in the
Hab-Rep (1.8.2.2) and/or Causative (1.8.3.2.2) fornt3 of the
above stems:
stem gloss derived
klerU mistaken klEra-klerU
nglEro-qakE
kopl coffee kOpa-kopI
merl envious, envy mEra-merl
2.5.2.3 A similar phenomenon is observable in stems
with lax mid vowels in Environment (3) (i.e. before a final
syllable containing gO. The lax mid vowel occurs in all
derived forms in which the conditioning £ is present; but
in the first member of Hab-Rep forms, where a occurs instead
of the preceding mid vowel is tense:
6 5
stem gloss derived

bos^n tired of bOsan-bos^n
empifr resemblance Empar-emp^r
orjrfk shake Orak-or^k
serifp a spare sErap-ser^p
2.5.2.4 Mid vowels in Environment (4a) (i.e. lax e
before a final syllable containing lax e) and in (4b) (i.e. lax
o before a final syllable containing lax o) behave similarly,
with the mid vowel of the penultimate syllable remaining lax,
as long as the conditioning vowel remains:
stem gloss derived
bobot weight bObat-bobot
entjer (of liquids) thin EntJar-entJer
kodoq frog kOdaq-kodoq
• • •
leren stop, rest lEran-leren

2.5.3 In all of the examples of mid vowels in derived
forms discussed so far, it has been the case that the environ
ments for laxness have been present in the primary stems,
while morphologically conditioned change in this environment
results in a tense mid vowel showing up in derived forms.
There exist other types of examples however, in which the
environments for laxness are not present in primary stems,
but are introduced in derived forms. Consider, for example,
the following primary stems and their Causatives:
stem gloss derived
bOdO stupid mbodo-qakE
66
bOdJO spouse mbodj o-qakE

dEdE sun oneself ndede-qakE
kErE beggar ngere-qakE
In these examples, although the primary stems all have two
identical tense mid vowels, the derived forms have two lax
vowels. This may be attributed to the fact that, in accordance
with the diagram given in 1.8.3.2.2 (and also discussion In
chapter 4), the formation of Causatives Is accompanied by a
change of stem-final E or 0 to e or o, respectively. This
change provides the environments for penultimate lax mid
vowels listed as (4a) and (4b) above.
2.5.3.1 It is also of interest to compare the primary
and derived Causative forms of stems like the following:
stem gloss derived
elU accompany ngElo-qakE
merl envious, envy mere-qakE
Notice that, in accordance with Environment (2), the primary
forms both have lax mid vowels In their penultimate syllables.
In the Causatives, however, there Is some divergence: while
the mid vowel in the stem-penultimate syllable of ngElo-qakE
behaves as seen above In 2.5.2.2 (I.e. It Is tense, In keeping
with the fact that the laxness-condltlonlng stem-final high
vowel Is not present In the Causative), in mere-qakE the
vowel of the stem-penultimate syllable is lax. This diver
gence, while perhaps puzzling at first glance. Is entirely
predictable on the basis of Environment (4a). That is.
67
although the stem-final high vowel which conditions laxness

in merl is not present in the Causative, it has been replaced
by a lax e, which also conditions laxness (Environment (4a)).
2.5.3.2 A similar situation arises in certain Hab-Rep
forms. Recall that, as was outlined in 1.8.2.2, when a stem
has a in its final syllable, this vowel is replaced by e in
the second member of its Hab-Rep forms, as in the following
examples:
stem gloss derived
bUbat horsehair bUbat-bUbet
Edan crazy Edan-eden
Endah fine, beautiful Endah-endeh
glmbal unkempt (of hair) glmbal-glmbel
gOmbal rag, scrap gOmbal-gOmbel
Notice that, as long as the vowel in the first syllable of
the stem is not E, there is no change in the tenseness of this
vowel in the Hab-Rep form; however, if the first syllable of
the stem contains E, as in rEwang for example, the second
member of the Hab-Rep form has the lax variant e in its first
syllable. Again, this divergence is attributable to the fact
that the morphologically introduced vowel creates the environ
ment for laxness described as (4a) above.
2.5.4 The fact that Environments (l)-(4) may be isolated
and used with great reliability to predict the occurrence of
the tense and lax variants of mid alternating vowels in derived
68
as well as primary forms suggests the likelihood that speakers

of the language possess a grammar with similar powers of
prediction; since the surface tenseness or laxness of a mid
vowel is always predictable on the basis of syllable shape
and/or co-occurrence with other stem vowels, there is no reason
to suppose that the basic vowel inventory of the language must
contain four distinct mid vowel sounds (/E/ and /e/, /0/ and
/o/). Rather, it would seem that at the most basic level there
must exist only two mid vowel segments, while the character
istic surface distribution of tense and lax variants would
be accomplished by general rules.
The question necessarily arises as to what these two
memorized segments might be— and, at the same time, what are
the characteristics of the general rules which predict surface
tenseness and laxness. There would seem to be three possible
answers to the first of these questions, with the answer to
the second of course depending on which of these is selected:
the basic shape of the mid vowels may be (1) tense /E/ and
/0/, (2) lax /e/ and /o/, or (3) neither tense nor lax (i.e.
13
with no specifications for the tenseness feature). In the
following pages, each of these three approaches will be con
sidered, and the advantages and disadvantages inherent in
each will be discussed, with the ultimate goal being to
discover which of the possible characterizations of Javanese
mid vowels permits the most natural and Insightful analysis
of mid vowel alternations. To facilitate discussion, I will
69
refer to the three approaches as (1) the Tense Mid Vowel

Approach, (2) the Lax Mid Vowel Approach, and (3) the Unspeci
fied Mid Vowel Approach.
2.5.^.1 Beginning with the Tense Mid Vowel Approach, one
can immediately see certain advantages. First of all, although,
as was noted earlier, the mid vowels exhibit a more varied
pattern of tenseness and laxness than the high vowels, it is
still the case that in the majority of environments (as may be
seen from Chart I), the tense variants of mid vowels do occur.
Thus, if mid vowels are assumed to be basically tense, then
only lax variants, which occur in well-defined environments,
need be derived by the rules of the grammar.
At the same time, adoption of the Tense Mid Vowel Approach
has the effect of grouping the mid vowels with the high vowels,
by virtue of shared underlying tenseness, and thus placing
these sounds in opposition to /a/ and /£/, which are basically
lax. The fact that this opposition would correspond exactly
to that between alternating vowels (E/e, O/o, I/i, and U/u)
versus non-alternating vowels (/a/ and /£/) might be regarded
as a point in favor of the Tense Mid Vowel Approach.
On the other hand, there are also certain disadvantages
inherent in grouping the mid vowels together with the basically
tense high vowels, since this grouping might lead one to
expect the mid vowels and the high vowels to exhibit parallel
behavior in certain other ways. If E/e and O/o are assumed
to derive from tense underlying segments, the grammar of the
7 0
language would presumably have to contain some process to

lax mid vowels in stem-final closed syllables (I.e. in Environ
ment (1) from 2.5.2.1), either by means of an extended version
of High Vowel Laxing (2.4.2.5), or by an independent but
similar rule. However, as was pointed out earlier, in certain
derived forms there is a significant difference in the behavior
of mid vowels in stem-final closed syllables versus that of
high vowels in the same environment. As can be seen from the
examples in 2.4.1, when, for example, a suffix like -(n)E
is added to a stem in final -iC or -uC, the high vowel shows
up in its tense variant (aplq; Dem. apIq-E); however, when the
same type of suffix is added to a stem in final -eC or -oC,
there is no change in the stem-final mid vowel (kodoq;
Dem. kodoq-E).
In 2.4.2.3 it was suggested that a plausible explanation
for the behavior of the high vowels might be that the addition
of a vowel-initial suffix (as opposed to the addition of
consonant-initial -kU or -mU) has the effect of opening up
the final syllable of the stem, thereby removing the closed-
syllable environment necessary for High Vowel Laxing to operate
and allowing the basic tense high vowels to be realized on the
surface. With mid vowels however, the presence of a vowel-
initial suffix appears to have no effect; the mid vowel is
in its lax variant in derived forms as well as In the primary
stem. If the Tense Mid Vowel Approach were to be taken, one
would want to account In some way for this divergent behavior.
71
This would not necessarily be an insurmountable problem;

however, it must be considered in light of the fact that, if
the Lax Mid Vowel Approach were to be taken, no problem would
exist at all. In the first place, if mid vowels were assumed
to be basically lax, there would be no reason to group them
with the basically tense high vowels, or to expect their
behavior to be parallel. In the second place, even if, on the
basis of the fact that both mid and high vowels are lax in
stem-final closed syllables of primary stems, one did expect
their behavior to be parallel in derived forms, the basic
laxness of mid vowels would afford an obvious explanation for
why stem-final vowels do not follow the pattern of high vowels
in derived environments. The high vowels are laxed by HVL
when they occur in primary stem-final closed syllables but
occur in their tense variants when the environment for laxing
is not present; taking the Lax Mid Vowel Approach, mid vowels
would be in their basic forms in stem-final closed syllables
and there would be no reason to expect them to become tense
Just because a suffix had been added.
So far discussion of the Tense Mid Vowel Approach has
been mainly limited to its general effect of grouping the mid
vowels with the basically tense high vowels, and in conjunction
with this, it was noted that such an approach necessitates
some type of rule to lax mid vowels in stem-final closed
syllables. In addition to this, attention must be given to
the rules necessary to produce the lax variants which are
72
seen to occur in other environments.

It would not be unreasonable to simply assume that the
grammar of Javanese includes two straightforward phonological
rules to lax mid vowels in precisely those environments
listed as Environments (2) and (3) in 2.5.1 above: that is,
in penultimate syllables before word-final high tense vowels
(e.g. merl, klerU, kopl) and before stem-final syllables
containing £ (e.g. bos^n, emp^r).
Treatment of the two environments listed as (4a) and (4b)
above can be accomplished in a similarly straightforward
manner. Upon consideration of these two environments, it is
not difficult to see that, although they were stated separately
for front and back mid vowels in 2.5.1, if one were to speak
in terms of phonological processes instead of environments,
the proper distribution of lax variants could be produced by
the operation of a single principle. This principle, which
I will refer to as Mid Vowel Harmony, could be something like
the following:
The tenseness of a mid vowel in the penultimate
syllable of a stem will be identical to the tenseness
of an otherwise identical mid vowel in the final
syllable of the stem.
Notice that this principle, as stated, characterizes
what would appear to be an important fact about the shape
of Javanese surface forms: that two otherwise Identical mid
vowels 'in a stem will always be identical, in. tenseness also.1
73
Mid Vowel Harmony would presumably operate after the process

which laxes mid vowels in stem-final closed syllables had
already applied, so that kodoq, for example, would be derived
by the following steps: /kOdOq/ -•>/kOdoq/ kodoq.*** With
• • •
input forms like /bOdO/, which have final open syllables,

Mid Vowel Harmony would apply vacuously, allowing the two
tense mid vowels to be realized in this shape on the surface.
2.5.^.1 Let us now turn our attention to discussion of
the Lax Mid Vowel Approach. Taking this approach, one can
see that, while claiming that mid vowels are basically tense
had the effect of grouping them together with the high vowels
in opposition to lax /a/ and //^/, the second alternative
groups the mid vowels together with /a/ and /£/, leaving /I/
and /U/ as the only basic tense vowels in the language. Since
the opposite grouping corresponded to the distinction between
alternating versus non-alternating vowels and was thus cited
as a possible advantage to the Tense Mid Vowel Approach, it
follows that the grouping inherent in the Lax Mid Vowel
Approach might be looked upon as a possible disadvantage. To
be sure, I can find no significant general similarity in the
behavior of the mid vowels to that of /a/ and /<£/.
Another point which might be looked upon as a disadvan
tage to the Lax Mid Vowel Approach involves the fact that most
surface occurrences of mid vowels in the language are tense;
taking the position that mid vowels are basically lax neces
sitates the presence of phonological processes to derive all
of these tense surface variants, while only the lax variants
reflect the basic form of the vowels.
This is not, however, an insurmountable problem. The
correct surface distribution of tense and lax mid vowels could
be produced from basic /e/ and /o/ by a set of ruleswhose
effect would be something like the following:
Mid vowels become tense:
(1) Word-finally
(2) In stem-penultimate syllables except
(a) before word-final high vowels
(b) before stem-final syllables with £
(c) before an identical tense mid vowel
in the stem-final syllable^
The above summarizes a set of rules which would appear to be

at least a plausible way to produce the correct surface
variants for mid vowels, although it would ultimately involve
vastly more complicated rules than were necessary in the
Tense Mid Vowel Approach discussed earlier.
One clear advantage to assuming that mid vowels are
underlyingly lax involves words of the shape CeC^C and CoCftC
(i.e. those in which the mid vowel is lax before a following
£). Recall that If tense /E/ and /0/ are assumed to be basic,
then some grammatical process is required to lax these vowels
In the environment in question; but if the mid vowels are
underlyingly /e/ and /o/, they then appear in their unaltered
basic shape before sh?2 so that no such process is necessary.
75
Because £ is normally considered to be the maximally neutral

vowel, it might be expected to have no influence on preceding
mid vowels— Just as it seems to have no effect on /I/ and
/U/, as well as /a/ and /£/ (cf. dln^s 'employed’; gUr^m
'chicken flea'; bang^t 'very (much)'; lgriSp 'subside'). And
this is exactly what appears to happen if one accepts /e/ and
/o/ as basic; if this approach is taken, then all vowels in
the language appear in their unaltered basic shape before £.
In my opinion, this advantage is more than neutralized
by the existence of a disadvantage to the Lax Mid Vowel
Approach involving its essentially negative treatment of
what seems to be a very Important principle within the grammar
of Javanese. Recall that in the discussion of the Tense Mid
Vowel Approach it was noted that Environments (4a) and (4b)
for laxing could be collapsed into a single principle of
Mid Vowel Harmony. The effect of this principle was to ensure
that in any stem containing two like mid vowels, these vowels
would match in tenseness on the surface. In the Lax Mid
Vowel Approach, no statement such as Mid Vowel Harmony is
necessitated. Although the correct surface forms may be
attained through a set of rules like that suggested above, the
important generalization that two occurrences of the same mid
vowel in a given stem will agree In tenseness is never
explicitly captured, but, rather, is only negatively implied
by the fact that penultimate mid vowels are tensed except
when they are in Environments (4a) and (4b).
7 6
The disadvantage of this negative approach can be seen

quite clearly by consideration of certain forms on Chart I
(2.1). Concentrating for the moment on the forms with closed
final syllables, note that in only two cases does one find
lax mid vowels in the penultimate syllables of stems: in
forms of the shape CeCeC and CoCoC and in those like CeC£C
and CoC&C. With the Lax Mid Vowel Approach, these particular
configurations share the distinction of being the only ones
in which the vowel of the initial syllable maintains its
original shape, with all other occurrences of mid vowels
being derived. Thus, there is an implicit grouping together
of two types of forms in which surface lax vowels occur in
entirely different environments; it is merely coincidence
that one of these cases involves an environment where two
identical mid vowels co-occur in the same stem.
Turning attention to words on Chart I with final open
syllables, one finds that penultimate mid vowels are tense in
all environments except those in which the final syllable
has a high vowel. All of these tense vowels would probably
be produced by a general rule to tense all penultimate mid
vowels except those which occur before a (tense) high vowel.
Mote, however, that such an analysis implicitly claims that
the tense 0, for example, which appears in the first syllable
of a word of the shape COCO has been tensed for the same
reason (i.e. by the same phonological process) as the 0 which
occurs in words of the shape COCE and COCo, as well as those
7 7
like COCaC, COCeC, etc., and not in order to harmonize with

the tenseness of the Identical stem-final mid vowels.
Thus, in forms with word-final closed syllables, as
well as those with final open syllables, the Lax Mid Vowel
Approach treats as coincidence the fact that two otherwise
identical mid vowels in a stem are always identical in tense
ness; it contains no single statement which can in any way
be felt to capture this generalization about the behavior
of identical mid vowels in the same stem. In derived words,
the disadvantage of this can be seen quite clearly by comparing
the two Causatives ngelo-qakE and mere-qakE, from melU and
merl, respectively. Taking the Lax Mid Vowel Approach, the
lax mid vowels in melU, merl, and mere-qakE must all be con
sidered to be on a par insofar as they are all unaltered
surface representatives of basic lax mid vowels, all theoret
ically reflecting segments memorized as idiosyncratic facts
about each lexical item. At the same time, this approach
forces one to regard the tense front mid vowel of ngElo-qakE
as derived, because it is in a open stem-penultimate syllable
followed by a back mid vowel.
Compare the above to the consequences of a Tense Mid
Vowel Approach. Here the grammar must include something like
the principle of Mid Vowel Harmony, thereby distinguishing
between, for example, the lax mid vowels occurring in words
of the shape CeCeC and CoCoC and the lax mid vowels in those
of the form CeCj[C or CoC0C, with the former and the latter
78
both being derived, but by two separate processes. Similarly,

this analysis distinguishes between the tense and lax vowels
in melU and merl and their Causatives ngElo-qakE and mere-qakE,
with the lax e of both primary forms being the result of a
phonological process which laxes mid vowels before I and U,
but with the lax e in the first syllable of mere-qake being
derived by Mid Vowel Harmony. At the same time, the tense
E in ngElo-qakE is still in its unaltered basic shape.
2.5.^.3 The one alternative approach to the analysis
of mid vowels which remains to be discussed is that which
assumes them to be unspecified for tenseness at the most
basic level, with, presumably, the rules of the language
filling in the appropriate tenseness specification for each
environment. Clearly this approach is more abstract than
either of those already discussed, in that it treats every
surface mid vowel as being derived from a basic memorized
segment which is originally lacking a feature necessary for
actual pronunciation.
Along with this added abstractness, the Unspecified Mid
Vowel Approach differs from those already discussed in that
it sets up the mid vowels as being unique at the most basic
level from all other vowels in the language. Assuming the
mid vowels to be basically unspecified for tenseness divides
the vowel inventory of Javanese into three parts on the basis
of underlying tenseness: basically tense /I/ and /U/i
basically lax /a/ and / W ? and the unspecified mid vowels.
79
Because the behavior of the mid vowels is in many respects

different from that of the high vowels on the one hand and
a and £ on the other, this grouping may be looked upon as
being to some extent advantageous. At the same time, however,
this advantage is perhaps neutralized by the fact that there
would seem to be no way to establish any natural direct
connection between the unique behavior of the mid vowels
within the grammar and an underlying representation which
treats them as being unspecified (as opposed to some other
analysis of their basic shape).
Going on to consider the sorts of phonological rules
which would be necessitated by the Unspecified Mid Vowel
Approach, one finds that these rules are more numerous and
more complex than those used by either the Lax or the Tense
Mid Vowel Approaches, since every occurrence of a mid vowel
must be derived. At the same time, the rules necessitated
by this approach allow one to capture some important gener
alizations about the language which are missed by at least
one of the other two analyses.
First of all, taking the Unspecified Mid Vowel Approach
necessitates a rule to provide mid vowels in stem-final
closed syllables with the feature value J}-tensej in order to
produce the correct lax surface variants. Note that, as in
the Tense Mid Vowel Approach, the presence of such a rule
in the grammar, along with High Vowel Laxing, captures an
important generalization about the shape of Javanese words:
8 0
surface alternating vowels in word-final closed syllables

are always lax. Secondly, this approach would necessitate
a rule to provide absolute stem-final mid vowels with a
featureJ+tenseJ, since, as was noted earlier in 2.3.4, all
stem-final vowels in the language are tense (except for o#,
which is derived from /a/). In addition to the above rules,
others are necessary to produce lax mid vowels in the pronun
ciation of words such as merl and kopl, empgr and or<?k; that
is, in the two environments listed as (2) and (3) in 2.5.1.
Some type of mid vowel harmony principle must also be
included in the grammar to account for alternations like those
in bOdO; Caus. mbodo-qakE and merl; Caus. mere-qakE. Recall
• •
that such a process was discussed in conjunction with the

Tense Mid Vowel Approach. An identical statement of Mid
Vowel Harmony (cf. 2.5.4.1) may be made using the Unspecified
Mid Vowel Approach. The only difference would be in the actual
application of the principle within the grammar. That is,
using the Tense Mid Vowel Approach, Mid Vowel Harmony would
actually apply only to produce a lax penultimate mid vowel
before an identical (lax) mid vowel in the final syllable of
a stem; the rule would apply vacuously whenever the final-
syllable mid vowel was tense, producing surface forms with
two tense mid vowels, like bOdO, for example. Taking the
o
Unspecified Mid Vowel Approach, Mid Vowel Harmony would

apply anytime two Identical mid vowels occur in a given stem,
either assigning a feature£+tenseJor L-tenseJto the penultimate
8 1
mid vowel, depending on the specification for tenseness which

had already been assigned to the final-syllable vowel (cf. the
rules in the preceding paragraph).
Finally, taking the Unspecified Mid Vowel Approach, some
way would be needed to produce tense variants of penultimate
mid vowels in all environments not covered by the preceding
rules: namely, for both mid vowels, before final syllables
which end in -aC, -o, -uC, and -iC; for the front mid vowel
E/e before final-syllable -oC or -0; and for the back mid
vowel O/o before -eC or -E. Clearly this diverse set of
environments would require the presence in the grammar of
either an extraordinarily complex rule or, alternatively,
a large number of less complex tensing rules. Whichever of
these alternatives is chosen, however, the result would be
a derivational apparatus which is significantly more complex
than that necessitated by the Tense Mid Vowel Approach, where,
it will be recalled, only four straightforward laxing rules
were needed.
2.5.5 After an examination of all the advantages and
disadvantages inherent in the various approaches to the
analysis of Javanese mid vowels discussed in the preceding
pages, it seems clear that at least one, the Lax Mid Vowel
Approach, has disadvantages which far outweigh any positive
points which might be made in its favor.
The Tense and Unspecified Mid Vowel Approaches are in
8 2
my opinion nearly equally good condidates, considering only

the degree to which each can capture what appear to be the
important generalizations about the realization of mid vowels
in the surface forms of the language. But in addition to
this, one must also take into consideration not only the
added abstractness of the Unspecified Mid Vowel Approach, but
also the fact that this approach seems to accomplish no more
than the Tense Mid Vowel Approach while requiring a much more
complex derivational apparatus. In view of all these factors,
the Tense Mid Vowel Approach is the one which will be followed
in the remainder of this dissertation.
It is possible that some readers, for one reason or
another, will not agree with this choice.' It should be •
pointed out for the benefit of such readers that, while In
the following discussions much will depend crucially on
acceptance of an analysis which does not posit basic lax mid
vowels, little will depend on acceptance of the Tense Mid
Vowel Approach over that which treats them as being basically
unspecified for tenseness.
2.6 The decision having been made In favor of an analy
sis which treats mid vowels as basically tense, the basic
Inventory of Javanese vowels may be represented as the
following six segments:
/I/ /U/
/E / /W /0 /
/a/
83
Although the rules which are necessary to take these

basic sounds to their correct surface realizations in various
environments have been briefly mentioned in previous sectionsx.
much remains to be said, both about the character and form
ulation of specific rules and about how certain of these rules
relate to other processes within the grammar. Because the
rules dealing with mid vowels are the most numerous and
complex, because these rules have been treated only in very
general terms so far, and, finally, because these rules can
serve as a vehicle for describing various interactions taking
place among the phonological and morphological processes of
the language, the following discussion will be organized
around those phonological processes needed to produce the
surface lax variants of the mid vowels in Environments (l)-(4),
isolated in 2.5.1 above.
2.6.1 Beginning first of all with discussion of the
realization of mid vowels in the final syllables of primary
forms, one can easily see from Chart I (2.1) and from comparing
examples like gs^dE ’large' versus k^teq 'monkey' and bOdO
• • •
'stupid' versus bodol 'come out (hair)* that, Just like the
high alternating vowels, mid vowels are tense in final open
syllables and lax in final closed syllables. With the high
vowels, however, it was noted in 2.^.1 that,- even after-the
addition of various derivational suffixes, this distribution
still holds true; the presence of a variable suffix in its
vowel-initial form opens the final syllable of the stem.
84
resulting in a tense stem-final high vowel (e.g. apiq;

Dem. apIq-E) while the addition of a consonant-initial suffix
causes the stem-final syllable to remain closed, so that it
contains a lax surface high vowel (e.g. apiq; Poss.l aplq-kU).
The rule of High Vowel Laxing (HVL) was accordingly formulated
to be sensitive to the open or closed nature of the stem-final
syllable in all forms.
2.6.1.1 With mid vowels in stem-finallsyllables, how
ever, the generalization that tense variants occur in open
syllables and lax variants occur in closed syllables does not
always stand up when derived forms are brought into the pic
ture. Thus, even though in a form like k^teq-E the stem-
final syllable is, strictly speaking, open, the lax variant e
occurs in the final syllable of the stem, Just as it does in
primary k^teq or derived k^teq-kU, where the final syllable is
• •
obviously closed. In fact, what seems to happen is that in

all derived forms built on stems containing mid vowels in
their final syllables, the occurrence of the tense or lax var
iant of the mid vowel is predictable, not on the basis of
whether it actually occurs in a closed or open syllable within
the word, but rather, on the basis of the configuration of the
final syllable as it occurs in the non-affixed stem. That is,
given a stem such as dOngeng ’story’, with a final closed syl
lable containing a lax mid vowel, one can predict that derived
forms built on this stem will have an identical lax mid vowel,
regardless of whether the stem-final syllable Is actually
85
closed (e.g. dOngeng-kU) or open (e.g. dOngeng-E) In that

particular derived form.
2.6.1.2 Thus, if one were to characterize the difference
between the occurrence of tense and lax variants of high vowels
in derived forms versus that of mid vowels, it might be said
that, whereas the former is clearly governed by the shape of
the stem-final syllable within the word, the latter appears
to be governed solely by the shape of the final syllable of
the non-suffixed stem, regardless of whether that syllable is
actually open or closed within a given derived word. The
most straightforward way to incorporate this restriction into
the grammar would be to include a rule of Mid Vowel Laxing,
formulated as follows:
Mid Vowel Laxing 1 V £-tensej / CT
mid ^stern
(MVL 1)
2.6.2 As it turns out, a similarly restricted environment
(i.e. one specifying that the crucial environment must occur
within a stem) seems to be necessary to account for the surface
realization of mid vowels in the second environment for mid
vowel laxing: namely, in words built on stems containing mid
vowels in their penultimate syllables and high vowels in their
final syllables. In primary stems of this particular vowel
configuration, the mid vowel will be lax just in case the stem
ends in an open syllable (i.e. has a surface tense high vowel)
and will be tense if the stem ends in a closed syllable (i.e.
has a surface lax high vowel). Compare the penultimate mid
8 6
vowels of the following pairs of'stems:

stem gloss
klerU mistaken
Edum shady, sheltered
kopl coffee
tOmis rice-accompanying dish
merl envious, envy

bEring unbalanced
wolU eight
kOdJur bad luck
2.6.2.1 The most obvious approach to forms like these
would be to assume that the grammar of Javanese contains a
rule or rules to lax a mid vowel in a stem-penultimate
syllable before a final syllable containing a tense high
vowel, while the same vowel would remain tense Just in case
the following high vowel is lax. It takes only a moment,
however, to realize that at the input to the phonological rules
all high vowels are tense, so that there would be no way to
distinguish between a form like /klErU/, where the mid vowel
will become lax, and one like /EdUm/, where the mid vowel
must remain tense, until after High Vowel Laxing had applied.
The prior application of this rule would lax a high vowel in
a stem-final closed syllable, leaving one In an open syllable
unchanged. Thus, at the point where the mid vowel laxing rules
87
would apply, two distinct environments would exist, as for

example In Intermediate /klErU/ and /Edum/.
2.6.2.2 Although this analysis is adequate for primary
stems, it encounters serious problems when certain derived
forms are considered: namely, those like the following, with
a variable suffix such as -(n)E:
stem gloss derived
klerU mistaken klerU-nE
Edum shady, sheltered EdU.m-E
• l •
kopl coffee kopI-nE

tOmis rice-accompanying tOmls-E
dish
merl envious, envy merl-nE
bEring unbalanced bErlng-E
wolU eight wolU-nE

kOdJur bad luck kOdjUr-E
Notice that in the surface forms of suffixed words like these,
even though the high vowel of the stem-final syllable is tense
in words like EdUm-E . as well as in klerU-nE, the distinction
between the tense and lax variants of stem-penultimate mid
vowels is still maintained. That is, klerU-nE, like primary
klerU, has a lax mid vowel, while EdUm-E has a tense mid vowel,
9
just like Edum.

Even if High Vowel Laxing as formulated is ordered to
apply before the mid vowel laxing rules, with derived forms
8 8
the input to the latter will still be intermediate forms like

/klErU-nE/ and /EdUm-E/, with tense high vowels in both forms.
Thus, there would be, at that point, no basis for laxing the
mid vowel of klerU-nE but allowing that of Edum-E to remain
tense.
2.6.2.3 It appears that there are two possible solutions
to the above problems. First of all, one might adjust the
formulation of HVL so that all high vowels in stem-final
closed syllables are laxed, regardless of whether or not a
suffix follows. If this were done, the output of such a rule
and the input to the mid vowel laxing rules would be forms
such as /klErU/ and /klErU-nE/, but /Edum/ and /Edum-E/,
• •
with the necessary distinctions between the high vowels in

the various forms to allow the correct assignment of tenseness
features to mid vowels. After the latter had been achieved,
a rule could be included in the grammar to re-tense the high
vowels just in case they occur in open syllables: that is, in
suffixed forms like EdUm-E .
Clearly there are certain drawbacks to such an analysis,
in spite of the fact that it would produce the correct surface
distribution of tense and lax alternating vowels. First of
all, the ploy of laxing high vowels which occur In open
syllables before suffixes, only to re-tense them later, once
mid vowel laxing has occurred, Is somewhat artificial. This
approach really fails, however, when one realizes that the
Inclusion of such a rule In the grammar Is in fact motivated
89
only by the necessity of providing a distinction between

stems like klerU and Edum so that the laxing of penultimate
mid vowels can be correctly accomplished. In doing so, it
Introduces an intermediate level into the derivation at which
lax high vowels occur in an open syllable environment— a sit
uation which never arises on the surface, where in all respects
the occurrence of the tense or lax alternant of the high vowels
is transparently dictated by syllable shape.
The second alternative analysis avoids this problem by
formulating the laxing rule itself in such a way that no
distinction between tense or lax high vowels is necessary to
its proper operation. Taking this tack, a feature value of
^-tense} might be assigned to mid vowels by a rule like the
following:
Mid Vowel Laxing 2 V F-tensel/ C V
mid highl
(MVL 2) stem
This rule would lax mid vowels in precisely those environ
ments where the primary stem is one which has a high vowel in
a final open syllable and would allow them to be realized in
their tense variants just in case the stem-final syllable in
which the high vowel occurs is closed within the stem Itself,
but not necessarily within the word. This has the desirable
effect of putting the burden of the problems involved in
producing lax mid vowels in the proper environments on the
formulation of the mid vowel laxing rule itself. It does not
necessitate the manipulation of high vowels just In order to
90
produce distinct environments in which mid vowel tenseness

assignment can operate correctly. Note, Incidently, that
such an analysis, using MVL 2 as formulated, does not Impose
any special ordering on this rule and High Vowel Laxing.
Looking at MVL 2 for the moment in terms of what it
actually does, one can see that Its primary characteristic
is that it laxes a penultimate mid vowel on the basis of the
shape of the stem-final syllable containing a high vowel and
not on the tesneness or laxness of that vowel. To be sure,
the shape of the stem-final syllable does determine the
tenseness or laxness of the high vowel It contains, but the
character of the high vowel itself is not really the focus
of the environment for the rule In question. Rather, it is
the fact that a given stem must contain a high vowel In its
final open syllable, regardless of whether that syllable is
in actual word-final position, or whether it is followed by
a suffix.
2.6.2.4 Clearly this situation has a great deal in
common with that which obtains in Mid Vowel Laxing 1, which
was discussed earlier in 2.6.1.2. In both cases, the rules
must be formulated so as to operate In an environment which
is defined in terms of the stem, and in both cases the inclu
sion of this environment may be interpreted as a guarantee that
the tenseness assigned to a mid vowel in a given stem will not
vary, regardless of whether the input is a word consisting
of the non-affixed primary stem or a word consisting of the
91
stem plus a suffix.

2.6.3 In addition to the two mid vowel laxing rules
formulated so far, another such rule with a simple environ
ment may be posited. This rule, which accounts for the
occurrence of the lax stem-penultimate mid vowel in forms
like bos^n 'tired o f and empgr 'resemblance* (cf. 2.5.1),
laxes a mid vowel before a stem-final syllable containing
Mid Vowel Laxing 3 V -> f-tense]/ C &C
mid °
(MVL 3)
Notice that all of the processes dealing with mid vowels which
have been posited so far do not have to be applied in any
particular order with respect to each other. However, as will
be seen in the discussion which follows, MVL 1 must apply
before the next rule to be discussed, Mid Vowel Harmony.
The necessity for this ordering is quite obvious when
one considers the nature of Mid Vowel Harmony. Recall that,
by means of this process, which was discussed briefly in
2.5.^.1, a tenseness value is assigned to a penultimate mid
vowel on the basis of the tenseness or laxness of an identical
mid vowel occurring in the stem-final syllable. Clearly the
operation of such a rule must not take place until after the
stem-final vowel has already received its specification for
tenseness by the application or non-application or MVL 1;
otherwise, either all penultimate mid vowels in the environment
for Mid Vowel Harmony would turn out to be tense, or the rule
of Mid Vowel Harmony itself would have to be given a very
92
complex formulation which would Include the environment for

MVL 1.
2.6.4 The operation of Mid Vowel Harmony can be seen in
many primary stems throughout the language, as well as in
certain Hab-Rep, Causative, and Locative derived forms. In
the former, its effect is to guarantee that in all cases where
a primary stem contains two identical mid vowels, they will
both be indentically tense or lax, depending on the tenseness
or laxness of the final-syllable vowel (which is determined
by the shape of the syllable). In Hab-Reps, Causatives, and
Locatives, the effect of Mid Vowel Harmony is the same; how
ever, in these derived environments, the penultimate vowel of
the stem is matched to a final vowel which has been morpho
logically introduced (cf. 1.8.2.2, 1.8.3*2.2, and 1.8.3.2.3):
stem gloss derived
bOdO stupid bOda-bOdO
mbodo-qakE
mbodo-nl
bodol come out (hair) bOdal-bodol
mbodol-akE
mbodol-I
ndede-qakE
ndede-nl
dedel rip dEdal-dedel
ndedel-akE
ndedel-I
93
Compare also:
stem gloss derived
merl envious, envy mere-qakE
2.6.4.1 In examples like these, the tenseness matching
function of Mid Vowel Harmony and Its relationship to MVL 1
are quite transparent. Before a final formulation can be
given to the rule, however, certain other forms must be
brought into consideration. In the following forms the
operation Cor, more precisely, the non-operation) of Mid Vowel
Harmony is not quite so transparent:
stem gloss derived
gOdo temptation gOda-nE
• ♦
pOlo design, pattern pOla-nE

rOso strong rOsa-nE
sOga bark of the sOga-nE
indigo tree
Notice that forms like gOda, rOsa, etc. represent surface
violations of Mid Vowel Harmony as It has been described so
far, since they contain two identical back mid vowels, one of
which Is tense and the other lax.
2.6.4.1.1 It will be recalled from 2.3.7 that the
word-final o in such forms is derived from basic /a/ by means
of the rule formulated in 2.3.1.3a which I will for convenience
refer to as "/a/ to o$”, so that g0do3 for example, comes from
94
basic /gOda/. Knowing this, one might simply assume that

the surface pronunciation of the word results from the
ordering of Mid Vowel Harmony before /a/ to off within the
grammar. Thus, at the point at which Mid Vowel Harmony would
apply, the word would still have its basic shape, with one
/0/ and one /a/, so that the environment for tenseness harmo
nizing would of course not be met. The rule /a/ to o# would
apply subsequently, producing the correct surface form.
2.6.4.1.2 Although it is possible that this ordering
does obtain in the grammar of Javanese, there is fairly con
vincing evidence to the contrary. There are a number of
things about the behavior of /a/ to off which indicate that it
may be a very early rule of the grammar, although most of
these are not directly relevant to the present discussion.
At the same time, it is clear that Mid Vowel Harmony must
apply at an intermediate point, since it must be preceded at
least by MVL 1, which determines the tenseness of mid vowels
in stem-final syllables. More specifically, however, there
is evidence to show that the morphological process of Hab-Rep
Formation must apply after /a/ to o# but before Mid Vowel
Harmony; thus, /a/ to o# must precede Mid Vowel Harmony.
2.6.4.1.2.1 As was discussed in 1.8.2.2, Hab-Rep Forma
tion involves root-doubling and an accompanying vowel change.
Because the process itself is quite complex, in order to
facilitate the present discussion I will speak in terms of
the end results of certain facets of Hab-Rep Formation: that
95
Is, in terms of what a properly derived surface Hab-Rep form

should sound like in relation to the surface form of the stem
it is built on.
One of the first requirements for a good Hab-Rep form
Is that thesecond syllable of the first member of the doubled
pair must be an a. That is, given a primary surface form of
the shape CV^CV^C or CV where is not a, the Hab-Rep
form will be CVâC-CV^CV^C or CV^Ca-CV^CV^, respectively.
Thus, a change of V2 to surface a is part of Hab-Rep For
mation. Furthermore, in case V, is an a, it becomes 0 in the
1 - -
first member of the doubled pair, so that the Hab-Rep for

adoh 'far* is Odah-adoh.
In cases where is already pronounced a in the primary
stem, the Hab-Rep form exhibits a vowel change in the second
member of the doubled pair. In such a case, the second vowel
of the second member of the doubled pair becomes e, so that,
for example, because a word like gOmbal ‘rag1 already has an
a in the position, its Hab-Rep is gOmbal-gOmbel.
2.6.5.1.2.2 Now, returning to the evidence that /a/ to
of must be ordered before Hab-Rep Formation, while Mid Vowel
Harmony must be ordered after,consider the following:
stem gloss derived
djlwo soul, spirit djlwa-djlwo
dOngo prayer dOnga-dOngo
mEdjo table mEdja-mEdjo
All of these stems have basic forms ending In /a/. As will
be argued in chapter 7, the fact that, in their Hab-Rep forms,
the Vg position in the first member of the doubled pair is
filled by a, along with the fact that no e has been Introduced
into position in the second member, can best be accounted
for by assuming that, at the time when Hab-Rep Formation took
place, V2 was not a— otherwise, the Hab-Rep of mEd Jo, for
example, would have ended up as something like *mEdJa-mEdJE.
This implies that, at the time Hab-Rep Formation took place,
the basic form of mEdJo had already been converted from /mEdJa/
to /mEdJo/ by /a/ to o# t so that a change of V2 from o to a
was possible and the introduction of e into the position
of the second member was unnecessary.
Consider, in addition to the above, the following?
stem gloss derived
Endah fine,beautiful Endah-endeh
In these examples, because the primary stem has a in the
position, e has been introduced into the V2 position of the
second member of the doubled pair In the course of Hab-Rep
Formation. The introduction of this (lax) e has In turn
caused the preceding mid vowel of the stem to become lax,
apparently as a result of the normal process of Mid Vowel
Harmony. Thus, the latter may reasonably be thought of as
having applied after Hab-Rep Formation has Introduced the
e which creates the necessary environment for its application.
97
This completes the evidence necessary to establish that the

rule of /a/ to o# must precede Hab-Rep Formation, while Mid
Vowel Harmony must follow it. It follows reasonably from
this that /a/ to o# must precede Mid Vowel Harmony.
2.6.4.2 Let us return now to the discussion of the
interaction of /a/ to o# with Mid Vowel Harmony which was
begun in 2.6.4.1. It had been suggested that one way to
account for the apparent violation of Mid Vowel Harmony In
words like rOso, pOlo, etc., would be to assume that Mid
Vowel Harmony applies before /a/ becomes o. But, as has
just been shown, the evidence presented by Hab-Rep forms
Indicates that just the opposite ordering obtains.
2.6.4.2.1 There is available an alternative approach
which does not Involve ordering. It Is clear from comparison
of examples like bOdO, from /bOdO/, versus rOso, from /rOsa/,
• o
that Mid Vowel Harmony operates only when a given stem con
tains two mid vowels which have an identical basic source
(i.e. /0/) and does not operate when a word contains one
basic /0/ and one o which has been derived from /a/. Thus,
one might claim that Mid Vowel Harmony, given an Input like,
for example, Intermediate /rOso/ from basic /rOsa/, has the
power to look back to the most basic form of the word, to see
that the stem-final o was not originally a mid vowel.
Assuming that Mid Vowel Harmony has the power to choose
between a basic and a derived segment (or. In other words,
that it is a global rule) not only avoids the problems which
98
are encountered in the ordering solution, but also provides
some explanation for the existence of forms like rOso, etc.
That is, while the originally proposed ordering of /a/ to o#
merely produces the correct surface forms, setting up Mid
Vowel Harmony as a global rule makes the intuitively correct
claim that speakers are aware of the difference between
segments which have distinct basic sources and, accordingly,
avoid matching up the tenseness of a basic /0/ with that of
an o from a different basic segment.
2.6.4.2.2 In keeping with this, Mid Vowel Harmony will
be formulated as follows:
Mid Vowel Harmony (MVL) A penultimate mid vowel in a
stem will be assigned a tenseness feature identical
to that of an otherwise identical mid vowel
occurring in the final syllable of the same stem,
Just in case both originated as mid vowels at
the input to the phonological rules.
Note that the wording "at the input to the phonological rules"
instead of "at the most basic level" has been purposely chosen
in order to distinguish between stem-final mid vowels which
are introduced morphologically, and thus are present in the
stem at the input to the phonological rules but not, in one
sense, at the "most basic level", and phonologically derived
segments like o from /a/. Clearly MVH does match a penultimate
mid vowel to a morphologically derived final vowel, as in the
examples In 2.6.4 and, as was just seen, in Hab-Reps such as
rEwang-reweng.
99
2.6.5 At this point, now that the four rules which

determine the surface tenseness of mid vowels have all been
discussed and given their final formulations, it is appropriate
to briefly consider the operation of these rules as a group.
One thing which is especially notable is that two of the
four mid vowel tenseness assignment rules (MVL 1 and MVL 2)
have been seen to operate in an environment which must be
defined in terms of the original stem and not, as is the case
with the rules not involving mid vowels (/a/ to o#, a-Vowel
Harmony, and HVL), in.terms of the stem as it may occur in
the word.
2.6.5.1 The fact that two out of the four mid vowel
rules must be restricted in this way might cause one to
wonder whether this perhaps reflects something more than a
simple detail of the individual rules* environments, and,
instead, could indicate a more general fact about the oper
ation of mid vowel rules in the language. That is, as MVL 1
and MVL 1 have been formulated, each with its environment
being defined in terms of the basic stem, the implicit claim
is made that this is a separate fact about the environment
for each rule, which must accordingly be learned separately
for each rule by children constructing a grammar of Javanese.
It might, however, be possible that, instead, it represents
a more general principle which speakers learn to apply to
the operation of all rules determining the surface realization
of mid vowels.
)
1 0 0
2.6.5.1.1 In case one wanted to make the latter claim,

the restriction in question could be factored out of MVL 1
and MVL 2 and a general convention for the application of
mid vowel rules could be set up, specifying that their
environments are determined only on the basis of the (original)
stem shape, and not on the basis of the stem as it may appear
within the word. Note that the inclusion of such a convention
would make, in addition to the above claims, certain predic
tions about future developments within the grammar which would
be entirely different from those made by restricting the
environment for each rule individually, as in MVL 1 and MVL 2
formulated above: that is, making use of a general convention,
one would implicitly claim that any new rules affecting mid
vowels which might be added to the grammar would also apply
only within the stem; and, similarly, that if one of the mid
vowel rules began to operate on the basis of environments
outside the stem (e.g. If speakers began to produce alter
nations like Edum; Dem.*c-;IUm-E), all of the mid vowel rules
" V •
would be likely to exhibit similar behavior. Including the

restriction in each rule Individually would of course make no
such predictions.
2.6.5.1.2 Before actually including such a convention in
the grammar, one would of course want to consider the operation
of the other mid vowel rules, MVL 3 and MVH, to see whether
they provide any favorable or unfavorable evidence bearing on
such a step. As it turns out, the application of the latter
1 0 1
rules in the grammar is entirely consistent with the proposed

convention, but, at the same time, their operation doe3 not
necessitate any such restriction on their application.
This can be seen first of all by consideration of the
environment for the application of MVL 3 (2.6.3). Notice that,
while this rule is very similar to MVL 2, here it is not
necessary to include the restriction that the rule must
operate only when its environment is met within the stem, as
it was with MVL 2. That is, for the latter, which laxes a
penultimate mid vowel before a high tense vowel, the restric
tion had to be included to account for forms like Edum;
Dem. EdUm-E
• 1
versus klerU; Dem. klerU-nE, where, even though
- ■- — ■ »
the addition of a suffix might disrupt the original pattern

of the stem by opening its final syllable and allowing the
vowel of that syllable to remain tense, the mid vowel rule
still applies as though it were dealing with the original
undisrupted stem shape. With MVL 3, however, this is not
necessary because the addition of a suffix can never alter
the original stem significantly, since the vocalixm of the
final syllable does not alternate and is unaffected by the
opening or closing of the syllable in which it occurs. Thus,
even though the operation of MVL 3 would not be inconsistent
with the possibility that mid vowel tenseness assignment
rules are governed by the proposed convention, there is no
need to restrict its application to environments defined in
terms of the original stem shape.
1 0 2
A similar situation exists for Mid Vowel Harmony. That

is, because the manner of operation of this rule on a pen
ultimate vowel seems to be dependent solely on the tenseness
or laxness of an identical mid vowel in the final syllable
of a stem, and the tenseness of laxness of the latter is
always determined by MVL 1, whose operation is of course
determined on the basis of the shape of the stem, there can
be no question of whether MVH is consistent with this restric
tion. But once again, while the operation of MVH is not incon
sistent with the presence of the proposed convention in the
grammar, as long as MVL 1 is appropriately restricted, it is
not necessary to specifically restrict the application of MVH.
2.6.5.2 Thus, it seems that, considering the operation
of all four mid vowel rules, the only ones which really
necessitate any restriction on their application are MVL 1
and MVL 2. For MVL 3 and MVH, although such a restriction
is not necessary, neither is it counter-indicated. Because
this situation does not, in my opinion, necessarily warrant
positing any general convention for the application of all
mid vowel rules in the grammar, MVL 1 and MVL 2 will be left
as originally formulated; however, it is at least worth
keeping in mind the possibility of the presence of a more
general restriction on the operation of all mid vowel rules.
2.6.6 There is one more rule which may or may not be
indicated for the grammar of Javanese, but which at least
103
merits some discussion. This rule would be responsible for

laxing any mid vowel In a closed stem-penultimate syllable.
Recall from the discussion In 2.4.2.5 that high vowels
usually do not occur In closed penultimate syllables In
native words. The same may be said for mid vowels; however,
Just as with the high vowels, I do know of a few cases where
my informant claims such a form to be "good Javanese". One
of these is borgol 'handcuffs'. Note, however, that in this
word, one can not really be sure whether the penultimate mid
vowel is lax because it is in a closed syllable, or whether
penultimate /0/ has been laxed as a result of Mid Vowel Har
mony. Another such word is elmU 'knowledge'; but here again,
one cannot tell whether the e is lax because it is in a
closed syllable or because of the following high tense vowel
(i.e. by MVL 2). The same type of problem arises with every
"good Javanese" stem containing a penultimate mid vowel in
a closed syllable.
I suspect that this situation reveals something about
the basis for my Informant's Judgements about what constitutes
"good Javanese". It may well be that whenever he is not sure
how a given stem should be pronounced, he simply refects It
as being "non-native" or "old", since. In general, he has
few Intuitions about how non-native vocabulary should be pro
nounced. (cf. 1.7 for a discussion of how non-native vocabulary
is treated.) Thus, it may be precisely because the mid vowels
of stems like borgol and elmU fall within the environments of
104
both Mid Vowel Harmony and MVL 2, and he accordingly has a

guide to their pronunciation, that he accepts them as being
good Javanese words.
Whether this is the correct explanation or not, the
fact remains that I have so far found it impossible to deter
mine how mid vowels in penultimate closed syllables are treated
in native words; in non-native words, it is the case that mid
vowels are usually lax in closed penultimate syllables, but
they may also be pronounced in their tense variants, with,
in most cases, no difference in acceptability.
2.7 Before discussion of the mid vowel tenseness assign
ment rules is concluded, it seems appropriate to give some
attention to the relationship between a-Vowel Harmony, posited
earlier (2.3.3.3) and Mid Vowel Harmony, as well as to the
relationship between these two rules and the rules producing
the segments which are to be harmonized with.
2.7.1 The possibility should at least be considered that
a-Vowel Harmony and Mid Vowel Harmony, which have so far been
treated as independent processes, are actually both facets of
a single more general vowel harmony rule. While it is true
that both a-Vowel Harmony (which I shall refer to as a-VH)
and MVH operate to produce surface forms with two identical
vowels, and it is also true that both rules match a penultimate
vowel to a stem-final vowel, it appears to me that this simi
larity is outweighed by the essential difference in their
operation: while MVH is formulated as a tenseness-feature
105
assignment rule, a-VH does not involve tenseness at all, but,

rather, Involves backing and rounding /a/ to o.
Another point on which a-VH and MVH might profitably be
compared is unfortunately inconclusive, but, in my opinion,
quite interesting nevertheless. Recall that a-VH will operate
across two internal consonants, but only if this cluster does
not close the penultimate syllable of the stem. Thus, one
finds djoro 'drill* and kont.jo 'friend', but dJarwo 'meaning'.
One might wonder whether Mid Vowel Harmony shares this
characteristic. As it turns out, however, the structure of
the stems in the language makes it impossible for the linguist
(and, probably, for the child learning the language) to find
an answer to this question. The problems involved relate to
the discussion in 2.6.6 above. As was noted there, even though
Javanese contains many stems with closed penultimate syllables
containing /a/, alternating vowels do not usually occur in
this position, except in non-native vocabulary. Because of
the situation described in 2.6.6 for the few "good Javanese"
words with mid vowels in this position such as borgol 'hand
cuffs', it is impossible to tell whether the lax o of the
penultimate syllable occurs because the vowel is in a closed
syllable, or whether it is lax because of the following lax
o and the operation of MVH. Consequently, one cannot tell
whether MVH does or does not apply across syllable-closing
clusters.
However inconclusive the evidence presented by the
106
application of the two rules across two syllable-closing

internal consonants might be, the fact that MVH is essential
ly a matching rule (which operates non-vacuously to lax a
basically tense mid vowel) while a-Vowel Harmony backs and
rounds /a/ to o, would seem to make it extremely unlikely
that the two should be collapsed into a single process.
2.7.2 Some consideration should also be given to the
question of whether /a/ to o§ and a-Vowel Harmony might be
accomplished by a single process. If one accepts the analysis
of Hab-Rep forms given in chapter 7, there is fairly convinc
ing evidence that this cannot be the case, since /a/ to o§
must apply before the morphological vowel changes associated
with Hab-Rep Formation take place, while it appears that a-VH
does not apply until after the morphological rules in question.
Because it would be of little value to go through all the
arguments for this ordering here, the reader is referred to
chapter 7, and specifically to Sect. 7.5 of that chapter.
The situation is not quite so straightforward if one
looks at the possibility that Mid Vowel Harmony and Mid Vowel
Laxing 1 might actually be accomplished by a single process;
here there is no evidence available Involving ordering. While
it is possible to establish that MVL 1 must provide the tense
ness specification which determines the actual operation of
MVH, there Is no convincing reason why the matching function
of MVH could not be taken over by a single rule combining the
functions of MVL 1 and MVH.
107
One rather Inconsequential problem involved In taking

this approach might arise when a penultimate mid vowel in .a
given basic stem must be harmonized to a final vowel which
has been introduced morphologically. It is not at all clear,
for example, whether one would want to claim that the lax o
in the stem-final syllable of derived forms such as Caus.
mbodo-qakE from bOdO ’stupid' is morphologically introduced as
• •
/0/, with the regular phonological rules of the language pro

viding its lax surface realization, or whether it is initially
introduced in its lax variant. If the former were the case,
there would of course be no problem, since the penultimate /0/
of the stem would be harmonized when the consolidated MVL 1-
MVH process laxed the /O/ of the final syllable; however, if
the latter is the ease, then some means must be found for
harmonizing a penultimate mid vowel to a morphologically
introduced mid vowel in the final syllable of the stem.
However, even this problem might be avoided by giving
a slightly different interpretation to the harmonizing process.
Alternatively, one could posit for the grammar of Javanese
some principle which says that, any time a grammatical process,
morphological or phonological, alters a relevant segment of
a stem-final syllable, an identical segment in the preceding
syllable will undergo the same change.
2.7.3 When the general effects of these possibilities on
the overall grammar of the language are.considered, it seems
that the choice of one over the other does not really have any
108
great consequence. Thus, having at least briefly described

some plausible alternatives, I have chosen to retain the
analysis already posited, which treats both a-VH and MVH as
independent processes in the’grammar, separate from /a/ to off
and MVL 1.
2.8 In summary, then, I propose for Javanese a grammar
containing the following rules for producing the correct
surface distribution of vowel sounds in the simplest and most
natural way. The connecting lines indicate rules which must
be ordered with respect to each other. Section numbers
Indicate where the final formulation of each rule was given.
/a/ to o# Sect. 2.3.1.3
a-Vowel Harmony tv
2.3.3.3 and 2.3.4.2
High Vowel Laxing tl
2.4.2.5
Mid Vowel Laxing 1 ff 2.6.1.2
Mid Vowel Laxing 2 H
2.6.2.3
Mid Vowel Laxing 3 It
2.6.3
or
Mid Vowel Harmony 2.6.4.2. 2
2.9 In the previous sections of this chapter, discussion
has been confined to the vowels which occur within the free
stems of the language (I.e. to morphemes which can serve as
a base for derivation). This course was chosen because the
latter are the only morphemes in which one finds the alter-
nations necessary to establish a phonological analysis. 17'
At the same time, however, it Is an undeniable fact that the
109
bound morphemes must also be provided with a surface reali

zation by the phonological processes of the language, and
thus must be given some attention.
2.9.1 The following list of productive Javanese suffixes
is based on a similar but much more complete listing which
occurs in Horne 197^:xxv:
-an (Substantive or verb-forming)
-kU (First Person Possessive)
-mU (Second Person Possessive)
-o (Imperative or Subjunctive)
-(n)anE (Locative Optative)
-(n)E (Third Person Possessive or Demonstrative)
-(n)£n (Passive Imperative)
-(n)I (Simple Locative)

-(n)ono (Locative Imperative)
-(q)akE (Simple Causative)
-(q)nE (Causative Optative)
-(q)no (Causative Imperative)
Notice that the surface realization of every one of these
suffixes, taken in isolation, is predictable using the rules
already posited for vowels which occur In stems. Thus, for
example, the vowel of the suffix -o, as well as that of -(q)no3
may be thought of as coming from basic /a/ by means of /a/ to
o#„ Similarly, the vowels of Loc. Imper. -(h)ono may be
derived from /-(n)ana/ by /a/ to op and a-Vowel Harmony. For
all other suffixes, the surface shape reflects the basic
110
vocalism, with none of the phonological rules having applied,

simply because the regular environments for their application
are not met.
In general, this is all that may, or needs to be, said
about the derivation of the suffixes taken in isolation.
There is, however, a need for some discussion of the operation
of the phonological rules (specifically, /a/ to o# and a-Vowel
Harmony) in the derivation of certain actual words consisting
of a stem and the Imperative suffix -o, from basic /a/.
2.9.1.1 It is significant that the Imperative is the
only formation in the language where a suffixed stem in final
/a/ appears on the surface with o as its final segment (e.g.
lUngo ’go (away)’: Imper. nglUngo-o ; t^ko ’come’; Imper.
n^ko-o ). In all other formations throughout the language,
stem-final /a/ does not appear as o on the surface unless it
is in absolute word-final position.
One way to account for this unexpected occurrence of
stem-final o would be to attribute It somehow to the operation
of /a/ to ojt. Taking this tack might necessitate some alter
ation of the /a/ to o# rule such as adding a grammatical
environment which specifies that the rule may operate before
the Imperative suffix -o. Alternatively, one might mark all
Imperatives built on stems In final /a/ as positive exceptions
to /a/ tc o#, undergoing the rule, even though its environment
Is not actually met. In addition, there are a number of other
purely mechanical ways to produce the correct surface forms
Ill
for Imperatives like nglUngo-o using the /a/ to off rule;

however, all of these must necessarily share the same faults
as those Just suggested in that all are ad hoc and offer no
explanation for what would seem to be the most striking fact
about the unexpected occurrence of stem-final o: namely,
that this segment occurs exclusively before a suffix which
Itself consists of -o, from basic /a/. Although this could
be coincidence, it could also be suggestive of an alternative
analysis.
2.9.1.2 A much more satisfying source for the occurrence
of surface stem-final o in Imperatives might be arrived at
by crediting the production of such segments to a-Vowel
Harmony, rather than to /a/ to off. That is, one might claim
that stem-final /a/ becomes o in Imperatives in order to
harmonize with the -o of the Imperative suffix. Note that
this approach would differ significantly from those considered
above by making the claim that the stem-final segment of an
Imperative like nglUngo-o is realized as o not Just because
the following suffix is the Imperative marker, but rather,
precisely because the Imperative marker happens to be surface
o, derived from basic /a/.
2.9.1.3 There are, however, certain problems involved
in using a-VH to account for stem-final o in Imperatives. It
will be recalled from 2.3.^.2 that, within stems, a-VH
operates across one or more consonants, as long as the partic
ular consonant configuration does not close the penultimate
112
syllable of the stem. Thus, one finds dowo 'long*, from

/dawa/, with the effects Of a-VH, but darmo 'duty', where the
rule has been blocked by a syllable-closing consonant cluster.
If a-VH were to be used to account for the stem-final o of
Imperatives like nglUngo-o , some means would have to be
found to account for why the rule does not appear to have
applied in Imperatives like mangan-o , from pangan ’food'.
That is, based on its operation within stems, a-VH might
be expected to apply in forming Imperatives from stems with
final syllables of the shape -aC, so that one might expect
to find, in addition to nglUngo-o , forms like *mangon-o .
There is, however, a fairly plausible explanation for
the non-occurrence of Imperatives like *mangon-o which
involves the difference between what closes a stem-internal
syllable, versus what closes a stem-final syllable. That is,
although two- or even three-consonant clusters may occur
stem-internally without closing the preceding syllable,
stem-finally, the occurrence of even one consonant will close
a syllable. Accordingly, it could conceivably be the case
that, although a-VH is not blocked by a single consonant when
the environment is stem-internal, the same rule, when operating
on the vowel of a stem-final syllable before a morpheme
boundary, is_ blocked by a single consonant. Another way to
look at this would be to compare the proposed operation of
a-VH across a morpheme boundary to that of mid vowel rules
like MVL 1 and MVL 2, whose operation, it will be recalled
113
from 2.6.4, is determined on the basis of the shape of the

stem-final syllable as it occurs in the isolated stem and not
as it occurs in a word consisting of a stem plus a vowel-
initial suffix.18
Although the preceding discussion of the possible source
for forms like Imper. nglUngo-o has been very sketchy, what
is important is that it would seem to have been shown suf
ficiently clearly that, in one way or another, the phonological
framework which has been set up in preceding sections of
this chapter can lend itself to the analysis of all types of
primary and derived forms of the language, including those
which are slightly problematic, as well as to those, represent
ing the vast majority of words in the language, where the
operation of the phonological rules as formulated is straight
forward and entirely regular.
2.9.2 Finally, it should be noted that the surface
vocalism of prefixes may also be produced by the rules already
posited.1^ Dealing with Javanese prefixes is especially
simple, since in most prefixation processes which are still
productive, no vowel except a or £ ever occurs. The £_ vowel
of course does not alternate, while in a situation where
prefixed /a/ might conceivably become o, this does not occur;
this is presumably because the environment for /a/ to o#
specifies that the /a/ must occur In a word-final environment
before it can become o.
In the two productive prefixes I know of containing
alternating vowels, dl-, the Simple Passive prefix, and koq-,
the Second Person Passive prefix, the vowels are always
realized In the shape cited. For dl-, this may easily be
attributed to the fact that the I In the prefix always occurs
in an open syllable, regardless of the shape of the following
stem, so that HVL may never apply to lax it. With koq-,
the lax o vocalism may be attributed to the operation of
MVL 1, if one makes the assumption that the latter may apply
separately to the prefix as well as to the stem. This assump
tion is not Incompatible with the restricted environment
for MVL 1 within words and may perhaps be attributed to the
fact that koq- (like dl- and all other productive prefixes in
the language) does not ever enter into any interactions with
the stems to which it Is affixed (unlike the suffixes), so
that speakers might accord prefixes more of a general "stem"
status and apply the phonological rules accordingly.
115
FOOTNOTES
Ît should be noted that I am making use of the binary
features in the above matrices for convenience of reference
and do not necessarily wish to claim, for example, that the
actual phonetic distinction between pairs like I and i is
exclusively one of tenseness versus the absence of tenseness;
instead, I would wish to leave room for the possibility that
the sounds of a language are actually distinguished by
relational or hierarchical criteria. To the best of my know
ledge, however, the use of such criteria instead of binary
features would have no crucial effects on the ultimate con
clusions reached In this dissertation, although some fairly
minor differences would arise in certain areas.
2
The reader is asked to disregard for the moment the
fact that the horizontal line indexed as a contains some
forms which have o and not a In their final syllables. The
reasons for this organization will become clear in the course
of later discussion.
3
The sole exceptions to this are, to the best of my
knowledge, Ora 'no' and the exclamation la or lha. I know
of no Javanese words with final
4
There are certain instances where it appears that e
may be optionally Inserted to break up some Initial consonant
clusters (cf. 1.8.1.1). In all such cases, however, this
appears to be the result of a very low-level optional process
which will not be treated in this dissertation.
5
Slashes will be used to distinguish any non-surface
form, including basic as well as intermediate levels.
6
There Is in fact a Javanese word sworo, which is unre
lated to swargo and which means ’noise, sound1.
7
These conclusions are reflected In the organization of
Chart I, where the forms contained in the column and line
indexed a do not always actually have a In their surface pro
nunciation, but instead exhibit o in appropriate environments.
8
In forms such as ngg^ntl-o, where a tense vowel and
a following lower or more front vowel are contiguous, Javanese
speakers normally pronounce a glide between the two vowels. I
have not indicated the presence of this glide in examples un
less it Is directly relevant to the discussion at hand. The
process of Glide Formation will be treated in 3-3.3.2.2.1.
116
9
There Is one other group of suffixes whose addition
does not result in a derived form with tense I or U in the
stem-final syllable. These are the Causative-forming suffixes
-(q)akE, -(q)no, and -(q)nE. Because Causatives are in many
interesting respects different from other derived forms of the
language, I will not deal with them here, but will refer the
reader to chapter 4.
10
It must be noted that there are two groups of words in
the language where tense high vowels do occur in closed stem-
final syllables. The first of these consists of certain
.xpressive or onomatopoeic primary forms, such as girfndUt
'bouncy, springy' or kgrklt 'creaking', which may be charac
terized as a group as being exceptions to High Vowel Laxing.
The other group consists of derived Elatives, which will be
discussed at length in chapter 5.
11
For the purposes of this discussion, word-final sur
face o which was shown earlier to derive from basic /a/ will
be disregarded, but cf. 2.6.4.1 below.
12
There are certain Elatives which at first glance may
appear to be exceptions to this statement. For discussion of
such forms, see chapter 5.
JIt is of course logically possible that there could
be a difference in underlying shape between the front and
back mid vowels, so that, for example, for E/e, tense /E/
might be basic, while for O/o, lax /o/ might be basic. This
possibility can, however, be easily eliminated on the basis
of the fact that, in all respects, the mid vowels behave
identically, sharing environments for surface tenseness or
laxness. It would thus be most unnatural to assign the front
and back mid vowels divergent types of basic representations.
I1!
Notice that, taking this approach, there would also
have to be some constraint on laxing of penultimate /0/ before
a word-final o which derives from /a/. This would be needed
In order to produce forms of the shape COCo (cf. Chart I).
Such a constraint could be built Into the grammar easily by
either including the limitation "In a stem-final closed syl
lable" in the formulation of the vowel harmony rule, or by
ordering the process responsible for producing o from /a/
after Mid Vowel Harmony.
15 Again, the operation of these rules would have to be
in some way constrained to guarantee that a penultimate /o/
will be tense before a final o from /a/. This could again be
accomplished in either of the ways suggested in the preceding
footnote.
117
Although the reader may be struck by the similarity

of the consonant configurations of these examples, this
similarity is entirely fortuitous. The stems in these
examples were chosen because they form minimal pairs which
nicely illustrate the effect of final-syllable shape on
mid vowels. Although other pairs could have been chosen,
the ones used were picked mainly on the basis of their
possessing all the derived forms needed for comparison.
17
The only possibility for the bound mrophemes to show
such alternations would be if a number of suffixes could occur
on the same stem. This does frequently happen but the only
suffixes which are ever followed by another suffix contain
non-alternating vowels (e.g. dj £r0 ’deep’; Subst. dj^rO-an;
Dem. Subst. dj^rO-an-E).
18
Notice that positing such a constraint on the oper
ation of a-VH across a morpheme boundary would be consistent
with the fact that, in contrast to nglUngo-o , derived forms
like Loc. Imper. nglUnga-nono do not have o as their stem-
final segment, even though the suffix contains surface o
from /a/. As will be discussed in 4.3-3 below, there is
evidence to indicate that the n which occurs in Locatives
built on vowel-final stems is indeed felt to be a part of
the (derived) stem itself, rather than being a part of the
suffix. If this is the case, then a word like nglUnga-nono
would have consonant-final /nglUngan-/ as its stem and would
thus, like mangan-o, be blocked from undergoing a-VH to
become *nglUngo-nono.
19
For a list of Javanese prefixes and their functions,
cf. Horne 1974 :xxi. For discussion of all productive pre
fixes and their usage, cf. Horne 1961.
118
CHAPTER III
CONSONANT PHONOLOGY
3.0 Compared to those Involving vowels, the morpho-
phonemic processes involving Javanese consonants are few in
number and relatively straightforward, although in some cases
there are interesting problems which merit discussion.
3.1 One striking fact about the pronunciation of
Javanese words is that, while both light and heavy consonants
occur word-initially and internally, only light consonants
occur in word-final position. At the same time, only light
consonants occur stem-finally before a consonant-initial suf
fix such as -kU or -mU. This distribution suggests the pres
ence in the grammar of the classic voicing neutralization in
word-final position or before a consonant. The presence of
such a process in the grammar is borne out by comparison of
simple stems and their suffixed Demonstrative counterparts:
stem gloss derived
gaip secret galb-E
kar^p wish kar^p-E
ngalap put to use ngalap-E
s^bap because, reason s^bab-E
bibit origin blblt-E

bl^b^t protective cover bl^b£d-E
djOget the classical dj Oged-E
Javanese dance
119
bUdjtfk deaf bUdj*g-E

gaduk a bump, thump gadUg-E
grUnj^k complaint grUnj^k-E
Ub^k Inharmonious Ub^k-E
For simplicity, and because It would be very difficult
to characterize the difference between light and heavy con
sonants accurately in terms of phonetic features, I will
formulate the rule which neutralizes the light versus heavy
consonant contrast simply in terms of "light" and "heavy"
Consonant Neutralization C -> light/
(Con-Neut) heavy
Note that as Con-Neut is formulated here, one would expect to
find no examples of a heavy consonant occurring before another
consonant within morphemes. This is in fact the case; even
though stems such as apsah 'valid* and patml 'first wife in a
polygamous royal marriage' are written in Javanese orthography
with internal clusters containing an initial heavy consonant,
apparently reflecting their pronunciation at an earlier stage
of the language, they are pronounced with light consonants.
Note also that Con-Neut actually affects only the stops b,d,
and £, since morphemes of the language never end in any other
heavy consonant, simple or nasalized (cf. 1.5.I).2
On the basis of the distribution of light and heavy stops
in the language and the obvious presence in the grammar of a
rule like Con-Neut, it seems quite reasonable to posit at least
the basic heavy stops /b/, /d/ and /g/ and the light stops /p/
120
and /t/. The situation Is not quite so obvious for the light
velar consonant, however, and requires some discussion.
3.2 In the examples listed above Illustrating the
neutralization of word-final consonants, the only examples
given for word-final non-alternating k have £ in their final
syllables (cf. grUn^k, Ub£k). This was not a mere coincidence
In the selection of examples; the £ environment for non-alter
nating k was chosen simply because that is the only vocalic
environment in the language where there is a merger of word-
final k which alternates with intervocallic £, and k which
does not alternate. (Note that word-final k is written, both
in modern and old Javanese orthography, after all vowels, but
it is pronounced as the glottal stop everywhere except after
) Thus, it is only final velar consonants after £_ whose
pronunciation merges; after all other vowels, /g/ does indeed
become k, but (written) <k> is pronounced c^. One finds exam-'
pies like:
stem gloss derived
Ub^k Inharmonious Ub^k-E
Ubr^k noise, racket Ubr^g-E
:ompare the following:
babaq skinned, bruised babaq-E
babak the equal of babag-E
dj^ruq citrus fruit dj^rUq-E

b^duk mosque drum bedUg-E
121
glltiq stick, elub glltlq-E
gllik cylindrical glllg-E
The distribution of surface £ and k in Javanese poses a
rather interesting problem for the linguist attempting to
establish the basic consonant inventory and phonological rules
of the language. Although there is clear comparative evidence
that at a chronologically earlier stage Javanese had only
basic /k/ and added a rule to realize this segment as £ in
certain surface environments, it is not at all clear whether
the synchronic grammar of the modern language has maintained
this relationship. In the following paragraphs, evidence will
be presented both for and against the hypothesis that Javanese
speakers possess a grammar which makes a connection between
surface k and deriving both surface segments from a single
base.
3.2.1 First of all, surface c[ and k are in complementary
distribution: k occurs freely word-initially and internally,
and word-finally it occurs after on the other hand, the
occurrence of £ is, with very few exceptions, restricted to
either absolute word-final position or stem-final position

3
before a suffix, when the preceding vowel is anything but
3.2.2 Although complementary distribution of surface ^
and k indicates the possibility that they might both derive
from a single segment, this is by no means a certainty; there
must be some evidence that speakers feel a real relationship
to exist between the two surface realizations before such an
122
assumption should be made.

3.2.2.1 As may be seen from the examples in 3.1 and 3.3,
and in various other places throughout this dissertation, in
general, there is no opportunity for alternations between
k and £, since in initial position and within stems, the
language does not produce consonant alternations, and in
stem-final or word-final position, final-syllable whose
presence or absence may be thought of as the crucial environ
ment for any possible alternation, is not usually altered by
morphological processes. There do exist, however, two forma
tions where the necessary change in the vowel of the final
syllable of stems is possible and does occur.
3.2.2.1.1 The first of these is in the Hab-Rep forms
of stems in final -0k3 where the is regularly replaced by
a in the first member of the doubled pair. In such forms,
based on the data collected from my informant, two equally
good pronunciations occur, one with the change from ^ to a
accompanied by a change of consonant from k to tj_:
stem gloss Hab-Rep 1 Hab-Rep 2
bUnjgk feel cramped bUnak-bUn^k bUnaq-bUnj£k
grUn^k complaint grUnak-grUn^k grUnaq-grUn^k
UbgJk inharmonious Ubak-Ub^k Ubaq-Ubfifk
3.2.2.1.2 The second formation where the alternation in
question has an opportunity to occur Involves Elatives of
a
stems ending in In data cited by Uhlenbeck (19^9:^!),

Elatives of some stems of this shape were characterized by a
123
change of the final-syllable vowel to £ and an accompanying

change in the pronunciation of the final consonant. Uhlenbeck
cites only two examples, but he seems to Imply that this was
the normal way of forming Elatives from stems In final £ at
the time his data was collected:
stem gloss Elatlve
£ndeq low gnd^k
• •
tj^daq near tj^d^k

• •
At the same time, however, he notes that some stems In

are subject to the more general type of Elatlve formation In
which the vowel of the final syllable Is replaced by the marker
vowel I or U. The latter formation would not, of course, give
any opportunity for an alternation between c[ and k to arise.
The situation differs somewhat in the dialect of my
informant, who supposedly speaks the same Standard Javanese
dialect as Uhlenbeck*s informants, but who must have learned
the language at least one or two generations later. On the
basis of my data, it appears that the type of Elatlve formation
using I and U as the characteristic marker vowels, in the time
elapsed since Uhlenbeck*s work, has been generalized to affect
all types of adjective stems, including those ending in glottal
stops, so that there is almost no opportunity for an alterna
tion between £ and k. Thus, for example, for my informant,
the Elatlve of ng^laq 'thirsty5 is not ang^lgk, as was probably
2j
the case for Uhlenbeck5s informants, but ng^llq.
124
However, there are still a few examples where the obsolete

type of Elatlve formation survives, so that my Informant
recognizes at least the following Elatives for commonly occurr
ring adjectives:
stem gloss Elatlve
Enaq easy En^k^
tj/^daq near tj£d£k

• •
tj^ndaq short (length) tj£nd£k

• «
tjjrfkaq short (distance) tj£k£k

It is interesting to note in passing that in at least one of
these cases, the retention of the old Elatlve form serves to
avoid homonymy in the Elatives of two closely related words.
That is, alongside tj^ndaq 'short (length)', there is another
word, tjjgndeq, which means 'short (height)1. The Elatlve of
the latter is the normal tj^ndlq. If tj^ndaq also underwent
• •
the normal Elative-forming process, Its Elatlve would also

come out *tj^ndlq.
The fact that most adjectives in final £ now follow the
general pattern of Elatlve formation casts some doubt on the
value of the few cj/vk alternations which do occur as evidence
in favor of modern-day speakers' possessing a feeling for any
relationship between the two segments. Although they do
appear to alternate in a few pairs like those listed above. It
Is quite likely that the Elatives of this type still In use
have been lexicalized and are no longer felt by speakers to
be derived.
125
Still, even If this Is true, the basic stems and their

Elatives do bear such a strong resemblance to each other that
one would not want to consider tjgdaq and tjgdgk, for example,
• •
to be memorized as totally independent lexical items with no

more relationship than, for instance, t.j^daq and Enaq; nor
would one want to treat it as mere coincidence that all Ela
tives of this type (i.e. reflecting the obsolete type of
Elatlve formation) end in -£k. Rather, Elatives like these
would probably have to be assigned some special status; and
to that extent then, perhaps speakers do feel some special
relationship to exist between the £ of tj^daq and the k of
the corresponding Elatlve tj^d^k.
3.2.3 Noting that sometimes the treatment of borrowed
words in a language can provide a clue to the presence or pro
ductivity of a rule in the grammar, in the course of discuss
ing the status of surface £ and k, Uhlenbeck (19^9:^5) devotes
some attention to the Javanese treatment of borrowed words
such as Dutch muzldk 'music', druk 'impression, print', and
boek 'book', among others, with the hope of discovering a clue
to whether Javanese has a productive rule which limits the
occurrence of -k to the position after the vowel Such words
have two Javanese pronunciations existing side by side: mUsik:
mUslq; druk: drug; and buk: bug, with the pronunciation listed
first mainly in use by more sophisticated speakers, or those
who are familiar with Dutch (cf. 1.7) 3 while the second is
used by more naive speakers. Thus, Uhlenbeck notes, one could
126
interpret this as indicating that the pronunciation with

reflects the real Javanese treatment. This would be evidence
that the grammar might indeed include a rule saying that k-
can be pronounced at the end of a word only after £, while
the pronunciation with -k simply reflects a more artificial
Dutch-influenced treatment.
On the other hand, he argues, this assumption might very
well not be valid. The two variant pronunciations of such
borrowed words could be the result of their coming into the
language at two separate periods: that with at an earlier
time when there surely was some synchronic process in the
grammar connecting k and a; and that with -k at a later date,
after the loss of the process in question. Thus, Uhlenbeck
concludes that the treatment of borrowed words, in this case
at least, is of dubious value. And I am inclined to agree
with this opinion.
3.2,4 In summary, then, Modern Javanese offers a certain
amount of evidence that surface k and £ might be derived from
a single basic segment. The two are in complementary distri
bution, with k occurring everywhere except in stem-final
position (after any vowel except g) . There is also an option
ally occurring alternation between and k in stems and their
corresponding Hab-Rep forms, such as Ub^k and Ubaq-Ub^k.
Furthermore, it is possible that speakers feel some relation
ship between the final segments of certain adjectives and
their Elatives, such as tj^ndaq and Elat, tj^nd^k. Finally,
127
It Is possible that the treatment of borrowed words In stem-

final velars, such as Dutch muziek, could indicate that
speakers have some feeling for a connection between k and c[.
3.2.5 At the same time, there is evidence which makes it
unlikely that present-day speakers really derive both k and c[
from a single segment. First of all, there is the undeniable
fact that in the vast majority of cases, £ and k do not alter
nate, so that, except for the Hab-Rep forms of stems in -£k,
no segment which is in a given stem pronounced £ or k, res
pectively, will ever, in any environment, be pronounced k or
£, respectively; and although Hab-Rep forms are heard often
in informal speech, by virtue of their very specialized meaning
(cf. 7.1), they are much less common than most other derived
forms. This fact alone casts some doubt on whether a child
learning Javanese would hear enough examples of Hab-Rep forms
from stems in -ffk to develop a feeling for a general connection
between £ and k.
3.2.5.1 At the same time, because of the effects of Con-
Neut described above, any process which derived k and c[ from a
single basic segment would be extremely opaque and thus quite
difficult for children learning the language to internalize.
As the concept of opacity was introduced in Kiparsky 1971a
a rule is opaque to the extent that the segments which are
the output of the rule occur in other environments on the
surface, or to the extent that one finds on the surface the
segments which form the rule’s input.
128
3.2.5.2 In Javanese, as a result of the presence of

Con-Neut In the grammar, a child learning the language would
encounter numerous cases where k is pronounced in a stem-final
environment other than after £, as in, for example, babak
'the equal of', b)#duk 'mosque drum', and gllik 'cylindrical',
from basic /babag/, /b^dUg/, and /glllg/, respectively. The
situation is further complicated by the fact that he will
hear forms like ambt*k 'assumption', where the final k alter
nates with £ in suffixed forms like Dem. amb^g-E , and others
like grUn^k 'complaint', where the k does not alternate. It
seems likely to me that a process whose operation is obscured
to such an extent would no longer be Included as a viable
rule of the grammar.
3.2.5.3 Finally, In the present-day language, many stems
which end in heavy consonants underlyingly, but which are
pronounced with light consonants because of the operation of
Con-Neut, seem to be falling together with those whose final
consonant Is basically light. That is, they are giving up
their alternations, so that, for example, for many speakers
the word which Is pronounced Udot 'smoke' has, in place of,
or in addition to, the standard suffixed Demonstrative UdUd-E,
a form UdUt-E. What Is perhaps crucially relevant to the
present discussion involves forms like godok 'boil', with
final k from /g/ by Con-Neut. While for some speakers the
addition of a vowel-initial suffix always results in a surface
derived form which reflects the basic stem-final heavy segment
129
(e.g. Subst. godog-an), for others, the suffixed form Is

either always or optionally pronounced godok-an. And for
those dialects where complete leveling has taken place and
pairs like godok; godok-an occur regularly, there is no longer
• •
any reason to assign godok a basic form with a final heavy

consonant. However, once this leveling had taken place, there
would be presumably no way for speakers to differentiate
between basic forms like /gOdOk/, which are realized on the
surface with a final velar, and all those forms where a stem-
final k may not occur on the surface except after If there
is still a process included in the language which produces
stem-final and k from a single basic segment, it is extremely
doubtful that forms like godok; godok-an would be allowed to
arise and remain in the language; instead, one would expect
to find pronunciations such as *godoq, with the proper variant
m
occurring in stem-final position.

3.2.6 Having considered all the above arguments, I have
chosen to include in the underlying consonant inventory of
Javanese both /k/ and /q/. Although the grammar of the lan
guage did at one time possess a rule to derive from basic
/k/, it seems to me that as the result of subsequent develop
ments this rule has become so obscured as to be virtually
lost from the phonology. This of course does not necessarily
mean that no trace of it remains In the grammar, since It
must presumably be a rule of Javanese morpheme structure that
130
basic /q/ occurs only stem-finally after vowels other than

while /k/ occurs freely, except stem-finally, where it is
found only after
If both /k/ and /q/ are posited, however, some explana
tion must be found for the alternation which appears optionally
in Hab-Rep forms such as Ubaq-Ub^k. Although I can find no
really satisfying explanation for this alternation, it is
possible that it is felt by speakers to have much the same
status as that of Elatives like tjjrfd^k, from tj^daq, where
• •
some relationship must be felt between the Elatives and their

stems, but where this relationship is not necessarily one that
is easily characterizable by the methods commonly used by
generative phonologists.
3.3 In addition to Con-Neut, the grammar of Javanese
includes another phonological process affecting consonant
segments, the operation of which may be seen by comparison
of the primary, Pirst Person Possessive, and Demonstrative
forms of stems like the following:
stem gloss Poss.l Dem.
aneh strange aneh-kU ane-E
arah direction arah-kU ara-E
baluh ballast baluh-kU balU-w-E
kokoh rice mixed kokoh-kU koko-E
Into soup
slsih side slsih-kU sIsI-y-E
As can be seen from examples like the above, stem-final
h appears in word-final position and before consonant-initial
131
suffixes, but not before vowel-initial suffixes (i.e. inter-

vocalically) J The consonant h does not occur initially in
native Javanese vocabulary, nor is it usually pronounced
internally, although it is written in certain words in the
latter position; moreover, h does not usually occur stem-
finally after £.
3.3.1 It should be noted that my data differs somewhat
from that of Uhlenbeck (1949:211-124) with respect to the
occurrence of h. According to Uhlenbeck, for stems in final
-ah, the h is retained (to a greater or lesser degree) before
all vowel-initial suffixes except the Simple Locative suffix
-(n)I. When the stem-final h is preceded by certain other
vowels, it may be deleted or realized in varying degrees of
audibility, depending on the nature of the following suffix-
initial vowel. In addition, Uhlenbeck claims that h is
pronounced intervocalically in certain stem-internal environ
ments. I can perceive no such occurrence of h in any inter
vocalic environment, and, at the same time, although my
informant freely admits to pronouncing an h in words like
arah-kU, he does not claim the presence of h in ara-S , for
example. From these differences, one might conclude that
perhaps in the time elapsed since Uhlenbeck's data was collec
ted, the environment for deletion of h has become more
generalized, so that it now includes all Intervocalic environ
ments. On the other hand, I regretfully admit that this
apparent generalization of the rule could be merely a result
132
of faulty perception on my part.

3.3.2 The surface distribution which has been described
for h, based on the author's own data, may be accounted for
by positing basic /h/ which occurs stem-finally and which is
deleted intervocalically by the following rule:
h-Deletion h J9 / V V
It is quite clear from examples like those above (3.3) that
the rule of Mid Vowel Laxing 1 (2.6.1.2) must apply to deter
mine the tensenss of final-syllable mid vowels before stem-
final h is deleted before a vowel-initial suffix, since one
finds forms like kokoh; Dem. koko-E , instead of *kOkO-E,
which might be expected if h-Deletion opened the stem-final
syllable before MVL 1 could apply. On the other hand, it is
impossible to tell anything about the relative ordering of
h-Deletion with respect to High Vowel Laxing, since the output
from basic /sIsIh-E/, for example, would be sIsI-y-E, regard
less of which of the two rules, HVL or h-Deletion, applied
first; the stem-final syllable would be open and, consequently,
the environment for HVL would not be met in either case.
3.3.3 Although the examples given so far represent
pronunciations which my informant seems to prefer as being
most correct, in informal speech, there do arise variant
pronunciations which are worth mentioning. These variants
occur only in forms derived from stems whose final syllable
contains a (lax) mid vowel followed by h 3 so that, in addition
to the "correct" pronunciation given for the relevant stems
133
above in 3.3, the following less acceptable variations may

occur:
stem gloss Poss.l Pern.
aneh strange aneh-kU anE-E
kokoh rice mixed kokoh-kU kOkO-w-E
into soup
3-3.3.1 It is interesting that in the above variants it
appears that the ordering of the rules MVL 1 and h-Deletion
is exactly the reverse of that which was seen to obtain in
the more standard pronunciations; the only ordering situation
which could produce the tense variants of the stem-final mid
vowels which appear on the surface in forms like anE-E and
kOkO-w-E would be one in which the stem-final consonant h
had already been deleted when MVL 1 applied.
3.3.3.2 At this point, some speculation seems in order
as to why such variants would arise. It would appear that
their origin may be traced to two factors. First, one of the
most striking features about standard forms like ane-E and
koko-E is their opacity (cf. 3.2.5.1) with respect to MVL 1.
That is, the stem-final mid vowel occurs in its lax variant
on the surface, even though it appears to be in the environ
ment where tense variants normally occur. Even though, in
general, orderings producing such opaque surface forms tend to
be eliminated from grammars, in this case (as in others dis
cussed, for example, in Kisseberth 1973 and in chapters 4 and
5 of this dissertation) there are extenuating circumstances:
13^
It Is precisely the opaque Interaction of h-Deletion with

MVL 1 which acts as an indicator of the basic shape of the
stem and thus avoids possible confusion. In derived forms
where the suffix is a variable one, such as -(n)E, this
function is not of such great Importance as in those where
the suffix Is always vowel-initial (i.e. -o and -an): in the
former, the surface presence or absence of the -n of the
suffix serves to signal the basic vowel- or consonant-final
character of the stem; in the latter, however, even this clue
is not available.
3.3•3.2.1 It is not unthinkable that, in spite of this
important role played by the opaque ordering, Javanese speakers
do have a tendency to choose the reverse ordering. The result
ing pronunciations are transparent in that the stem-final vowel
is in an open syllable and is, accordingly, tense; and to
achieve this transparency, it might be worth risking possible
confusion and homonymy by sacrificing preservation of the
character of the basic stem.
3.3.3.2.2 A second factor which might contribute to
production of the sort of variants In question is the greater
ease with which they may be pronounced In fast colloquial
speech. Aside from the fact that it is simply easier to
pronounce two vowels in succession if the first Is tense,
perhaps because there Is a natural tendency to develop an
off-glide between the two sounds under these circumstances,
the grammar of Javanese possesses a rule which actually
135
inserts a clearly pronounced glide between certain pairs of

contiguous vowels. Before continuation of the present discus
sion, it will be helpful to devote some attention to describing
the operation of this rule in more detail.
3*3.3-2.2.1 Glide Insertion may be characterized as the
insertion of a front or back glide between two contiguous
non-identical vowels, just in case the first is tense and
o
the second is either lower or farther front than the first.
The choice between jr. or w as the inserted glide depends on
whether the first vowel is itself front or back. Notice
that this rule never applies when a or ^ is the first vowel,
since these sounds are never tense, while it will always
apply if the first vowel is high (except when an identical
high vowel follows), since in all cases where the contiguous
vowel environment for Glide Insertion is met, high vowels are
invariably tense and are also always followed by a lower vowel.
3.3.3*2.2.2 Note also that the rule almost always
operates across the boundary between stem and suffix, since
the proper environment never arises within morphemes. The
latter is attributable to the fact that the language possesses
very few native morphemes containing contiguous vowels, and
most of these have a as the first vowel sound. This situation
results from various historical developments, In the course of
which, any morphemes which arose with other contiguous vowels
(those which would have met the environment for Glide Insertion)
were either contracted into one syllable or were lexicalised
136
with Intervocalic w or depending on the historical stage

at which they arose.
3.3.3.2.2.3. The effects of Glide Insertion may be seen
clearly in the following examples of vowel-final stems and
derived forms with vowel-initial suffixes:
stem gloss Subst. Imper.
bantu help bantU-w-an mbantU-w-o
dadl become dadl-y-an ndadl-y-o
dj£r0 deep djj2r0-w-an ndJjrfrO-w-o
gawE Job, task gawE-y-an nggawE-y-o
3.3.3.2.2.H The above do not represent the only places
In the language where Glide Insertion operates; In certain
cases, h-Deletion feeds Glide Insertion by deleting a stem-
final consonant and thus creating vowel sequences:
stem gloss Subst. Imper.
llrih low ngllrl-y-o
lUngguh sit lUnggU-w-an nglUnggU-w-o
pllih choose plll-y-an mlll-y-o
3.3.3.2.3 Returning now to the discussion of h-Deletion
and the variant forms which sometimes occur, one can begin to
see the role which may be played by Glide Insertion In pro
ducing less acceptable variant pronunciations of stems in
final -eh and -oh, such as anE-y-E and kOkO-w-E, instead of
the more correct ane-E and koko-E . By comparing the derivation
of the standard pronunciation with that of the variants, it
can be seen that, while the ordering of h-Deletion after the
137
rule of MVL 1 results in vowel sequences which are not eligible

to undergo Glide Insertion, the reverse ordering always feeds
the latter, so that all the variant forms shew the effects
of Glide Insertion. There is no doubt that, in the fast infor
mal speech where variants are likely to occur, vowel sequences
broken up by glides are much easier for speakers to pronounce.
3.4 Based on the various discussions in this chapter,
the following phonological rules affecting consonant segments
may be posited for the grammar of Javanese:
Consonant Neutralization Sect. 3«1
II
3.3.2
f»
3.3.3.2.2.1
Of these, only h-Deletion and Glide Insertion are ordered
with respect to each other. These two rules must also be
ordered with respect to Mid Vowel Laxing 1, applying after
the laxing of mid vowels in stem-final closed syllables has
taken place.
3.5 It will also be helpful to include in this chapter
on consonant phonology a brief discussion of the processes
involved in producing the various surface realizations of the
nasal prefix commonly used to mark Simple Active verb forms
and, in conjunction with certain suffixes, other verbal forms.
Although it is not altogether clear whether all the processes
involved are, strictly speaking, part of the morphological or
phonological component of the grammar of Javanese, analysis
of nasal prefixation will be important to later discussions
138
(cf. chapter 6) and would seem to be Just as appropriate here

as anywhere else.
As was mentioned briefly In chapter 1, the general effect
of nasal prefixation on the surface appears to be the replace
ment of a basic voiceless stem-initial consonant with its
homorganic nasal, and the addition of a homorganic nasal to
stems with a basic initial voiced consonant.^ With vowel-
initial stems or those beginning with a liquid, the prefix is
realized as ng- added to the basic robt:
stem gloss Act,
kel^m become submerged ngel^m
pangan food mangan
tjampur mix, mixed nj ampur
totoq back of the hand notoq
tUlis write nUlis
sllih borrow njIlih
adj ar learn ngadjar

bakar roast mbakar
djogo guard, watch ndj ogo
dUduq place, position ndUduq
0 0 • 0
findoq
0
egg ngj&idoq
godok boil nggodok
o 9
IbU mother nglbU

lUngguh sit nglUngguh
OmbE drink ngOmbE
139
3.5.1 Because the realization of the nasal prefix

before vowel-initial stems would presumably reflect the basic
prefix in its unaltered form, I will assume that the nasal
prefix is added to the basic stem as /ng-/. For convenience,
the whole process by which nasal prefixed forms of all types
are produced will be referred to as ng-Preflxation. Notice
that, for vowel- and liquid-initial stems, the addition of
the /ng-/ to the stem is all that appears to be involved
in deriving the verbal forms in question. For other types
of stems, the addition of the prefix must either be accompanied
or followed by certain alterations in the shape of the prefix
and the initial consonant of the basic stem.
3.5.1.1 These alterations may be characterized as
probably involving, first of all, assimilation of the /ng-/
prefix to a stem-initial consonant. This Nasal Assimilation
process is necessary to account for the fact that the basic
prefix /ng-/ shows up unaltered only with stems which have
initial vowels, liquids, or velar consonants; in all other
cases the nasal is realized on the surface as m, n, etc.,
depending on the point of articulation of the stem-initial
consonant.
3.5.1.2 Secondly, there must be some process by which
a stem-initial voiceless consonant is either replaced by its
homorganic nasal or is deleted when a homorganic nasal replaces
it. The operation of this replacement or deletion process
with stems in initial voiceless consonants is necessary to
1H0
account for the fact that with stems like pangan, for example,
the initial voiceless consonant of the stem does not occur
at all in the ng-Prefixed form mangan, while in forms based
on stems with voiced initial consonants, such as mbakar, from
bakar, the stem-initial consonant as well as the nasal are
reflected (cf. 3.5.3) on the surface. Because, for all prac
tical purposes, the nasal does appear to replace the initial
voiceless consonant of a stem (cf. chapter 6) and to be in
no way separable from the stem once this has occurred, I have
chosen to characterize this relationship in the most direct
way possible and will refer to the process as Replacement.
In general, it would seem that this is as much as one
can say with any certainty about the proper analysis of the
process of ng-Preflxatlon. There is no non-arbitrary way
to decide whether addition of the prefix, Nasal Assimilation,
and Replacement should properly be treated as a single unitary
process, or whether ng-Prefixation should be treated as the
result of the operation of two or three separate processes
within the grammar. At the same time, there is no way to
tell whether all processes involved in producing ng-Prefixed
forms are part of the morphological marking system of the
language, or whether such forms are in part derived by phono
logical processes.
3.5.2 It is worth noting, however, that just in case the
process referred to above as Replacement (of stem-initial
voiceless consonants with their homorganic nasals) is in fact
m
phonological, its environment would have to be stated specif

ically in terms of interaction between the /ng-/ of the nasal
prefix and a stem-initial consonant; the replacement could
not be allowed to take place freely whenever a voiceless
consonant and its homorganic consonant come together over a
morpheme boundary. Thus, one finds numerous forms like
Poss.l klntang-kU 'my betel chew' or rangsang-kU 'my desire’,
where Recplacement has obviously not taken place. Compare
these with Act. ngel/m from kel^m 'submerge', which at some
stage involves intermediate /ng-kel^m/.
3.5.3 In summary, then,..it will be assumed that, in
general terms, the following processes are responsible for
deriving surface ng-Prefixed forms: Nasal Prefixation; Nasal
Assimilation; and Replacement. Their operation may be seen
in the following sample derivations:
/adj ar/ /pangan/ /bakar/
Nasal Prefixation ng-adjar ng-pangan ng-bakar
Nasal Assimilation --------- m-pangan m-bakar
Replacement mangan ----
other rules ngadjar mangan mbakar
While it is somewhat unsatisfactory to have to assign an
analysis which is so general and devoid of details to a deriv
ational process which occurs as frequently in the language as
ng-Prefixation, there is simply no evidence available on which
to base a more detailed treatment. At the same time, I am con
vinced that the suggested analysis Is correct, as far as it
142
goes. The situation is also somewhat tempered by the fact

that the analysis which has been suggested is sufficient for
the purposes of later discussions (chapter 6) involving
ng-Prefixed forms, with none of the unanswered questions
having, to the best of my knowledge, any crucial bearing on
those discussions.
3.5.4 There Is one other small point concerning ng-
Prefixation of stems beginning with a voiced consonant which
deserves mention. It will be recalled that in such derived
forms, the (assimilated) nasal occurs on the surface along
with the Initial consonant of the stem, as In, for example,
mbakar. There is some question whether this combination of
nasal and obstruent is in actuality realized on the surface
as a cluster or as a nasalized version of the obstruent.
Although there are many arguments which might be brought to
bear on this point, and, Indeed, on whether Internal sequences
of a nasal followed by its homorganic voiced obstruent are
actually pronounced as clusters or nasalized obstruents, none
of the arguments is very compelling. At the same time, the
problem is, while interesting, nevertheless a very minor one
which is of little consequence for other discussions contained
In this dissertation.
Consequently, I will not go into all the arguments for
one view or the other, but will simply follow Horne (1961) and
assume that the mb of words like mbakar Is Indeed pronounced
as a nasalized labial obstruent, and not as a cluster. The
143
adoption of such a view necessitates adding some type of rule

to the grammar which converts sequences consisting of a voiced
obstruent preceded by its homorganic nasal, such as those
which arise in the course of n^-Prefixation of stems beginning
with voiced obstruents, into the nasalized version of the
obstruent. For ease of reference, this process will be re
ferred to as Voiced Consonant Nasalization.
144
FOOTNOTES
^There exists one surface exception to this environment.
In Causative forms, before the suffixes -(g)akE, -(q)no, and
-(q)nE, only light consonants occur in what is, on the surface,
an intervocalic environment. See chapter 4 for a thorough
discussion of Causatives, including their surface excep
tionality to Con-Neut.
2
For a very thorough and scholarly description of the
distribution of consonants, as well as vowels, in Javanese
morphemes, along with their frequency of occurrence, the reader
is referred to Uhlenbeck 1949.
•3
JThere do exist a very few stems in the language where
q occurs internally. Most of these, however, may apparently
be treated as contractions composed of two syllables taken,
one each, from two primary stems. Thus, one finds daqwo
’of different, disparate lengths', which is cited by Horne
(1974) as being from tjg^ndaq ’short' and dowo 'long'. It is
not unthinkable that such*forms must be treated as containing
some sort of internal word boundary, so that £ would indeed
be in stem-final position.
^My informant does not form Elatives at all from stems
whose final vowel Is
Âlthough the tense E In this word violates MVL 2, this
is explainable. See chapter 5 for discussion of the vowel
realization rules and Elatives.
^The source of the intervocalic glides in forms like
these will be discussed below in 3.3.3.2.2.1.
^The single environment where surface h occurs before a
vowel-initial suffix is In Causatives, which will be discussed
in chapted 4.
Ûhlenbeck (1949:214) was well aware of this process
and the formulation which occurs here Is virtually Identical
to his.
9
In certain cases when the language does not have a
homorganic nasal In Its Inventory, the closest nasal Is used.
Thus, for example, since there is no alveolar nasal, £ Is
replaced by the palatal nasal nj. Similarly, alveolar t is
replaced by dental n.
CHAPTER IV
CAUSATIVES AND LOCATIVES
4.0 It was noted in 1.8.3.2.3 that there are many
ways In which Causatives and Locatives behave similarly with
respect to segmental changes accompanying the addition of
the characteristic Causative or Locative suffix to a given
stem. In this chapter, the two processes of Causative and
Locative Formation will be discussed. The reader should
made aware that the aim of this discussion will be mainly to
bring to light some of the issues Involved In providing an
analysis of derived forms belonging to these categories.
No one integrated analysis will be brought forward as the most
desirable way of treating the various aspects of Causative
and Locative Formation, although, along the way, several alter
natives will be examined and certain approaches will be singled
out as being more desirable than others.
4.1 The great similarity between Causative and Locative
Formation may be seen by comparing the Simple Causative and
Simple Locative forms of the following vowel-final stems:
stem gloss Caus. Loc.
dj^rO deep ndj^ro-qakE ndj^ro-nl
gawE job, task nggawe-qakE nggawe-nl
g^ntl change ngg^nte-qakE ngg^nte-nl
klro thought nglra-qakE nglra-nl
lakU walk nglako-qakE nglako-ml
146
Notice that, In Causatives as well as In Locatives, addition

of the characteristic suffix to the stem Is accomplished by a
change in the stem-final vowel. One such change, exemplified
by klro; Caus. nglra-qakE; Loc. nglra-nl, will not really be
of concern to the present discussion, since it is not a phenom
enon peculiar to the derived forms in question. Instead, it
may be attributed to the regular operation of the phonological
rule of /a/ to off; precisely the same change occurs when any
suffix (except Imp. -o, as discussed in 2.9.1.1) is added to
a stem ending in /a/, destroying the word-final environment
necessary for the change from basic /a/ to surface o.
4.1.1 Thus, eliminating changes resulting from the oper
ation of /a/ to off, one may summarize the relationships between
primary stem-final vowels and their realizations in Causatives
and Locatives as follows:
primary root vowel Caus. and Loc.
0
4.1.2 These changes are of a somewhat different nature
from that attributed above to the operation of /a/ to off in
that they occur only In specific morphological environments:
namely, in Causatives and Locatives, with all other productive
types of derivation in the language involving no change in
stem-final vowels. It should be noted that identical changes
147
also appear in certain forms which were originally derived

by suffixation of -an (cf. 1.8.3.3.2). However, in the modern
language, the vowel changes in question no longer appear to
be a productive part of -an suffixation, with -an now being
added directly to all types of stems with no change in a stem-
final vowel, except of course those in -o§ from basic /a/.
4.1.3 That the vowel changes which occur in the formation
of Causatives and Locatives no longer occur elsewhere as part
of the productive -an suffixation process is evidenced by the
fact that, invariably, when given a stem and asked to form
its -an derivative, my informant will only produce forms with
the suffix added directly to the unaltered stem. If asked
about specific -an forms with accompanying vowel change, such
as Subst. klEron, from klerU 'mistaken', he will readily accept
them as good words, but does not voluntarily seem to produce
them as -an derivatives.
At the same time, it seems very likely on semantic grounds
that many -an derivatives with accompanying vowel change are
not actually derived synchronically, but are instead stored in
the lexicon as unitary items. This view is supported by the
fact that a great many -an derivatives with vowel changes
have meanings which would represent such extreme extensions
of their (original) basic stems that, semanticily, they may
reasonably be treated as Independent lexical Items (e.g. banJU
'water'; banjon 'specially prepared water for blackening the
teeth').
148
Finally, there appears to be a strong tendency among

Javanese speakers to produce over-marked forms such as
mElonan, alongside mElcm, from melU 'accompany’, when -an
has apparently been added to a stem with accompanying vowel
change. This could reasonably be attributed to a desire to
regularize the marking of -an forms so that the suffix always
occurs in its full form, instead of Just -n, as might appear
to be the case in derived forms like mElon; or, alternatively,
it might Just as easily be taken as an indication that -an
is beginning to be treated like the variable suffixes, with
an initial -(n) which shows up only after vowel-final stems.
In addition, however, the occurrence of over-marked forms
like mElonan could also, Just as reasonably, be an indication
that speakers no longer feel forms like mElon to have under
gone the productive process of -an suffixatlon and are simply
■\
adding the suffix -an to such forms to produce over-marked

mElon-an. The latter interpretation of the facts would of
course support the contention that the vowel changes outlined
in 4.1.1 above are no longer felt to be a part of the produc-
2
tiv.e -an suf fixation process of the language.
4.2 On the strength of the above arguments, I am con
vinced that the vowel changes in question may thus be treated
as being limited to Locatives and Causatives built on vowel-
final stems. At the same time, it seems clear from a general
examination of derived forms in the language that the environ
ment for these changes is not phonological in nature, but
l!»9
must somehow be defined solely in terms of the two categories

Causative and Locative. There are three possible ways to
look at such a process within a generative framework: (1) It
might be considered to be a part of the phonological component
of the grammar, but different from other phonological pro
cesses in that its environment is morphologically defined;
(2) It might be considered to be a part of the marking sys
tem for the morphological processes of Causative and Locative
Formation, and thus on a par with the addition of the approp
riate suffix for each category; or, finally, (3) It might be
considered to be a stem-forming process which takes place as
a sort of preparation for the addition of Causative or Loc
ative suffixes to vowel-final stems.
4.2.1 Admittedly, the difference between the above
means of characterizing the changes in stem-final vowels
exhibited by Causatives and Locatives is not great. Essen
tially, it appears to be a choice between claiming the vowel
changes in question to be the result of a process which is
separate from the morphological derivational apparatus of
the language (as in Choice (1) above) or claiming it to be
an integral part of the morphology (Choices (2) and (3))*
If the former view is taken, then a number of more sub
tle claims are made. First of all, the claim would be made
that the process responsible for altering stem-final vowels
could, but of course does not necessarily have to, apply
after other phonological processes of the language. Secondly,
150
Just like other phonological processes, such a rule might

be expected to be subject to possible loss or generalization,
which would be independent of the morphological processes of
Causative and Locative Formation themselves.
On the other hand, if one chooses to regard the process
in question as being in some way a part of the morphological
marking of Causatives and Locatives, quite different claims
and predictions would be made. Such a characterization would
probably rule out the application of the vowel changing pro
cess anywhere but at the same stage of the derivation at
which the other processes involved in Causative and Locative
Formation (i.e. addition of the characteristic suffixes) apply,
and although this is not necessarily always the case for
derivational processes, there is no reason to believe that
Causative and Locative Formation do not take place before
the rules of the phonological component come into play.
Similarly, taking this approach would make the claim that
speakers feel the vowel change to be an essential part of
the Causative and Locative marking process, so that, it
would seem, the loss or generalization of the change might
be much less likely than would be the case if it were accom
plished by a phonological rule.
4.2.2 In general, on the basis of the attitude of my
informant toward Causatives and Locatives, I would lean
toward taking the view that the vowel change is not a phono
logical process. There is little doubt that my informant
151
feels a real difference to exist between the phonological

rules discussed in chapter 2 and the stem-final vowel change
which occurs in Causatives and Locatives. This difference
is evidenced by his rather casual attitude toward the correct
application of the phonological rules on the one hand, and
his steadfast insistence on the appropriate vowel change in
Causative and Locative forms on the other. That is, when
given a pronunciation in which any one of the phonological
rules is violated, he is likely to say that the word would
be "acceptable" or "understood", even though the pronuncia
tion might not be entirely correct; but when given a Caus
ative form such as *ngg£ntI-qakE or *nggg*nt 1-qakE, from
g#ntl 'change', where the necessary stem-final vowel change
has not taken place, he is likely to reject such words, saying
that they might be spoken by a little child who has not yet
learned good Javanese.
This difference in attitude may be seen especially
clearly in my informant's evaluation of forms like
*ndjoro-qakE and *ndjoro-nI, from djoro (/djara/) "drill1
on the one hand, and those like *nUkU-qakE or anUku-qakE
and *nUkU-nI or anUku-nI, from tUkU "buy", on the other.
While all of these Simple Causatives and Locatives are deviant
to some degree, the former represent violations of the gen
eral phonological process of /a/ to qji, and the latter have
not undergone the vowel changes described in 4.1.1, which
are limited to Causatives and Locatives. It is significant
152
that my informant will accept forms like »ndJoro-qakE

(instead of correct ndjara-qakE) as being "possible” (but not
"correct”) but completely rejects those like *nUkU~qakE
(instead of correct nUko-qakE) as being simply "wrong".
On the basis of such judgements by a native speaker, it
seems likely that the vowel change in question is felt to
be a more necessary part of derived Causatives and Locatives
than would be implied by treating it as a phonological pro-'
cess. Rather, my informant’s attitude toward the vowel
changes would seem to indicate that one of the two morpho
logical approaches suggested in 4.2 above should be chosen.
4.3 Recall that the two morphological analyses Involved
(1) treating the vowel change as being on a par with the addL
tion of the appropriate suffix as a marker of Causatives and
Locatives, or (2) treating the vowel change as serving to
create a Causative or Locative stem to which the proper suf
fixes would be added. It is of course very difficult, if
not impossible, to choose between these two very similar
approaches; however, it seems to me that there is some moti
vation for choosing the second over the first.
i
4.3.1 Although so far in this discussion examples have

been given only of the Simple Causative and Locative forms,
it has been mentioned that there are other Causative and
Locative forms in which the same vowel changes occur, but
which are marked by different suffixes: one also finds
Causative Imperatives, marked with -(q)no3 and Causative
153
Optatives, marked with -(q)nE; similarly, there are Locative

Imperatives and Locative Optatives, marked with -(n)ono and
-(n)anE, r e s p e c t i v e l y .3 if the change of stem-final vowels
described above were to be treated as a marking process which
is on a par with the addition of the proper suffix for each
type of Causative or each type of Locative, the morphological
rule responsible for the vowel change would presumably have
to be stated in the grammar six different times, once for
each different type of Causative or Locative.
4.3.2 On the other hand, if the vowel change were
treated as a more general Causative and Locative stem-forming
process, whose output is a marked stem to which any of the
Causative or Locative suffixes may be added, the vowel change
would need to be stated only once. A great deal of redundancy
would thus be avoided, and, at the same time, the generaliza
tion that the same type of stem-final vowel change takes
place in all Causative and Locative derived forms would be
captured.
4.3.3 There is further evidence indicating that, in
addition to the change of stem-final vowels, a bit more may
be involved in the proposed stem-forming process. Consider
the following Causative and Locative forms, where doubling
has taken place, Indicating plurality of actions or actors:
stem gloss doubled
dj&rO deep ndj^roq-ndj^ro-qakE
ndjjgron-ndjj^ro-nl
154
gawE job, task nggaweq-nggawe-qakE

nggawen-nggawe-nl
g^ntl change ngg£nteq-ngg£nte-qakE
ngg£nten-ngg/£nte-nl
lakU walk nglakoq-nglako-qakE
nglakon-nglako-nE
Notice that in the doubled Causative and Locative forms
the doubling process has apparently copied not only the stems
(with the proper changes in stem-final vowels), but also
the initial consonants of the suffixes. It is striking that
Causatives and Locatives are the only derived forms in the
language where Doubling involves copying this suffix-initial
consonant, so that, even though the Locative suffixes have
initial -(n), Just like other variable suffixes such as
Demonstrative -(n)E, there is a divergence in the treatment
of this n in doubled forms: compare Dem. bOdO-bOdO-nE with
• •
Loc. mbodon-mbodo-nl, both from bOdO ’stupid'.

• •
4.3.3.1 Although Causatives and Locatives are the only

forms where Doubling appears to copy a part of the suffix
along with a vowel-final stem, there is evidence throughout
the language Indicating a strong tendency to keep the two
members of doubled words identical (cf. chapter 6 for a thor
ough discussion of this tendency) which might possibly be
related to the peculiar operation of Doubling in Causatives
and Locatives. Thus, for example, one finds doubled forms
like plll-plll-y-an from pllih ’
'choose9 and njala-njala-I
155
from salah 'make a mistake'. In both cases, the stem-final

h has been deleted in the second member of the doubled pair
because it occurred in the intervocalic environment for
h-Deletion (cf. 3.3.2); in the first member of the doubled
form, however, there was never any chance for h to become
intervocalic, and yet It has been deleted. As will be shown
in chapter 6, the best explanation for this deletion Is that
the grammar of the language regulates the application of
various processes in order to maintain Identity between both
members of the doubled pair.
M.3.3.2 It is possible to explain the copying of the
suffix-initial consonant In the first member of doubled
Causatives and Locatives as simply being another manifestation
of this same desire on the part of Javanese to maintain
identity In doubled stems if one claims that the morphological
Causative and Locative stem-forming process spoken of earlier
involves, over and above the change of stem-final vowels, the
addition of a characteristic consonant to such stems: g for
Causatives and n for Locatives. Taking this tack, at the
time when Doubling took place, the Causative or Locative
version of an original vowel-final stem would differ from
the basic version of the stem in two ways: it would have
undergone the appropriate change of Its final vowel; and it
would have acquired a characteristic Causative (g) or Locative
(n) marker as Its final consonant. The process of Doubling
would then simply copy the derived stem, complete with vowel
156
and marker consonant.**

If one does not accept the occurrence of and n In the
first member of doubled Causatives and Locatives as being a
result of the fact that they, like the changed stem-final
vowels, are felt to be part of the characteristic Causative
and Locative stem, then I can see no other reasonable way
to account for why doubled forms belonging to these two
particular derived categories are the only ones which have
what has been characterized as the initial consonant of the
suffix (which should perhaps not be treated as part of the
suffix at all) in both members. That is, there would seem
to be no other reason for the existence of forms like Loc.
nggawen-nggawe-nl with n in both members, but others like
Dem. gawE-gawE-nE, with the only occurrence of n being in
the suffix.
4.3.4 Note, however, that if this explanation is accep
ted, then the original concept of stem-forming suggested in
4.3.2 is inadequate. There it was suggested that the process
responsible for changing stem-final vowels could simply be
stated once, operating to form a general derived stem to
which any Causative or Locative suffix might be added. If,
however, a characteristic consonant must also be added to
the stem, then at least one, and perhaps two, additional
stem-forming processes are needed, depending on how one
chooses to treat the consonant-adding process. There appear
to be two plausible ways to characterize the stem-forming
157
process in question.
^.3.^.1 One alternative would be to characterize the
addition of the stem-forming consonants and n to vowel-
final stems as being accomplished by a process which does
simply that: it adds a characteristic to the altered
vowel-final stem to form a unique Causative stem and an n
to prepare an appropriate Locative stem. The individual
suffixes would then be added to these stems.
4.3.2 Adopting such an analysis for vowel-final stems
has certain consequences when one moves on to consider Caus
ative and Locative formations built on stems ending in con
sonants. In general, for all variable suffixes in the
language other than those used in Causatives and Locatives,
the simplest and most straightforward way to account for
the fact that such suffixes appear to have an Initial con
sonant which occurs on the surface with a vowel-final stem,
but not with a consonant-final stem, is to assume that when
the suffix is added, It Is In its fullest (i.e. consonant-
initial) form. An early rule of the grammar would then
apply to delete this suffix consonant (which is always n
except in Causatives) just in case it is preceded by a stem-
final consonant. Whether this deletion rule were treated as
a phonological process or a morphological process would have
little consequence for the present discussion.
^.3*^03 Notice, however, that if one accepts the anal
ysis suggested above for Causatives and Locatives of vowel-
158
final stems, with a consonant (c[ for Causatives and n for

Locatives) being added as a stem-formant to stems in final
vowels, it is not necessary to assume that the Causative
and Locative suffixes, which have been cited so far as
(optional) consonant-initial -(q)akE, -(q)no, -(q)nE and
~(n)I, -(n)ono, and -(n)anE, ever, at any stage of the
derivation, occur in their full or n-initial forms. That
is, the optional suffix consonant which occurs only with
vowel-final stems would, taking the approach outlined above,
be added by a rule to the stem itself, and thus would not
ever have to be treated as an actual part of the suffix.
The same consonant never shows up at all with consonant-
final stems. Thus, it would seem that there would be abso- -
lutely no reason to assume that the basic Causative and
Locative suffixes actually have the full forms cited earlier,
but, instead, one could characterize them as /-akE/, /-na/,
and /-nE/ for Causatives and /-I/, /-ana/, and /-E/ for
Locatives, with the characteristic consonants and n being
part of the Causative and Locative stems and not part of the
suffixes.
If this approach were taken, although Causative and Loc
ative suffixes might appear on the surface to act (in non
doubled forms) like all other variable suffixes in the
language, there would be a difference in the course of their
derivation. While Causatives and Locatives would involve
the addition of a stem-forming j or n to vowel-final stems
159
and subsequent addition of the appropriate suffixes /-akE/,

/-na/, /-nE/, or /-I/, /-ana/, or /-E/ to all types of stems,
all other surface forms with variable suffixes would probably
be derived by means of addition of the suffixes in their full
n-initial forms (e.g. /-nE/, /-n£n/, etc.) to all stems, and
subsequent deletion of the n after all stem-final consonants.
This' difference may be seen by comparing the following sample
derivations, the first for Causatives, the second for Demon
strative forms:
/dj^rO/ /djUpUq/
Vowel Change dj^ro ----
Causative Stem-Forming dj^roq ----
Suffixation dj/groq-akE djUpUq-akE
other rules ndj^roq-akE ndjUpuq-akE
/djf*rO/ /djUpUq/
Suffixation dj£rO-nE dJUpUq-nE
Suffix-Consonant Deletion ---- djUpUq-E
other rules dj/rO-nE dJUpUq-E
4.3.4.^ An alternative analysis of the addition of the
stem-forming consonants and n to vowel-final stems would
differ from that just discussed in that, instead of simply
adding c[ and n from out of nowhere, as it were, the appropri
ate Causative or Locative suffixes could be added to the stem
in their fullest forms (i.e. /-qakE/, /-nl/, etc.) and a rule
of stem-forming would then apply to consolidate the initial
160
consonant of the suffix with the stem. Note that this

analysis allows one to treat Causative and Locative suffixes
as being added in their fullest forms to vowel-final and
consonant-final stems, Just as one would probably want to
treat all other variable suffixes in the language, with a
subsequent rule deleting the initial consonant of all such
suffixes Just in case they occur on consonant-final stems:
/djUpUq/ /dJ^rO/ /djUpUq/ /dj*r0/
Vowel Change dj^ro ---- ----
Suffixation dJUpUq-qakE dJ^ro-qakE dJUpUq-nE dj£rO-nE
Caus. and Loc. ---- dj£roq-akE ---- ----
Consolidation
Suffix-Consonant dJUpUq-akE ----- dJUpUq-E -
Deletion
other rules ndJUpuq-akE ndJ^roq-akE dJUpUq-E dJgfrO-nE
On the other hand, this analysis seems to be much less
general than its alternative discussed above in that it'
implies that speakers have to memorize the as part of each
Causative suffix and the n as part of each Locative suffix,
when it would seem more economical to simply learn a rule
which adds £ as a stem-formant for Causatives and n as a
formant for Locatives and then add the individual suffixes
to these prepared stems.
In spite of this possible disadvantage arising with
vowel-final stems, when one brings into consideration Causa
tives and Locatives built on consonant-final stems, there is
some evidence to Indicate that the second alternative (i.e.
161
addition of the Causative and Locative suffixes in their

consonant-initial form and subsequent consolidation) could
be the correct one; however, as will be seen, even this
evidence is open to an alternative interpretation which
ends up shedding little light on how the c[ and n of Causa
tives and Locatives should ultimately be treated.
So far, it has been possible to speak of Causa
tive and Locative derived forms together, since in all
respects discussed so far, their behavior has been identical.
However, when one looks closely at the surface pronunciation
of Causatives and Locatives built on stems ending in con
sonants, a number of differences become evident. These
differences involve the apparently aberrant behavior of
certain phonological rules in Simple Causatives, while
Locatives (and also Causative Imperatives and Causative
Optatives) exhibit, like all other forms in the language,
strictly regular behavior of these rules. Note that, through
out the following discussion, unless otherwise indicated, I
will be talking exclusively about Simple Causatives, formed
with the suffix -(q )akE. It should be emphasized that it
is only in these forms that the aberrant behavior of phono
logical rules shows up; in Causative Imperatives, marked with
-(q)no«, and Causative Optatives, which have the suffix -(q)nE3
the operation of the phonological rules is perfectly regular
and transparent.
4.4.1 Consider the following examples of Causatives and
162
Locatives built on stems with final /-IC/ or /-UC/:

stem gloss derived
b£tJ iq good mb£tJiq-qkE
mb^tjIq-I
dJUpuq go get ndjUpuq-akE
ndjUpUq-I
Etung arithmetic ngEtung-akE
ngEtUng-I
lUlus pass (an exam) nglUlus-akE
nglUlUs-I
raup wash the face ngraup-akE
ngraUp-I
tUlis write nUlis-akE
nUlIs-I
In these examples, the final-syllable high vowel of the
primary stem is lax as a result of High Vowel Laxing, which,
it will be recalled from 2.4.2.5, laxes high vowels in
closed syllables. On the other hand, this same vowel Is
tense in Locative forms, since, as was argued in 2. *1.2.3,
the presence of the Locative suffix -(n)I, which would have
the shape -!I at the point at which HVL applies, has the
effect of opening up the stem-final syllable so that the
environment for HVL is not met. One would also expect the
Simple Causative forms, like the Locatives, to have tense
high vowels In their stem-final syllables, since the syl
labification would appear to be the same in both types of
163
derived forms; nevertheless, all Simple Causatives built on

stems in final /-IC/ or /-UC/ have lax high vowels in this
syllable on the surface.^
4.4.2 Similarly, Causatives and Locatives behave dif
ferently with respect to the rule of Consonant Neutralization.
Recall that this rule changes heavy consonants to light
word-finally or before another consonant (cf. 3.1). Thus, in
the following examples of stems in basic final heavy conso
nants, as would be expected, the primary stems are all
pronounced with light (word-final) consonants, while the
Locatives have stem-final heavy consonants, the latter occur
ring because Con-Neut does not operate in intervocalic envi=-
ronment s :
stem gloss derived
7
godok boil nggodok-akE
• •
nggodog-I
rantap flake off ngrantap-akE
ngrantab-I
wUdjut shape, form mUdJut-akE
mUdjUd-I
In Simple Causatives, however, contrary to what would be
expected in the surface intervocalic environment which is
in all appearances identical to the environment occurring
in Locatives, the light variants of basic heavy stem-final
consonants occur.
4.4.3 Finally, Causatives exhibit unexpected behavior
I6i|
with respect to the rule of h-Deletion (3.3.2) which regu

larly deletes h intervocalically:
stem gloss derived
adoh far ngadoh-akE
ngado-I
akeh much, many ngakeh-akE
ngake-I
lUngguh sit nglUngguh-akE
nglUnggU-w-I
pllih choose mllih-akE
mlll-I
salah mistaken njalah-akE

njala-I
In the above examples, just as with the other rules discussed
so far, Locatives exhibit regular behavior of the rule in
question, while even though Simple Causatives present pre
cisely the same phonological surface environment, the rule
does not appear to have operated in forms of the latter
category.
In summary, then, it seems that Simple Causatives are
distinct from Locatives and from all other derived forms in
the language by virtue of their behavior with respect to
three phonological rules: High Vowel Laxing; Consonant
Neutralization; and h-Deletion. What is especially striking
is that in every case the surface phonological environment
provided by Simple Causatives and Locatives Is identical, so
165
that one would normally expect these two types of derived

forms, whose behavior Is absolutely parallel In forms built
on vowel-final stems, to exhibit parallel behavior with
respect to the phonological processes of the language in
forms built on consonant-final stem3.
4.5 If this apparently aberrant behavior of the phono
logical rules in Simple Causatives could be shown to be
totally random and without pattern or motivation, one might
reasonably adopt some sort of formal analysis such as marking
Simple Causatives as positive or negative exceptions to the
appropriate rules or re-defining the environments of the
relevant rules to be sensitive to the catetory of Simple
Causatives. These approaches do not, however, appear to be
warranted in this case, since it Is not difficult to find
a pattern behind the apparently Irregular behavior of HVL,
Con-Neut, and h-Deletion in Causatives.
4.5.1 Consider the operation of these three phono
logical rules and the behavior of Simple Causatives with
respect to each: HVL operates in closed syllables and in
Simple Causatives; Con-Neut operates on heavy consonants
word-finally, before another consonant, and. in, Simple Causa
tives; and the process of h-Deletion operates intervocalic^
ally except In Causatives. In each case, it seems that the
phonological rules are applying or not applying, as the case
may be, just as they would If there were another consonant
after the final consonant of the stem. If such a segment
166
were present, HVL would be expected to apply, since the

final syllable of the stem would then be closed, just as
it is in derived forms such as Poss.l aplq-kU, from aplq-
’good, nice’. Similarly, the presence of another consonant
after the final consonant of the stem would account for the
otherwise unexpected application of Con-Neut in Simple Caus
atives, since one of the environments for the rule consists
of a heavy consonant followed by another consonant. Finally,
the failure of h-Deletion to apply would also be explained,
since the latter rule operates only in intervocalic environ
ments. Thus, it seems that an analysis positing a consonant
which occurs after the final consonant of the stem at the
point in the derivation where the relevant phonological rules
apply, but which is somehow deleted before the word is real
ized on the surface, would allow one to treat.the behavior of
the phonological rules in Simple Causatives as being entirely
regular.
4.5.2 It is not difficult to see how such an analysis
could be tied in with other facts about Causatives on the
one hand and all other derived forms with variable suffixes
(including Locatives) on the other. It would first have to
be assumed that all variable suffixes, Including the Causa
tive suffixes, are added to stems in their fullest consonant-
initial form. A morphological or very early phonological
rule would then apply to delete the initial n from all
suffixes occurring on consonant-final stems, before the
167
phonological rules of HVL, Con-Neut, and h-Deletion apply.

The Causative suffixes -(q)akE, -(q)no, and -(q )nE, being
the only variable suffixes In the language which do not
begin with n, would still have their initial consonants
when the relevant phonological rules apply. Thus, HVL,
Con-Neut, and h-Deletion would be prefectly regular in their
application, with the £ of the Causative suffixes pro'ucing
the necessary environments for the operation of HVL and
Con-Neut, but guaranteeing that the intervocalic environ
ment required for h-Deletion will not be met. Subsequently,
a rule Just like that which deleted the n of other variable
suffixes (including Locatives) would delete the from all
Causative suffixes.
4.5.2.1 Using this analysis, the derivation of Simple
Causatives and Locatives built on stems with final consonants
would proceed as follows:
/wUdjUd/ /wUdJUd/ /lUnggUh/ /lUnggUh/
Suffixation wUdJUd-qakE wUdjUd-nl lUnggUh-qakE lUnggUh-nl
n-Deletion ---- wUdJUd-I --------- lUnggUh-I
HVL wUdjud-qakE ---- lUngguh-qakE ----
Con-Neut wUdjut-qakE --- — ---- ----
h-Deletion -------------------- l
g-Deletion wUdjut-akE ----- lUngguh-akE -
other rules mUdJut-akE mUdjUd-I nglUngguh-akE nglUnggU-w-I
^.5.2.2 The suggested analysis appears to have some
advantages in that it not only neatly explains the strange
168
behavior of a number of phonological rules In Simple Causa

tives built on consonant-final stems, but it also accounts
for why this strange behavior should show up only In Causa
tives and not in Locatives as well, In spite of the fact
that these two types of derived forms exhibit parallel
behavior when built on vowel-final stems. That is, the
suggested analysis attributes the different behavior of the
phonological rules in Simple Causatives versus Locatives to
a difference in the stages of derivations at which the two
rules deleting suffix-initial consonants apply. Locative
suffixes, which begin with n, lose their initial consonants
before the phonological rules apply; Causatives, with initial
do not lose this consonant until after HVL, Con-Neut, and
h-Deletion have applied.
4.5.2.3 Furthermore, the suggested analysis allows
one to decide between the two alternative sources for the
stem-final ^ and n which show up in Causatives and Locatives
built on vowel-final stems. That is, if the above analysis
for Causatives and Locatives built on consonant-final stems
Is accepted, it must be assumed that the Causative and Loca
tive suffixes are added to consonant-final stems In their
fullest forms (i.e. with c[ and n present). But this is really
consistent only with the analysis given in above
(I.e. that which assumes the and n which appear with origi
nally vowel-final roots to have been, in effect, taken from
the Causative and Locative suffixes and consolidated with the
169
original stem). Although It would be possible to assume

that stem-forming q and n are added to consonant-final stems
as well as to vowel-final stems, this would make little
sense: since the stem-forming consonants would always
ultimately be deleted with a consonant-final stem, it is
not likely that speakers would ever add them in the first
place. The latter would be ad hoc and in fact inconsistent
with the notion of "stem-forming".
4.5.3 In spite of the fact that the proposed analysis
is fairly plausible and neatly accounts for a number of
otherwise puzzling phenomena, I am nevertheless hesitant to
accept it as it stands as offering the best explanation for
the application of the phonological rules in Simple Causa
tives; while it is a very neat mechanical solution, I am
not sure that it is a very insightful one. There are sev
eral drawbacks to accepting such an analysis.
4.5.3.1 First of all, it is somewhat unlikely that,
while the initial consonant of all other variable suffixes
occurring on consonant-final stems would be deleted at an
early point in the derivation, which is probably what one
would expect to find in most grammars, the initial c[ of
Causatives would be allowed to remain until after h-Dele
tion has applied, the latter being a rather late rule in
the grammar (cf. 3.3.2). It is true, however, that this
unusual retention of the Initial consonant of Causative
suffixes does correlate with the fact that Causatives are
170
unique in beginning with and not n. Thus, the situation

is not nearly as unlikely as it might be if Causatives began
with n, like all other variable suffixes.
5.3.2 Another problem with the proposed analysis is
the obvious opacity of Simple Causatives with respect to
HVL, Con-Neut, and h-Deletion. It strikes me as highly
unlikely that a language would allow to persist an ordering
which produces such blatantly opaque forms; especially when
the ordering is of the unusual type proposed in this analysis.
That is, since the language already possesses an early rule
which deletes suffix-initial n after a consonant-final stem,
it would seem quite reasonable to expect that rule to become
generalized to delete g_ as well— or at least to expect the
g-Deletion rule to be re-ordered to apply at an early stage
of the derivation. Both courses would avoid opaque surface
Causatives. Nevertheless, the derivation of Simple Causa
tives as described for the modern language seems to be quite
stable, to the best of my knowledge not having changed in
the course of at least the last century. While it seems to
be the case that speakers do occasionally tolerate opaque
orderings in grammars, this Is usually only when the opacity-
producing ordering can be seen to have some definite function,
such as preserving the identity of underlying forms or marking
a grammatical category (cf. KIsseberth 1973 and chapter 5 of
this dissertation). And, as it turns out. It is possible to
see some functional advantage which Is gained by the presence
171
of the ordering in question within the grammar of Javanese.

Recall that it was only in Simple Causatives that the
aberrant behavior of HVL, Con-Neut, and h-Deletion was seen
to occur; in both Causative Imperatives and Causative Opta
tives, the application of the rules in question is completely
transparent. Thus, for example, while the Simple Causative
nUlls-akE (from tUlls 'write') has an unexpected lax high
vowel occurring in a stem-final open syllable, the Causative
Imperative nUlls-no also has a lax high vowel in this posi
tion, but the syllable is closed, as it should be for the
regular operation of HVL, by the combination of a stem-final
consonant with the initial consonant of the suffix. Similarly,
one finds Caus. mUdJut-akE (from wUdjut (/wUdjUd/) 'shape,
form'), with Con-Neut apparently having applied irregularly
in an intervocalic environment, but Caus. Imper. mUdJut-no,
with the rule not having applied because of the following
consonant-initial suffix. And, again, h-Deletion, which
ought to have applied in the intervocalic environment of
Caus. nglUngguh-akE (from lUngguh 'sit'), but did not, would
not be expected to apply before the consonant of the Causa
tive Imperative suffix in nglUngguh-no.
In every case, the Simple Causative form has irregularly
either undergone or not undergone the relevant phonological
rules In a manner exactly parallel to the regular derivation
of both of the other types of Causatives (I.e. Causative
Imperatives and Causative Optatives). Thus, Simple Causatives
172
undergo HVL, Just like Causative Imperative and Causative

Optative forms; they undergo Con-Neut, Just like the other
two Causative categories; and they do not undergo h-Deletion,
again, Just like the other two types of Causative forms.
The big difference is of course that in Causative Imper
atives and Causative Optatives, whose suffixes may be cited
as -(q)no and -(a)nE, and which are accordingly consonant-
initial even with consonant-final stems, the operation of
all these rules is quite clearly regular and transparent
on the surface. Notice, incidently, that no such divergence
occurs with the various Locative forms, since all of the
three Locative suffixes -(n)I, -(n)ono, and -(n)E, create
precisely the same type of environment: that is, all of
these suffixes are vowel-initial when occurring with conso
nant-final stems, and the behavior of the phonological rules
is completely regular in all three Locative forms.
4.5.4 Returning now to the ordering analysis suggested
above in 4.5.1, it is not difficult to see the possibility
that the unusual ordering described earlier might in fact
be maintained within the grammar simply because it does
result in Simple Causatives whose stems are identical to
the corresponding Causative Imperative and Causative Optative
stems. (Note' that, although this might be the reason .the
ordering is maintained, no claims are made about how or
why the ordering in question first arose in the grammar.)
173
4.6 There Is, however, another way to think of the

derivation of Simple Causatives with respect to that of
Causative Imperative and Causative Optative forms, which
would make a much stronger claim about the source of their
apparently deviant behavior with respect to the phonological
rules than would be made by the ordering approach already
discussed. I would suggest the possibility that Javanese
speakers may perhaps be applying a sort of synchronic
analogy to the derivation of Simple Causatives built on
consonant-final stems, using the Causative Imperative and
Causative Optative forms as their model. That is, the
irregular behavior of the phonological rules in Simple
Causatives could conceivably be explained as an attempt
on the part of speakers to keep the realization of consonant-
final stems in all three Causative forms identical by
having the stems In Simple Causatives irregularly undergo or
not undergo the same phonological rules which apply or do
not apply regularly In Causative Imperatives and Causative
Optatives.
4.6.1 The desire to maintain identity among the various
forms of paradigms or derivational categories is frequently
9
encountered in languages. Furthermore, In Javanese, the
natural tendency to maintain the identity between the reali
zation of a given consonant-final stem In all three types
of Causatives would probably be strengthened, first of all,
by the fact that vowel-final stems are regularly treated
174
Identically in all three Causatives (e.g. Caus. nggawe-qakE;

Caus. Imper. nggawe-qno; and Caus. Opt. nggawe-qnE, from
gawE ’Job, task').10 Thus, if all rules were allowed to
apply regularly in all Causatives, it would be only in
Simple Causatives built on consonant-final stems where
anomalous forms would occur. Secondly, as was mentioned
earlier, the three realizations of a given consonant-final
stem are identical in all three Locative forms; this could
perhaps influence speakers to strive for the same identity
in all three Causatives, especially in light of the fact
that Causatives and Locatives built on vowel-final stems
undergo parallel derivations.
Thus, I am suggesting that the apparently irregular
behavior of the phonological rules in Simple Causatives
is in fact a result of an attempt, to maintain regularity
of a higher order: that is, regularity or identity between
the realizations of a given consonant-final stem in all
three Causative forms.
4.6.2 If one accepts this motivation for the behavior
of phonological rules with respect to Simple Causatives, .
there are a number of different ways to construct a formal
grammar which could be made to produce the correct surface
forms, for the correct reasons: one could provide Simple
Causatives with features marking them as positive exceptions
to HVL and Con-Neut, and as negative exceptions to h-Dele
tion; or, the environments for these phonological rules could
175
simply be altered to apply or not apply In Simple Causatives,

as would be necessary; or a special boundary could even be
Inserted In Simple Causatives, to which the appropriate
phonological rules could be made sensitive; it is even pos
sible that the approach described earlier using a late rule
to delete the £ of Causative suffixes could somehow be made
to work. But requisite to the adoption of all these possible
formalisms would be the inclusion in the grammar of a state
ment something like the following:
Identity must be maintained between all reali
zations of a stem in the three types of Causatives.
As long as some such statement were included, giving the
functional motivation behind the otherwise apparently aber
rant behavior of phonological rules in Simple Causatives, it
would not seem to make very much difference which formal
device was used to implement this behavior. The essential
principle behind the operation of the grammar would still
be captured.
4.6.3 Note that there is a definite, although subtle,
distinction between taking this tack and simply adopting the
rule ordering approach discussed earlier. Simply assigning
the rules of the grammar the ordering in 4.5.1 and saying
that this ordering Is allowed to persist because It maintains
regularity between surface realizations of the Simple Causa
tive stem and the realization of the same stem in all other
176
Causatives does not necessarily make any claims about what

would happen if, for example, a new rule were added to the
grammar which might disrupt this regularity. On the other
hand, assuming that there is some type of synchronic analogy
at work which always functions to keep Causative stems iden
tical would make the prediction that, regardless of what
types of new rules might be added to the grammar, this reg
ularity would be maintained. The operation of a disruptive
rule would always be controlled within actual derivations
to keep the realization of a given stem in all three types
of Causatives identical.
4.6.4 It should be noted, however, that this direct
functional explanation does leave something to be desired.
One might ask why the language should have generalized the
behavior of the phonological rules In Causative Imperative
and Causative Optative forms, which are of somewhat secondary
importance within the language, to the Simple Causative,
which Is surely much more frequently used, and not the other
way around. One answer to this question might be that the
majority simply has won out. That is, even though the Simple
Causative may be more frequently used, It Is still only one
of three derived Causatives, two of which share the same
Causative stem. It is also conceivable that a better answer
to this question Is to be found in the history of Javanese;
however, for various reasons, I have not been able to learn
the necessary facts about the source and development of the
177
relevant categories to determine whether or not this is the
case.
4.6.5 It is also true, though of secondary importance,

that adopting the analogical approach does not aid one at
all in choosing between the two analyses for the derivation
of Causatives built on vowel-final stems suggested in
4.3.4.3 and 4.3.4.4 above. Indeed, the approach Just des
cribed could be consistent with either treatment of stems
in final vowels, since it does not necessarily hinge in any
way upon whether the a which appears in Causatives built on
vowel-final stems is added as a stem formant, or whether it
is added as a part of the suffix and only later consolidated
with the stem.
178
FOOTNOTES
^■Recall from chapter 1 that In all cases, Causative
and Locative Imperatives and Optatives can be formed by
substituting -(q)no and -(q)nE for Simple Causative -(q)akE
and, similarly, -(n)ono and -(n)anE for Simple Locative -(n)I.
2
By this I do not mean to imply that speakers make
no connection between forms like mElon and the root elU;
clearly this connection must be made somehow in the grammar.
Rather, I would claim, such a connection may be reflected
in the organization of the lexicon instead of by assuming
the productivity of vowel changes as a part of -an suffix-
ation.
3
Although there is a striking resemblance between
the two Imperative suffixes and the two Optative suffixes,
in that the former both end in -no and the latter both end
in -nE, I am quite sure that this similarity is more cor
rectly attributed to historical developments than to the
synchronic existence of some sort of -no Imperative suffix
and a -nE Optative, since numerous problems arise if one
tries to derive Locative Imperative -(n)ono, for example,
from Simple Locative -(n)I plus some sort of Imperative
-no. Even if, however, it should turn out that the Causa
tive and Locative suffixes can be broken down into two
separate morphemes, this would probably have little effect
on the validity of discussions contained in this disser
tation.
An alternative approach would be to allow Doubling
to take place before Causative and Locative Formation, with
the characteristic vowel changes and the addition of £ or n
taking place in both members of the doubled stem. (As will
be seen later, there are other rules in the grammar which
seem to operate In this way to maintain identity In doubled
forms (chapter 6).) In general, I see no way to decide
conclusively between the two possible approaches to the
occurrence of and n in both members of doubled Causatives
and Locatives; the choice of one over the other would, how
ever, have no great effect on the present discussion.
5
Final forms are shown with the effects of ng-Prefix-
ation, although I have not Included this step In the deriva
tion. For discussion of ng-Prefixatlon, cf. 3-5.
^Note that Causative Imperatives and Causative Optatives
also show the effect of HVL, but this is entirely predictable
since, even if the initial c[ of the suffixes is not present,
these two suffixes are still consonant-initial (i.e. -no and
-nE, instead of -qno and -qnE).
179
7
All forms of this type are subject to the lexical-
lzation with final light consonants spoken of in 3.2.5.3.
It should be noted also that neither Horne (1961) nor
Sumukti (1971) specifically mentions the fact that Causa
tives, like primary forms, have stem-final light consonants.
This fact is noted in Uhlenbeck 1963:76, however. Consequent
ly, at the time I did my informant work, I paid particular
attention to Causatives and did indeed find that the
pronunciation my informant considered to be correct for
Causatives did, in accordance with Uhlenbeck's observations,
involve neutralization of stem-final consonants.
8
In the following derivations, the final forms are
again given with ng-Preflxes, although this step is not shown
in the derivation. Note also that, although I have shown
Con-Neut applying before h-Deletion, these rules are not
crucially ordered with respect to each other.
9Cf. Dudas and O'Bryan 1972, Wilbur 1973, and,
especially, Hock 1973, where the author actually proposes
the concept of "synchronic analogy" as a way to deal with
cases where phonological rules exhibit apparently aberrant
behavior in order to maintain certain types of morphologically
significant identity.
10It may be significant that, in fast informal speech,
Javanese speakers make use of a variant pronunciation for
Simple Causatives in which the a of the suffix -(q )akE is
not pronounced. Thus, informal variants like nggawe-qkE
and nglUngguh-kE may occur instead of the more formal
nggawe-qakE and nglUngguh-akE. There is no doubt, however,
that for my informant at least, the real suffix is felt to
be - (a)akE. It may well be that the shortened form represents
a further attempt by speakers to bring Simple Causatives into
line with corresponding Causative Imperative and Causative
Optative forms. That is, deletion of the a of the suffix
results in complete parallelism between Caus. -(q )kE, Caus.
Imper. -(q)no, and Caus. Opt. -(q )nE, all of which have initial
consonant clusters when they occur with vowel-final stems and
a single initial consonant when affixed to consonant-final
stems. Notice that the shortening of -(q)akE to -(q)kE also
makes the application of the phonological rules in forms like
nglUngguh-kE transparent. On the other hand, it is entirely
possible that the shortening of the Simple Causative suffix
represents nothing more than an attempt by speakers to make it,
like all other suffixes in the language, monosyllahic.
180
CHAPTER V
ELATIVE FORMATION1
5.0 It is the purpose of this chapter to devote some
attention to the Javanese derivational process of Elative
Formation. In itself, the process by which Elatives are
formed may not be especially noteworthy; what is worth looking
into, however, is the unusual and interesting way in which
this morphological process interacts with certain phonological
rules included in the grammar of Javanese.
As will be seen in the following sections, contrary to
what is frequently taken for granted by generative linguists
and what appears to be the case in most languages (and, in
fact, except for EJative Formation, obtains in Javanese also),
Elative Formation, a morphological process, must be said,
in terms of rule ordering, to follow the operation of a
2
number of phonological processes. Secondly, it is of inter
est that surface Elative forms exhibit definite opacity with
respect to certain phonological rules; however, this apparent
opacity, it will be calimed, turns out to be well motivated
when the grammar of the language is looked at as a function
ing whole, with the surface phonological opacity not only
operating to preserve certain semantic contrasts, but, at the
same time, serving in many cases as the distinguishing sur
face marker of Elative forms.
5.1 Elative Formation is a highly productive process
In spoken Informal Javanese, although it does not appear In
181
either the written language or In the more formal levels of

speech. By the operation of this process, intensives are
formed from primary adjectives, so that, for example, the
Elative of adoh 'far', which is adUh, may be glossed as
'very far'. In very general terms, the processby which
Elatives are formed may be described as a change of the last
vowel in the primary adjective to tense IE or U, the choice
of the front or back vowel depending on whether the primary
stem has a front or back vowel in its final syllable. One
may thus speak of final-syllable tense I and U as the charac
teristic surface markers for Elatives.
At the same time, Elatives may (usually— cf. 5.2) option
ally exhibit an exaggerated drawing out of the characteristic
I or U and an accompanying rise in pitch on the final syllable.
In fact, in most Elative expressions, these prosodic devices
are used. However, it should be recognized that these prosodic
phenomena, while they occur very frequently and do help to
distinguish Elative froms from primary adjectives, especially
where there may be ambiguity otherwise, are not In general
sufficient to distinguish Elatives; the change in final
vowels must be regarded as the true process of Elative
Formation, with the prosodic phenomena playing only an ancil
lary role.
Elatives may be formed from nearly all primary adjectives
In the language, with a few logical exceptions. Elatives may
not be formed from adjective bases which have some other
182
commonly occurring Intensive form (e.g. tUwo ’old1 has the

Intensive form tUweq 'very old' and not the expected Elative
*tUwU). In addition, adjectives whose meanings are logically
not subject to intensification (e.g. patI ’dead') do not have
corresponding Elatives. Furthermore, no Elatives are formed
from adjectives whose final-syllable vowel is these may be
intensified by Juxtaposition with bang^t 'very' or by doubling
(cf. 1.8.H.2).
5.2 In most cases, the operation of Elative Formation
is clear and may readily be seen from examples like the
following:
abot heavy, hard abUt
adoh far adUh
alus refined, smooth alUs
angel hard, difficult angll
dJtfrO deep dJ^rU
IdJO green IdJU
lUgU ordinary lUgU
lUrus straight lUrUs
ramE noisy rami
r£siq clean r^slq
rlndiq slow rlndiq
o o
wan I bold wanl

Note that for primary adjectives with (tense or lax)
flnal-syllable mid vowels, the change to the Elative markers
183
I and U Involves raising of the final vowel. In addition,

for adjectives with final closed syllables, the Elative is
distinguished from the primary form by virtue of the tense
ness of the final vowel In the former (cf. adoh with lax
final-syllable vowel, but Elat. adUh, with a tense high vowel
in the final syllable).
For primary adjectives whose final-syllable vowel is
already high, there can of course be no question of raising
that vowel. In such cases, for adjectives with a final
closed syllable, the only feature which distinguishes Elatives
from primary forms Is the tenseness of the marker vowel; and
for adjectives with word-final I or U, the Elative forms are,
with respect to segmental characteristics, absolutely identical
to the primary form. It Is in Elatives like the latter that
the prosodic devices of drawing out the marker vowels and
raising the voice pitch must come Into play if the listener
is to recognize the intensified nature of the adjective he
hears. (Recall that these phenomena are more or less optional
for Elatives built on adjectives with other final-syllable
vowels.)
5.3 When the primary adjective has underlying /a/ In
Its final syllable (which appears as surface o by /a/ to ojt
(2.3 .1.3), Just In case It Is word-final) the operation of
Elative Formation Is slightly less straightforward. Consider
the following examples:
184

gampang easy gamplng
kirfras hard, harsh k^rls
larang high in cost larlng
rOso strong rOsU

kj^mbo Insipid k£mbU
With adjectives ending in high or mid vowels, it was
possible to say that the choice of Elative marker was deter
mined by the frontness or backness of the vowel of the primary
adjective. All adjectives with a final-syllable front mid
vowel, for example, whether that vowel was in an open (surface
E) or in a closed (surface e) syllable, formed Elatives by
replacing that front vowel with I. However, for adjectives
with /a/, such a simple statement is impossible, and it
appears that a distinction must be made between /a/ in a
closed syllable,•which is replaced by the Elative marker I
(cf. larang; Elat, larlng), and /a/ in an open syllable (i.e.
sufface o ), which is replaced by U in Elatives (cf. k^mbo;
Elat. kjgmbU). ^ As long as one considers only the underlying
forms of the primary adjectives, this split seems rather
puzzling. Granted that /a/ is neither a front nor a back
vowel in Javanese, why should speakers, having decided on
the front vowel marker for Elatives built on adjectives with
/a/ in their final syllables, further complicate matters by
choosing the back vowel marker for /a/ when it occurs word-
185
finally-— or vice versa?

The reader has probably already noticed that the choice
of U as the Elative marker for adjectives in word-final /a/
is not so puzzling if, taking a rule ordering approach, one
assumes that Elative Formation applies after the phonological
rule of /a/ to off has already applied. At this point, although
/a/ in a closed syllable is still neither front nor back,
and may therefore apparently arbitrarily be assigned I as an
Elative marker, word-final /a/ has already become o, which is
clearly a back vowel (and may be treated along with o from
/0/) and therefore should logically have an Elative which is
marked by U.
5.4 Now consider the following examples of Elatives
built on stems with two basic /a/vowels:
dowo long dowU
loro ill, painful lorU
ombo large ombU
Note that in these Elative forms, not only is the Elative
marker the U which in general replaces back vowels, but, at
the same time, the penultimate /a/ of the underlying form
(e.g. /dawa/, /lara/, etc.) shows up as surface o. The lat
ter Is a situation which only arises through the operation
of a-Vowel Harmony (2.3.3.3 and 2.3.4.2), harmonizing a pen
ultimate basic /a/ to a word-final o which has been produced
from /a/ by the operation of /a/ to off. Thud it seems that
186
not only must /a/ to off precede Elative Formation, but, in

addition, a-Vowel Harmony must also apply before the morpho
logical process takes place.
5.5 If the Elatives of stems containing penultimate
mid vowels are brought into consideration, similar con
clusions may be reached concerning the ordering of Elative
Formation and the phonological rules which determine the
surface tenseness of mid vowels:
Edan crazy Edln
eleq bad ellq
groboh rough, crude grobUh
kEkuq clumsy kEkUq
kosong empty kosUng
5.5.1 Looking at the above examples, one can see that
in Elatives the tenseness or laxness of the penultimate mid
vowel does not appear to be predictable in any way from
the surface environment, so that, for example, one finds
Elat. Edln with a tense penultimate E, but Elat, ellq with
a lax e in an identical vocalic environment. Thus, if
Elative Formation were to apply, as morphological processes
usually do, before the phonological rules which determine
mid vowel tenseness, there would be no way for the latter to
distinguish between cases like Edln, from basic /Edan/, and
those like ellq, from basic /ElEq/?
5.5.2 The only way for the relevant mid vowel rules to
187
apply properly to derive the surface Elative forms In the

above examples Is for the morphological process to apply
after the phonological rules in question. Thus, the
derivation of Elative forms like Edln and ellq would proceed
as follows:
/Edan/ /ElEq/
Mid Vowel Laxing 1 ---- Eleq
Mid Vowel Harmony ---- eleq
Elative Formation Edln ellq
other rules Edln ellq
5.5*3 Notice that these derivations indicate that
Elative Formation must follow the rules of Mid Vowel Laxing 1
(2.6.1.2) and Mid Vowel Harmony (2.6.4.2.2). It also appears
that the morphological process may follow High Vowel Laxing
(2.4.2.5) as well. Recall, that the surface Elative markers
are always tense I and U occurring in place of the vowel of
the last syllable of theprimary stem. In manycases (as In
the preceding derivations) thismeans that Elativeforms have
tense vowels in final closed syllables: that Is, in precisely
the environment for HVL. Thus, If Elative Formation intro
duced its tense vowel markers before the application of HVL,
the latter might be expected to apply, producing a lax sur
face Elative marker. This, of course, Is not what happens.
What does happen (i.e. the tense Elative markers come through
to the surface unchanged) is consistent with the introduction
of Elative markers after HVL has already applied. It should
188
be noted, however, that the same surface result might plausibly

be obtained by claiming that all Elatives must be exceptions
to HVL.
5.6 The reader has no doubt already discovered the
obvious surface opacity with respect to certain phonological
rules exhibited by Javanese Elative forms. The fact that
Elatives have tense vowels in the exact environment for HVL
is a clear case of one type of opacity (cf. 3.2.5.1). That
is, Elative forms like angll, rlndiq, adUh, etc..have high
tense vowels in their final (closed) syllables which are in
the environment for HVL, but which nevertheless appear to
have remained tense.
A second type of opacity is exhibited by Elative forms
like dowU, kosUng, and ellq (primary adjectives dowo, kosong,
and eleq) where the penultimate vowels appear to have under
gone changes connected only with a-Vowel Harmony or Mid Vowel
Harmony even though the final-syllable vowels which must be
present to condition such changes are not present on the sur
face, having been replaced by the Elative markers I and U.
It is quite apparent that both of the above instances
of opacity with respect to phonological rules arise through
the proposed ordering of the morphological process of Elative
Formation after the phonological rules in question. Although
this situation is quite unusual in several respects, there Is
good motivation for its presence In the grammar of Javanese.
5.7 I would In fact find it quite reasonable to claim
189
that the unusual Interaction of Elative Formation with the

phonological rules serves the purpose of producing the very
surface forms which appear to be opaque and that, in this
case, the apparent opacity actually has two functions:
that of maintaining the distinctness of Elative forms from
primary forms; and that of preserving the identity of under
lying lexical items. Both of these functions would of course
contribute to the successful utilization of Elatives by
speakers of the language.
5.7-1 To see first of all how opacity in Elatives
functions to keep them distinct from primary forms, one
needs only to consider a few Elatives with final closed syl-
5
lables, such as rlndiq and alUs (primary rlndiq and alus).
• •
A hypothetical Javanese speaker encountering such forms would

know immediately that they are not primary words, but have
undergone intensification by Elative Formation. What signals
this information to him is the presence of the tense high
vowel (i.e. the Elative marker) in the final closed syllable;
in primary forms, no tense vowels occur In this environment,
since they are all laxed by the regular operation of HVL; and
the speaker accordingly knows that the only tense vowels in
that environment which reach the surface are those introduced
by Elative Formation. In other words, it Is precisely the
fact that Elatives are opaque with respect to the phonological
rule of HVL that allows the speaker to distinguish between
primary forms with final closed syllables containing surface I
190
and u and Elatives.

At the same time, opacity of a different type plays a
role In distinguishing certain other Elatives from primary
forms (namely, thoj^ Elatives which correspond to primary
forms with underlying shapes like /CECEC/ or /C0C0C/). Con
sider, for example, the possible reaction of the hypothetical
speaker when he encounters an utterance which Includes the
words geplng or kosUng. He knows first of all that he has
heard an Elative by virtue of the tense marker in the final
closed syllable, as was discussed directly above. However,
In this case, In addition to the character of the final vowel,
he has yet another signal that what he has heard is an Ela
tive: the vowel configuration of the whole word. He knows
that the penultimate lax e or o (from: basic /E/ or /0/)
could have come about only by the operation of Mid Vowel
Harmony. However, in Elatives like geplng and kosUng
(primary gepeng and kosong) there is no vowel In the final
syllable which might have brought about the laxness of the
penultimate mid vowel, the conditioning vowel having been
replaced by the Elative marker. The fact that Mid Vowel
Harmony is opaque in such forms, together with the fact that
this opacity occurs only In Elatives, helps to distinguish
Elatives from primary forms.
5.7*2 The above, however, is more or less a redundant
function of Mid Vowel Harmony opacity in Elatives like geplng
and kosUng, since the presence of the tense vowel in a final
191
closed syllable might be of itself sufficient indication

that such a form is an Elative. The real significance of
opacity with respect to Mid Vowel Harmony in such Elatives
is, in my opinion, that it functions to allow the Javanese
speaker to distinguish between the various possible under
lying forms on which an Elative like geplng or kosUng might
be built.
Since Elative Formation involves replacement of final-
syllable front vowels with I and back vowels with U, it must
necessarily result in a certain amount of neutralization of
underlying contrasts, with e, lt and a in primary final
closed syllables all coming out as I in Elatives, and, simi
larly, primary o and u falling together into Elative U. Such
a situation could result in a great deal of confusion about
the primary source for a given Elative; however, because of
the organization of the grammar of Javanese, which results in
phonological opacity of certain Elatives with respect to Mid
Vowel Harmony, a great deal of this potential homonymy and
its resulting confusion may be avoided. 6
Consider, for example, the Elatives Eklr (primary Ekar
'lopsided') and eklr (primary eker 'impatient for something').
Notice that, because of the neutralization-brought about by
the introduction of the Elative marker vowel, these two forms
are identical except for the difference in tenseness of the
penultimate vowels; and this difference in tenseness of the
vowels in question has the effect of making each Elative sound
192
more like its corresponding primary adjective and thus giving

the hearer an immediate clue to its semantic import. This
distinctive difference in tenseness, however, is a consequence
of the opacity of the Elative eklr with respect to Mid Vowel
Harmony. That is, the initial e of this form is lax, even
though the lax final-syllable mid vowel (cf. primary form
eker from /EkEr/) which must have originally conditioned
this laxing has been replaced by the Elative marker I. If
the grammar of Javanese were organized differently, so that
the opacity in question were eliminated, it would lead to
possible homonymy of the Elatives of primary forms of the
shape (C)eCeC, (C)ECiC, and (C)ECaC, all of which would have
7
Elatives of the shape (C)ECIC. The same possibility of
confusion would exist between the Elatives of primary forms
of the shape (C)oCoC and (C)OCuC, but this confusion is
again avoided by the ordering of Elative Formation after
Mid Vowel Harmony and the resulting opacity which functions
to make the penultimate vowel of the Elative identical to
the same vowel in the corresponding primary.
Thus, it seems that, as a result of the unusual position
of Elative Formation In the grammar of Javanese, a fair
amount of the potential confusion which could result from
the neutralizing effect of the derivational process itself
is avoided. Although this of course serves mainly to dis
tinguish Elatives corresponding to primary forms of the
shapes 'C)eCeC and (C)oCoC (but cf. 5«8.1 below), this Is
193
quite significant in light of the fact that, according to

Uhlenbeck (19^9:98), roots having identical vowels in both
syllables are by far the most prevalent in the language;
and thus the forms affected by the opacity in question could
be quite numerous.
5.8 So far in the discussion of the functional role of
opacity in Javanese Elatives consideration has been limited
to Elatives corresponding to primary adjectives with final
closed syllables. This has been done mainly to facilitate
discussion, but also because such forms illustrate so clearly,
not only the functional role of opacity with respect to Mid
Vowel Harmony, but also the way in which opacity with respect
to HVL functions to mark Elative forms. At the same time,
morphemes with final open syllables are actually much less
frequently occurring in Javanese. There are, however, cer
tain facts about the latter which are relevant to discussion
of the functional role of opacity in Javanese Elatives.
5.8.1 First of all, just as was the case with Elatives
built on primary adjectives with two identical mid vowels,
Elatives of primary forms with two open-syllable /a/ vowels
also exhibit opacity— this time with respect to a-Vowel
Harmony. And this opacity also functions to avoid possible
confusion by making the Elative sound more like its corres
ponding primary form. That is, in the Elative dowU, for
example (primary dowo from /dawa/), the penultimate o from

underlying /a/ can only have arisen through harmonizing to
194
to a final o which has been derived from /a/ by /a/ to o#.

But in the Elative this final o has been replaced by U and
the form is therefore opaque with respect to a-VH (thus dowU
instead of *dawU). This opacity, however, has the effect of
making the penultimate vowel of the Elative identical to the
penultimate of its corresponding primary form and thereby
avoiding the confusion arising from neutralization of the
final vowel, which would otherwise allow the possibility that
the Elatives of CoCo (/CaCa/) primaries might be segmentally
identical to those of CaCU and CaCO primaries.
5.8.2 There is yet another way in which a type of
opacity in Elatives of adjectives with final open syllables
serves to avoid the possible confusion resulting from
partial neutralization of underlying contrasts. Consider
the following examples:
bOdO stupid bOdU
• •
lOgrO loose lOgrU

sEdjE different sEdjl
Notice that, if Elative Formation were to apply before Mid
Vowel Laxing 2 (2.6.2.3), which laxes a penultimate mid vowel
before a final-syllable high vowel, one would expect the
above Elatives to have lax penultimates, since the Elative
markers create the precise environment necessary for applica
tion of MVL 2. However, the Elative corresponding to lOgrO
’loose’, for example, is not alogrU, with the expected lax o.
195
but lOgrU, with penultimate tense 0. This of course rep

resents opacity with respect to the rule of MVL 2. At the
same time, however, Just as with the opacity of certain
Elatives discussed above, the result is minimization of the
possible confusion caused by Elative neutralization by means
of producing a surface Elative with the same penultimate
vocaiism as its primary counterpart. The opaque vowel
configuration of such words may function as a signal to
speakers that they are dealing with Elatives and not pri
mary words. This is an especially important consideration
for certain Elatives— namely, those corresponding to primary
adjectives with final U or I— with final open syllables,
because, as was mentioned above in 5.2, such forms are not
readily recognizable as Elatives otherwise.
5.9 In summary, then, it has been shown that many
Javanese Elatives exhibit opacity with respect to a number
of phonological rules. Because this opacity must arise
through (taking a strict rule ordering approach) the unusual
ordering of a derivational morphological process after a
number of phonological rules, it might be generally concluded
that such a situation is extremely unlikely. It was claimed,
however, that, unlikely as it seems, there is good motivation
for the existence of such a situation in the grammar of the
language, since the opacity of certain Elatives probably
serves to distinguish them from primary forms and at the
same time helps to avoid some of the potential confusion about
196
underlying distinctions that might otherwise result from

the partial neutralization of final vowel contrasts inherent
in the process of Elative Formation itself.
5.10 Although I have spoken only in terms of rule
ordering so far, there are some interesting alternative
approaches to the analysis of Javanese Elatives which ought
to be considered.
5.10.1 First of all, there is the possibility that
Elative Formation is a rather unusual derivational process
that simply operates on the surface forms of primary adjec
tives and replaces their final vowels with the characteristic
Elative markers. This would be entirely consistent with the
data and would perhaps account for why all the examples of
opacity that I have claimed help to maintain underlying dis
tinctions actually result in making Elatives look as much as
possible like their corresponding surface primary forms.
Thus, such an approach would basically be very similar to
one using ordering, with the difference being that simply
claiming that Elative Formation is ordered after certain
phonological rules still leaves room for the possibility
that it may be ordered before certain others— even though,
to the best of my knowledge, there are no phonological rules
that must be ordered after it. The surface derivation
approach of course by definition precludes this possibility.
At the same time, the rule ordering approach makes no claims
about what the relationship between Elative Formation and
197
any new rule entering the grammar might be; the surface
derivation approach, however, would claim that such phono
logical rules would of course precede Elative Formation,
which would take only surface forms as its input.
5.10.2 An essentially similar, but somewhat more
insightful, approach would involve the operation of syn
chronic analogy (cf. Hock 1973 and chapter 4 of this dis
sertation) within the grammar of Javanese. Taking this
approach, one might include in the grammar a principle
governing the application of Elative Formation which
specifies that the process will apply in any way necessary
to ensure that the surface forms of Elatives will look as
much as possible like their corresponding primary forms,
but, at the same time, be distinguishable as Elatives. Or,
alternatively, the proposed principle might be set up to
govern the application of the phonological rules, specifying
that they may over- or under-apply in any way necessary to
produce Elative forms with the desired characteristics.
Note that this approach would differ from both the rule
ordering approach and the surface derivation approach in that
it does not entail setting up a specific point in the deriva
tion where Elative Formation will always apply; rather, it
claims that the morphological process, or, alternatively,
the phonological rules, will apply in any way necessary to
produce the correct surface Elative forms.
198
5.10.3 A third approach that might be taken to account

for the facts of Javanese Elative formation is that of
positing some sort of transderivational constraint for
the language.® It seems that the major effect of ordering
Elative Formation after Mid Vowel Harmony, for example, is
the avoidance of homonymy among certain Elatives by making
them sound more like their primary counterparts. However,
these same results might be obtained by alternatively
claiming that, instead of the suggested ordering, the
grammar of Javanese possesses some device that operates
during the derivation of an Elative to "look over” to the
derivation of its primary form in order to ensure that the
final output of the former will have the same penultimate
vowel as the final output of the latter. Thus, for example,
at that point in the derivation of Elat, geplng (cf. 5.5)
at which Mid Vowel Harmony should apply (but ordinarily
could not because the input at this point would be /gEpIng/,
which does not meet the environment for this rule) the
grammar would look over to the same point in the derivation
of primary gepeng (which at this stage would be /gEpeng/),
see that Mid Vowel Harmony does apply in the latter, and
accordingly, lax the penulatimate vowel of the Elative to
ensure that there will be the necessary surface resemblance
to primary gepeng. Mote that if this approach were taken,
some way would have to be found to account for the opacity
of Elatives with respect to HVL (cf. 5.5.3). This problem.
199
however, could be taken care of by claiming that all Elatives

are automatically exceptions to this rule. Another problem
would be that the choice of the I Elative marker for forms
with a in a final closed syllable and the choice of U to
mark Elatives of primaries with /a/ word-finally would both
seem to be completely arbitrary (cf. 5.3 above).
200
FOOTNOTES
^This chapter appeared, in a slightly different version,
as Dudas 197^.
2
Although the literature contains numerous analyses
where morphological processes are ordered among phonological
rules, as was noted in Wilbur 1973s these special orderings
usually involve the morphological process of reduplication
and can be eliminated by acceptance of an identity constraint
which exists between the component parts of reduplicated
forms (cf. chapter 6 of this dissertation).
As was mentioned earlier,, in chapter 3, Uhlenbeck
(19^9) gives a description of Javanese Elatives which differs
slightly from that presented here. The main difference lies
in the fact that, as Uhlenbeck describes it, Elative Forma
tion involves replacement of any vowel in a final syllable
ending in q with the marker vowel instead of the usual
I or U. However, as was noted in chapter 3, my informant
seems to have generalized the more common type of Elative
Formation (i.e. that using !E and U exclusively as Elative
marker vowels) to all types of stems, including those in
final There is also one other difference between
Uhlenbeck*s description of Elatives and what I found to be
true for my informant: Uhlenbeck claims that primary forms
in final /a/ may have Elatives with either I or U as the
marker. I found this to be the case only for a very small
number of adjectives (e.g ombo has both ombl and ombU as
permissible Elatives), with the U marker clearly being, for
my informant at least, the productive one for such forms.
k
It should be noted that there do exist a few primary
forms in the language which are also exceptions to HVL. Such
words belong to the category of expressive or onomatopoeic
words and may thus be treated as categorial exceptions to
HVL. Cf. chapter 2, footnote 7, for more discussion of such
forms.
5
For a discussion of some ways in which surface opacity
may function to preserve underlying contrasts, cf. Kisseberth
(1973).
6
The language appears to escape confusion in many other
cases by what appears to be more or less an accident of dis
tribution. In Javanese (according to Uhlenbeck (19^9:98))
bisyllabic roots with certain vowel configurations are very
common, while others are less common, and still others are
extremely rare. It turns out to be the case that In many
instances where a given Elative could theoretically correspond
to more than one primary form, one of these forms has a vowel
201
configuration that is at least twice as common as the other,

which occurs very rarely. Thus, for example, an Elative of
the shape CUCUC could conceivably correspond to either of
the primary forms CUCuC or CUCoC. However, Uhlenbeck gives
the frequency of occurrence of the former type as 356,
while the latter type only occurs 35 times. So that, given
an Elative like CUCUC, there should in actuality be very
little possibility of confusion as to primary forms, with
one of the possibilities being very likely to actually occur
as a primary adjective, and the other extremely unlikely.
7
The reader may have noted that the Elatives of CECiC
and CECaC forms still fall together. This problem is, how
ever, partially avoided by the distributional statistics
mentioned in the previous footnote, with the frequency of
morphemes of the former type being, according to Uhlenbeck,
only 26, while the latter occurs 126 times.
g
For discussion of a rather different situation in
which a transderivational approach to the avoidance of
homonymy is suggested, cf. Kisseberth and Abasheikh 197*1.
202
CHAPTER VI
REDUPLICATION
6.0 In preceding chapters, attention has been devoted
mainly to discussing the phonological and morphological
processes involved in deriving forms consisting of either
a simple stem by itself, or a simple stem plus one or more
affixes. In this chapter, the focus will be on certain
aspects of those forms whose derivation involves, in addi
tion, the morphological process of reduplication, or more
specifically, the type of reduplication referred to as
Doubling, in which a copy is made of the entire stem.
Recall from 1.8.4 that the language makes use of two
different types of Doubling: simple copying of the entire
stem; and copying of the stem with an accompanying morpho
logically conditioned vowel change. Of these two types,
only the first will be relevant here; Doubling with accom
panying vowel change will be discussed in chapter 7.^
Doubling is one of the most frequently used processes
in the language and can, to the best of my knowledge, occur
in conjunction with any of the affixation processes. Usually,
Doubling is used to indicate various sorts of plurality—
either of objects or actions— although there are certain cases
where it is used in conjunction with a specific affix with
a slightly different semantic import. Thus, for example,
while slnaU means ’study5, and slnaU-sInaU can mean simply
’keep studying’, slnaU-sInaU-w-an means something like
203
'pretend to be studying'. As with other categories, the

major concern in this chapter will be with the phonological
or segmental marking aspects of Doubling and not with its
exact semantic function.
6.1 In all cases, when a word consists of a stem, with
or without affixes, which has undergone Doubling, both
realizations of the stem in the doubled word are identical.
This can be seen, first of all, in simple, non-affixed,
doubled words:
base stem gloss doubled
/abad/ abat century abat-abat
/abUr/ abur flight abur-abur
/atl/ atl heart, mind atl-atl
/bOdO/ bOdO stupid bOdO-bOdO
• • • •
/dawa/ dowo long dowo-dowo

/gjtfdE/ gf*dE big g£dE-g£dE
• • • •
/kOdOq/ kodoq frog kodoq-kodoq

• • • •
/mEdj a/ mEd jo table mEdj o-mEdj o

/mUrld/ mUrit student mUrit-mUrit
/plllh/ pllih choose pllih-pllih
Notice that, just in case the simple stem undergoes
any of the phonological rules discussed in chapters 2 and 3»
the effects of the rule will also show up in both members of
of the doubled word, even though, strictly speaking, the
rule's environment may not be met in both members. Thus, for
example, /a/ to of (2.3.1.3) was set up as converting a
204
word-final /a/ to o and, accordingly, is clearly responsible

for the surface o of mEdJo, from /mEdJa/. However, in the
doubled form mEdJ o-mEdJo, even though the final /a/ of the
first member is not in absolute word-final position, it
still is identical to the second member in having final o.
Similarly, abat-abat, from /abad/, shows the effects of
Consonant Neutralization (3.1) in the final consonant of
both members, even though the rule’s word-final environment
is met only by the rightmost member. The same is true for
abur-abur, from /abUr/, with respect to High Vowel Laxing
(2.4.2.4): only the /U/ of the second member is actually in
a closed syllable; but yet both members have lax u in their
final syllables.
6.2 Traditionally, generative linguists have used
several different ways to account for situations like that
presented by Javanese doubled forms. Among these are (1)
positing a special reduplioation boundary between the members
of doubled pairs and making the phonological rules sensitive
to this boundary, (2) positing a word boundary between the
two members, and (3) imposing special constraints on the
?
ordering of Doubling with respect to the phonological rules.
6.2.1 In the case of the Javanese examples given above,
it is possible to eliminate from consideration the first of
these alternatives on the grounds that It would be an unneces
sary complication to set up a special reduplication boundary
in the grammar. Since, In every case, the leftmost member
205
of the doubled pair has, like the rightmost, undergone a

rule whose environment specifies word-final position, it
would seem perfectly plausible and much more economical
to simply claim that it is a word boundary which separates
the two members of doubled forms (i.e. utilize the second
alternative listed above) and allow the rules to apply as
they were originally formulated. Taking this tack, a form
like mEdjo-mEdjo, for example, could derive from intermediate
/mEdja#mEdja/, which has already undergone Doubling, with
/a/ to q# applying regularly to both members.
6.2.2 Adopting the third alternative would involve
ordering the morphological process of Doubling after all
of the relevant phonological rules have applied, so that
the process would not make a copy of the stem until after
the necessary phonological changes had taken place. Thus,
mEdjo-mEdjo would derive from basic /mEdja/, which would
become /mEdjo/ by /a/ to o#, and only then would Doubling
take place to produce mEdjo-mEdjo.
6.3 Although both the word boundary approach and the
ordering approach just described would seem to be equally
effective in producing the correct surface forms for doubled
non-affixed stems, when words derived by means of Doubling
and affixation are examined, it becomes evident that only
the ordering approach is feasible.
6.3 .I Consider, for example, the doubled forms, with
3
and without affixes, for the following stems in final /a/;
206
stem gloss doubled doubled affixed

dOngo prayer dOngo-dOngo dOnga-dOnga-nE
dowo long dowo-dowo dawa-dawa-nE
mEdjo table mEdj o-mEdj o mEdj a-mEdJ a-nE
Like the simple doubled stems discussed earlier, the above
doubled stems with the Demonstrative suffix -(n)E exhibit
identity between the two parts of the doubled form, with
both the right and leftmost members having the same surface
realization. But note that, while it is possible to account
for the application of /a/ to o# in both members of the
simple doubled forms by positing a word boundary between
them, this approach is not able to account for the non-
application of the rule in the first member of doubled
affixed forms, such as mEdj a-mEdJ a-nE. In fact, if a word
boundary were present between the two members, one would
expect /a/ to o§ to apply to the final /a/ of the first
member. Note also that the first alternative given in 6.2
above (i.e. positing a special reduplication boundary between
the members of doubled forms) will not work either. Using
this approach, there would be no way to account for why
/a/ to o# appears to have applied in the first member of
forms like mEdjo-mEdjo, but not in the first member of
mEdj a-mEdja-nE in precisely the same type of environment
(I.e. before the special reduplication boundary).
If, however, the ordering approach is taken. Doubling
would copy the stem as It occurs in the Intermediate form
207
/mEdja-nE/, after /a/ to o# has already had a chance to

apply, but, of course, has not applied, since the presence
of the affix would rule out the necessary word-final environ
ment for /a/. Copying the stem at this stage of the deriva
tion would produce the correct surface form mEdj a-mEdJ a-nE.
The reader is advised at this point that it will be crucial
to later discussion (6.5) that adoption of this ordering
approach to the analysis of doubled suffixed forms involves
the implicit claim that Doubling makes a copy to the left of
(or preceding) the basic stem. That is, a surface form such
as mEdj a-mEdJ a-nE must be thought of as consisting of a copy
of the stem followed by the original suffixed realisation of
the stem.
6.3.2 The identical situation exists with respect to
stems with high vowels in final (basic) closecf syllables
and the operation of High Vowel Laxing:
abur flight abur-abur abUr-abUr-E
apiq good, nice apiq-apiq apIq-apXq-E
dUduq place dUduq-dUduq dUdUq-dUdUq-E
• •
gllik cylindrical gllik-gllik glllg-glllg-E

mUrit student mUrit-mUrit mUrld-mUrld-E
Although HVL appears to have applied to both members of the
simple doubled stems, in affixed forms it has applied to
neither, with both stem-final syllables having tense high
vowels. In such cases, positing a word boundary between
208
the two parts of the doubled word will not work, since one
would then expect to get forms like *abur-abUr-E , with HVL
having operated on the high vowel of the first member.
And again, positing a special reduplication boundary Is also
not feasible, for the same reason that was cited In 6.3.1.
Ordering Doubling to apply after HVL is again a possible
solution, since this would provide forms such as intermediate
/abUr-E/ as the input to Doubling, which would then copy the
stem to produce surface abUr-abUr-E.
6.3.3 Exactly the same problem occurs when one tries
to account for the differences between the stem-final conso
nants in forms like mUrlt-mUrlt and mUrld-mUrld-E , given
in the preceding section: Consonant Neutralization has taken
place in both members of the former, but in neither the copy
nor the original in the latter, indicating that, once again,
one might assume that Doubling has taken place after the
operation of the relevant phonological rules.
6.3.4 In all of the examples given so far, doubled
forms with affixes have not undergone rules which have
applied in the derivation of doubled non-affixed stems.
There is, however, still another rule, h-Deletion,(3.3.2),
which must be discussed, whose environment is not met at
all in simple doubled forms, but which applies in those which
have undergone certain types of affixation:
b^dah broken b^dah-b^dah b^da-b^da-E
209
dajoh guest dajoh-dajoh daJo-daJo-E

• • • • •
gadjih salary gadjIh-gadhih gadjI-gadJI-y-E

kUkuh solid kUkuh-kUkuh kUkU-kUkU-w-E
In cases of this type, there is of course again no way
to use an approach positing a word boundary between the parts
of the doubled stem to produce the correct surface form for
the first member, since h-Deletion, operating strictly in
an intervocalic environment, does not apply before word
boundaries. And, as in the other cases discussed, positing
a special reduplication boundary to which h-Deletion could
be made sensitive would work in suffixed forms like
b^da-b^da-E, but would leave no way to account for the non-
• •
application of the rule in the first member of non-affixed

bj^dah-b/gdah. Thus, once again, it appears to be necessary
to adopt an approach ordering Doubling after the rule of
h-Deletion has already removed the stem-final consonant in
order to derive the correct surface forms.
In summary, then, It seems that the ordering approach
Is the only one of the three suggested alternatives which
will produce the correct surface realizations for all the
doubled forms discussed In preceding sections, both suffixed
and non-suffixed. Taking such an approach to the derivation
of these Javanese simple and suffixed doubled forms, one
would have to defer the operation of the morphological
process of Doubling at least until after the following
phonological rules had applied: /a/ to o#; HVL; Con-Neut;
210
and h-Deletion. Notice that, because MVL 1 must precede

h-Deletion (3*3*2), this means that the operation of Doub
ling must be deferred until after nearly all phonological
rules in the grammar have applied. In addition, it also
seems to be the case that such an approach to the analysis
of suffixed doubled forms makes the claim that copying is
done to the left of the original stem, so that in a doubled
form like b^da-b^da-nE, for example, the first member may
• •
be thought of as a copy, while the second would represent

the original suffixed stem.
6.4 So far, discussion has been limited to simple and
suffixed doubled forms. In addition, one must also consider
what happens when Doubling occurs with a stem which also
undergoes prefixation. Because most Javanese prefixes are
like Passive dl- and do not ever enter into phonological
interactions with the stem, the only prefix relevant to
this discussion will be the Simple Active prefix ng-.
It will be recalled from 3*5 that ng-Preflxation may
very generally be described as Involving the following:
(1) addition of the prefix /ng-/ to the stem; (2) assimi
lation of the nasal to a stem-initial consonant; (3) replace
ment of a voiceless stem-initial consonant with the assimi
lated nasal. Subsequently, the nasal of the prefix and a
stem-initial voiced consonant merge by means of Voiced
Consonant Nasalization, with the result being a nasalized
version of the initial consonant of the stem. The following
211
illustrates the derivation of qg-Preflxed forms;

/adOh/ /pangan/ /bakar/
Nasal Prefixation ng-adOh ng-pangan ng-bakar
Nasal Assimilation ---- m-pangan m-bakar
Replacement ---- mangan ----
Voiced Consonant mbakar
Nasalization
other rules ngadoh mangan mbakar
6.4.1 It is significant that, in prefixed doubled
forms, not all types of stems behave alike: specifically,
it is only with stems beginning in a voiceless consonant,
which is deleted by Replacement, that the nasal prefix
obligatorily occurs in both members of the doubled form;
with vowel-initial stems, or with stems beginning with a
liquid or a voiced consonant, where there is no such dele
tion, the nasal may occur in both members of the doubled
word, but does not necessarily have to occur twice. Thus,
one finds the following ng-Preflxed doubled forms:
stem gloss doubled affixed
kondo
•
say, tell ngondo-ngondo
« •
pangan food mangan-mangan

sUwE long (time) njUwE-njUwE
tUlls write nUlis-nUlis
4
adang steam ngadang-ngadang
ngadang-adang
212
adJar learn ngadjar-ngadJ ar

ngadj ar-adJ ar
IdJO green ngldj O-ngldj 0
ngldJO-IdJO
r^siq clean ngr^siq-ngr^siq
ngr£slq-r£siq
bakar roast mbakar-mbakar

mbakar-bakar
dadl become ndadl-ndadl
ndadl-dadl
djaran horse ndjaran-ndjaran
ndJ aran-dj aran
godok boll nggodok-nggodok
• « •
nggodok-godok
• •
Notice that for the first group of stems (I.e. those like
kondo, which have Initial voiceless consonants), only one
doubled prefixed form is given, but for stems in the other
two groups (i.e. those like adang, which are vowel-initial,
and those with initial voiced consonants, such as bakar),
two alternative pronunciations for doubled prefixed forms
are given: one with the nasal appearing on both members of
the doubled form and one where the nasal occurs only on the
leftmost member.
6.4.2 As with the doubled suffixed forms discussed
earlier, one might consider producing doubled prefixed forms
213
by using a special reduplication boundary to which relevant

rules can be made sensitive; however, as before, such an
approach turns out to be less satisfactory than an ordering
analysis. Using the boundary approach, one could assume
that Doubling applies before Nasal Prefixation takes place.
(Note, incidently, that this would impose no necessary direc
tionality on the copying involved; Doubling could go to the
right or to the left.) The prefix would then be added to
the beginning of the doubled word, as usual, and, in addition,
it would obligatorily be added to the second member of the
doubled form (i.e. the ng-Prefixation process would be
sensitive to both word boundaries and the special reduplica
tion boundary) just in case the stem began with a voiceless
consonant; and the /ng-/ would be added to the second member
optionally if the stem began with a vowel, liquid, or voiced
consonant. Or, using the same approach with a slight varia
tion, /ng-/ could be added to both members of all doubled
forms, but a later rule would optionally delete the prefix
from the second member just in case the stem began with a
vowel, liquid, or voiced consonant.
6.4.3 Both of these analyses using boundaries may be
eliminated on the grounds that they are extremely ad hoc and
redundant. In the first variation, in order to make sure
that the ng-Prefix will always turn up in both members of
doubled stems like mangan-mangan, from /pangan/, but does
not necessarily have to be part of the second member of
214
doubled forms like mb akar-(m )bakar, from /bakar/, the

environment for replacement of the stem-initial consonant
by the nasal must also be built Into the ng-Preflxatlon
process Itself. In the second variation, the entire
environment for non-operatIon of Replacement has to be built
into the optional rule which deletes the nasal from the
second member of doubled stems like mb akar-(m )bakar.
Clearly, in both cases, even with all the required
redundancy, all that is achieved is a mechanical solution,
which offers no possibilities for explanation at all for
what might be the source of the difference between those
ng-Preflxed doubled stems like mangan-mangan and those like
mbakar-(m)bakar.
6.4.4 However, a much less redundant and ad hoc analysis
can be achieved by using an ordering approach similar to
that already suggested for non-prefixed doubled forms. Taking
such an approach, one need only impose an ordering on the
rules of the grammar which guarantees that Doubling will not
apply until after Replacement has taken place (i.e. until
after all stem-initial voiceless consonants have been replaced
by their homorganic nasals). This would offer an explanation
for the difference between doubled stems like mangan-mangan
and those like mbakar-(m)bakar. Taking a stem like /pangan/,
at the point where Doubling takes place. Nasal Assimilation
and Replacement would already have applied to give inter
mediate /mangan/; and it would be this form, with the nasal
215
already made a part of the stem, which would be copied, to

give mangan-mangan. With a stem like /bakar/, Nasal Assimi
lation (but not Replacement) would apply to give /m-bakar/,
with the prefix and the stem still separate, as the input
to Doubling. The process would then apply, with the option
of copying only the original stem or the stem and the nasal•
of the prefix. 5 Thus, the derivation of mangan-mangan and
mb akar-(m )bakar would proceed as follows:
/pangan/ /bakar/
Nasal Prefixation ng-pangan ng-bakar
Nasal Assimilation m-pangan m-bakar
Replacement mangan
Doubling mangan-mangan m-bakar-(m)-bakar
other rules mangan-mangan mbakar-(m)bakar
6. *1.5 It is important to note that in the above analy
sis, Doubling must be formulated as a rule which makes a
copy to the right (or after) the original stem. Although
this is not apparent with forms like mangan-mangan, where,
with an input like /mangan/, Doubling could go either to the
left or to the right with no ultimate difference, it is
crucial in deriving forms like mbakar-(m)bakar, where the
leftmost member must have the nasal prefix reflected on the
surface, although this Is only optional In the rightmost
member. That is, given an Input like /m-bakar/, if Doubling
were to make a copy to the left, there would be no good way
to account for the fact that the nasal on the leftmost member
216
(the copy) must obligatorily appear, but the rightmost

member (the original) does not necessarily have to have the
nasal. Thus, the only way to get a surface form like
mbakar-bakar (with only one nasal) using a Doubling rule
which copies to the left would be to start out with /m-bakar/,
copy to the left to get /m-bakar-m-bakar/ and then delete the
nasal of the second member after Doubling had taken place.
This of course would be an extremely ad hoc analysis, since
the only reason to ever include the optional nasal on the
rightmost member in the derivation of a form like mbakar-
bakar would be to give Doubling an opportunity to get the
obligatory nasal onto the leftmost member.
6.5 The reader has probably noticed by now that
adopting the suggested ordering approaches to the analysis
of both prefixed doubled forms and suffixed doubled forms
would be impossible, since they involve contradictory
formulations for Doubling: to use the suggested ordering
analysis for suffixed doubled forms, Doubling has to copy
to the left; on the other hand, the ordering analysis pre
sented above for prefixed doubled forms involves copying
to the right. At the same time, there can be no doubt that
only one process of Doubling is at work in the derivation
of both types of forms.
One might conceivably sidestep this problem by claiming
that Doubling copies either to the left or to the right,
depending on what sort of affix has been added. To be sure.
217
one would probably not expect a language to copy to the

right over a suffix and then move the suffix into word-
final position; nor would one expect it to copy to the left
over a prefix. But if it were the case that the direction
of copying could be dictated by the presence of a prefix or
a suffix, how would one deal with stems which are doubled
and have both a prefix and a suffix? It would seem that
some conflict would surely result. Such forms are found
frequently in Javanese and are illustrated by the following
examples of doubled Simple Locatives:
stem gloss doubled Loc.
dJUpuq go get ndJUpUq-(n)djUpUq-I
godok boil nggodog-(ng)godog-I
• • •
pangan food mangan-mangan-I

pllih choose mlll-mlll-I
Notice that in all of the above cases, the surface forms
are entirely predictable on the basis of the behavior of both
doubled prefixed forms and doubled suffixed forms. There is
no point where any conflict appears to be produced by the
combination of prefixation and suffixation on a single
doubled stem. In fact, the only point where any conflict
arises is within the two suggested ordering analyses, with
respect to the formulation of Doubling. It appears that in
order to predict the behavior of the prefixes, one must copy
to the right; but in order to predict the behavior of the
phonological rules such as HVL in the first member of forms
218
like ndjUpUq-(n)dJUpUq-I, one must copy to the left; and I

can see no way to construct any type of satisfactory analysis
using both formulations for Doubling. Thus, It seems reason
able to conclude that In the case of Javanese doubled forms,
some alternative to rule ordering other than those already
eliminated.should be sought..
6.6 Fortunately, a viable alternative has been sug
gested by Wilbur (1973). In her dissertation, Wilbur
examines a number of different reduplication processes In
various languages, including Javanese, and argues that in
a great many cases where reduplicated forms exhibit appar
ently aberrant behavior on the part of. phonological rules
(i.e. rules not applying where their environment is met or
over-applying to segments in other environments), this
behavior is not really aberrant at all, In that it apparently
serves the purpose of maintaining identity between the orig
inal stem and its reduplicated copy.*’ Furthermore, she
argues, none of the traditional devices used for treating
the exceptional behavior of phonological rules in reduplicated
forms (exception features, boundaries, and rule ordering)
can adequately cover all of the different types of exception
ality.
On the strength of the above, Wilbur suggests that
there is a universal tendency among languages to maintain
Identity between the original and copied parts of redupli
cated forms; and in order to implement this principle in
219
grammars, she suggests positing global constraints on phono

logical rules which can function to maintain the desired
identity between copy and original by allowing the rules
to over- or under-apply, as may be necessary for each spe-
7
cific environment. This, she argues, will not only produce
the correct surface forms and at the same time capture the
functional role of what would otherwise be looked at as
"exceptional" behavior of the phonological rules, but would
also allow one to keep the morphological and phonological
components of the grammar separate, with the relevant redu
plication processes applying, like most morphological pro
cesses, before the operation of the phonological rules of
the grammar.
6.6.1 It appears that Wilbur's proposed identity
constraints could be utilized to successfully account for
the observed identity between both members of affixed, as
well as non-affixed, doubled stems in Javanese. In suffixed
doubled forms, Doubling would apply, along with (nearly all
(cf. chapter 5)) other morphological processes of the lan
guage, providing the Input to the phonolpgical rules.
Subsequently, the phonological rules would be allowed to
scan their Immediate Input forms, but, at the same time,
these rules would have the power to look back into the
derivation to see if Input forms had undergone Doubling;
just in case they had been doubled, the rules would affect
both the copy and the original, either applying or not
220
applying to both, depending on whether the rightmost (i.e.

the one whose immediate environment would be altered by the
addition of a suffix) met or did not meet the environment
for the rule in question. Thus, for example, /a/ to off
would receive as its input forms like /mEdJa/, /mEdJa-mEdja/,
and /mEdJa-mEdja-nE/. The rule would then look back through
the derivation of each form to see if it had undergone
Doubling; for those forms which had been doubled, the rule
would scan the rightmost member of the doubling and, depending
on whether its environment were met, apply or not apply to
the stem-final /a/ of both members, regardless of whether
both meet or do not meet its environment. For stems which
had not undergone Doubling, the rules of course would apply
regularly, only when their specific environments are met.
It is clear how positing such constraints on the appli
cation of rules provides an elegant explanation for what
would otherwise appear to be quite exceptional behavior on
the part of phonological rules in simple doubled forms and
doubled suffixed forms by capturing the functional principle
behind this behavior: that Is, the preservation of surface
identity between the members of doubled stems. It remains to
be seen how global constraints on rules can be used in the
analysis of prefixed doubled forms.
As was seen In the preceding paragraphs, the motivation
for Including global constraints on the operation of rules
was to allow them to apply or not apply to a given segment in
221
both members of a doubled stem, even though, strictly .

speaking, the rule’s application or non-application was
in fact determined only on the basis of that segment’s
surrounding environment in one member.
6.6.2 The same approach can be applied to the pro
cesses involved in producing the surface realizations of
prefixed doubled words. 8 Thus, one would give the processes
involved in ng-PrefIxation the power to look back into the
derivation of their input forms to see if they had undergone
Doubling." Just in case they had, and the environment for
the rule in question was met by the leftmost member of the
stem, the rule would apply not only in that member where
its environment was actually met, but would also apply in
the other member. Using this approach with Replacement, for
example, the rule, whose environment specifies a voiceless
consonant preceded by its homorganic (assimilated) nasal
prefix, would receive as its input a form like /m-pangan-
pangan/. Seeing that this form has undergone Doubling, the
rule would apply to replace not only the initial consonant of
the first member with m, but also that of the second member,
giving mangan-mangan. If the input form had not undergone
Doubling, the rule would of course apply normally.
In order to derive surface forms like mbakar-(m )bakar,
a similar constraint could be placed on Voiced Consonant
Nasalization. Thus, when the rule receives an input such
as /m-bakar-bakar/, it can optionally over-apply to produce
222
a nasalized version of the initial voiced consonant of both

members of the doubled stem (i.e. mbakar-mbakar) or it can
apply regularly to produce mbakar-bakar.
The fact that the over-application of the rule is
only optional, and not, as with Replacement, absolutely
obligatory, may very reasonably be attributed to the func
tional motivation behind the over-application: namely, to
maintain identity between the two realizations of the stem
in doubled forms. That is, once Replacement applies, it
results in a very obvious alteration of the stem, replacing
its original initial consonant with a nasal; however, Voiced
Consonant Nasalization results in a relatively minor change,
creating a nasalized version of the initial consonant, but
at the same time leaving the stem more or less intact..
6.6.2.1 In spite of the success so far, certain prob
lems arise when one tries to use such constraints on phono
logical rules in the derivation of doubled forms like
ngadoh-(ng)adoh (i.e. those built on vowel- or liquid-initial
stems). The problem of course is that no phonological rules
apply in the derivation of such forms, since it would natu
rally be assumed that the ng- which occurs with such stems
represents the unassimilated nasal prefix. This leaves
nothing on which to impose any kind of constraints except
the process which is responsible for attaching the /ng-/
prefix to stems in the first place.
6.6.2.2 This could be done by claiming that the
223
ng-PreflxatIon rule has the power to determine whether an

input has undergone Doubling; if it has, the rule can add
the /ng-/ to both members of the doubled stem. The addition
to the rightmost member would have to be optional if only
vowel- or liquid-initial stems are involved. Notice that
taking this tack represents a departure from the treatment
of doubled suffixed forms, where only phonological processes
were subject to global constraints. Although there is some
doubt about whether the other processes involved in ng-
Pref ixation are phonological or morphological, there can be
little doubt that the Nasal Prefixation process responsible
for producing the actual basic segmental form of the prefix
is anything but morphological.
It seems, however, that if it is necessary to make
Nasal Prefixation global anyway in order to handle vowel-
and liquid-initial stems, it should at least be considered
whether positing global constraints for this rule would
allow one to account successfully for the derivation of
prefixed consonant-initial stems like mangan-mangan and
mbakar-(m )bakar, without involving the constraints on
Replacement and Voiced Consonant Nasalization proposed earlier.
6.6.2.3 As it turns out, assuming that Nasal Prefixation
is global does permit satisfactory treatment of all types of
stems. To do this, it must be claimed that, just in case
an Input stem has undergone Doubling, Nasal Prefixation has
the power to assign the nasal prefix to both members of the
224
doubled stem. In order to determine whether this double

assignment of /ng-/ is optional or obligatory, the rule
has to look ahead in the derivation to see whether the
input form in question will undergo Replacement; just in
case it will (i.e. it will end up having its initial con
sonant replaced by the nasal of the prefix) Nasal Prefixation
obligatorily assigns the prefix to both members of the
doubled stem, and all other rules apply regularly, with the
subsequent derivation of both members of the doubled stem
proceeding identically. Just in case the input form is
built on a stem which will not undergo Replacement and will
thus make it through to the surface without undergoing any
crucial alterations, Nasal Prefixation may optionally assign
the nasal to both members of the doubled stem.
Taking this tack, one could produce all the required
doubled prefixed forms in the language and at the same time
characterize the functional basis behind their ultimate
realizations. It will be noted, incidently, that this
analysis of doubled prefixed forms is really very similar
to the one suggested and rejected in 6.4.2 and 6.4.3 above.
The crucial difference between that approach, which used
boundaries, and the global one suggested here Is that the
former was purely mechanical, serving only as a vehicle to
produce the correct surface forms, but offering no explana
tion for the existence of such forms; the latter, on the
other hand, is thoroughly motivated by the tendency to
225
maintain identity between the two realizations of the stem

in doubled forms which was seen to be at work in suffixed
doubled forms also.
6.7 Admittedly, the above discussion of how Wilbur's
proposed identity constraints might be utilized within the
grammar of Javanese is rather sketchy. This is attributable
to the fact that it is difficult at this point to tell.how
innovative notions such as Wilbur's identity constraints
should actually be incorporated into a grammar. In spite of
the tentative nature of the preceding discussions, however,
several facts emerge:
Using global constraints allows one to eliminate the
problems encountered by an ordering approach with regard to
the formulation of Doubling as a right- or left-copying rule;
the global approach to doubled suffixed forms does not con
flict in any with the approach suggested for doubled prefixed
forms, since neither makes any claims about the direction in
which Doubling takes place.
The global analysis presented for prefixed doubled forms
may involve the claim that Doubling applies before Nasal Pre
fixation, so that rule can look back into the derivation of
its input forms in order to determine if they consist of
10
doubled stems. This, however, does not conflict with the
analysis of doubled suffixed forms, since the latter does not
depend in any way on the ordering of Doubling.
The above facts indicate that there Is no reason why
the global analyses suggested for prefixed doubled forms
and suffixed doubled forms should not be combined In the
derivation of doubled forms which are both prefixed and
suffixed (6.5), and thus it appears that a global approach
is able to successfully apply where the other, more tradi-»
tional approaches discussed earlier proved to be totally
inadequate.
227
FOOTNOTES
^There Is also another type of reduplication which
occurs In the language, Involving copying of the first con
sonant of the stem and addition of £ to this consonant to
form the reduplication syllable, as in t^tUkU, from tUkU
’buy’. This type of reduplication will not be specifically
discussed in this dissertation, but, it should be noted, its
analysis would appear to be quite straightforward and may
easily be extrapolated from the discussion of doubled forms.
2
In certain situations, positive and negative
exception features have also been used by linguists in the
analysis of doubled forms. In cases like that of Javanese,
however, where a number of different rules are involved,
and where, it shall be seen, the behavior of these rules in
doubled words belonging to a number of different categories
is involved, there is really no satisfactory way to use
exception features exclusively to even produce the correct
forms. One could perhaps mark the first member of a form
like mEdjo-mEdjo as being a positive exception to /a/ to o#,
but, as will be seen, this will not work in forms like
mEdJa-mEdj a-nE, so that, somehow, it would have to be claimed,
no exception feature is Introduced in the latter case. Even
if such problems could be overcome, the fact would still
remain that exception features, by their very nature, Imply
that the behavior of the marked form with respect to the
relevant phonological rules is irregular and unpredictable.
Making such a claim about Javanese doubled forms would
simply be incorrect, since, as has been seen to some degree,
and as will become even more obvious, the behavior of these
forms appears to be dictated by a desire to keep the two
realizations of stems in doubled forms identical. Thus, in
the following discussions, exception features will not be
considered as a possible way of dealing with doubled forms.
3
Throughout the following discussions, I will be refer
ring to "doubled suffixed forms", "doubled forms with
suffixes", "doubled prefixed forms", etc By this termi
nology, I do not mean to make any claims about the actual
order in which affixation and doubling take place within the
grammar; I wish only to refer to surface forms which exhibit
the effects of both processes. Thus, in the following, there
is no difference between "doubled suffixed forms" and "suf
fixed doubled forms".
Although in all cases my Informant seems to prefer
the doubled forms with the nasal prefix present in both
members, he will also accept forms such as ngadang-adang as
being perfectly correct, while he completely rejects mangan-
pangan.
228
One might reasonably wonder why the language would

ever (optionally) produce forms like mbakar-mbakar, with
the nasal occurring on both members of a doubled voiced-
consonant-initial stem, if it really is the case.that, at
the time copying takes place, the nasal and the stem-
initial consonant are still separate. One fairly plausible
way to account for this would be to claim that such forms
are being produced by analogy, on the model of those like
mangan-mangan. That is, because the nasal must be copied
as a part of a stem with a voiceless stem-initial consonant,
it might reasonably be assumed that speakers, in an effort
to keep the surface characteristics of all doubled ng-
Prefixed forms parallel, are also copying the nasal in both
members of the doubled form of stems like bakar. Note that,
even though the ordering analysis for ng-Preflxation in
doubled forms will ultimately be rejected, the above analo
gical explanation for the optional occurrence of forms like
mbakar-mbakar will still be entirely consistent with the
approach which will eventually be chosen at the end of this
chapter.
^Because her Javanese data was extremely limited,
Wilbur's discussion of Javanese reduplication is not really
very valuable in its specifics. This, however, turns out
to have no effect on the general validity of her claims
about the tendency to maintain identity in doubled forms.
Ît should be pointed out that in Hock 1973 a similar
proposal is made. In that paper, the author suggests the
use of a principle of "synchronic analogy" governing the
application of phonological rules in order to maintain
morphological identity among certain derived forms.
Q
Note that, as was mentioned in 3.5, there is no
way to tell whether Nasal Assimilation and Replacement are
phonological or morphological In nature. Since they could
be phonological and do lend themselves to the same type of
global application exhibited by the phonological rules
discussed In connection with suffixed doubled.forms, there
is no reason not to treat them as such (but cf. 6.6.2.1).
g
It also represents an extension of Wilbur’s notion
of the operation of such constraints, since, I believe, she
envisioned them as being used only to govern the application
of phonological processes; she did not, however, rule out
their extension to include morphological processes.
10
However, even this ordering may not really be neces
sary, since Nasal Prefixation, being global, would have the
power to look ahead to determine whether its input forms
would undergo Doubling in the future.
229
CHAPTER VII
HABITUAL-REPETITIVE
7.0 In the preceding chapter, It was claimed that a
simple ordering approach was not feasible for the analysis
of Javanese prefixed and suffixed doubled forms, and an
alternative approach was outlined based on a universal
tendency among grammars to preserve Identity between the
parts of reduplicated forms.
The present chapter will be devoted to discussion of
yet another class of Javanese words, Habitual-Repetitives,
or Hab-Reps, whose derivation involves the process of
Doubling and which can also undergo certain types of suf-
fixation and prefixation. There Is, however, a notable
difference between the ordinary doubled forms already dis
cussed In chapter 6 and Hab-Rep forms: In ordinary doubled
forms, the tendency to preserve identity between the two
members of the doubling was seen to play an important role
in governing the application of numerous processes within
the grammar; with Hab-Rep forms, It Is In fact by controlled
disruption of identity between the two members that the
category Is uniquely marked. For this reason, the analysis
of Hab-Reps represents a very Interesting area of Javanese
grammar.
7.1 The Hab-Rep of a given stem may be glossed as "be
continually doing or saying X ’, where X may either be the
action denoted by the basic stem or represent the basic word
230
Itself. Thus, from pEngln 'want, desire', the Hab-Rep

pEngan-pEngin can be glossed as 'continually wanting some
thing'; at the same time, Iba-IbU, from IbU 'mother' can
mean 'continually saying "mother"'. The specific semantic
force of a Hab-Rep depends greatly on its base and the
situation In which it is used.
Hab-Reps constitute what one might call a much more
"informal" category within the grammar of Javanese than any
other type of derived form. The application of the word
"informal" to Hab-Reps has motivation in several areas.
First of all, the use of such forms is most frequent in
informal colloquial speech. Although Hab-Reps can be used
in written Javanese, their usage there is probably limited;
it is only in Informal conversation that such forms are
freely created. Secondly, although Javanese derivational
categories are much freer than those of many other languages
with respect to which grammatical category (i.e. noun, verb,
etc.) can serve as their input, no other derivational process
in the language is as free as Hab-Rep Formation, which can
apply to Items belonging to any category. Finally, probably
because Hab-Reps are used mainly in Informal speech, Javanese
speakers are quite casual about what is the correct Hab-Rep
for a given form. Although there are certain minimum require
ments (which will be discussed below), within the bounds set
by these standards, a number of different Hab-Reps might be
acceptable for a single Input form.
231
7.2 The derivation of all Hab-Reps built on (non

affixed) steins whose final vowel is not /a/ may be des
cribed in the following steps
(1) The stem is doubled.
(2) The last vowel of the first member of
the doubling is changed to /a/.
(3) If the first vowel of the stem is /a/,
it is changed to /0/ in the first
member.
The resulting Hab-Reps are illustrated by the following
stem gloss Hab-Rep
adus take a bath Odas-adus
ball return bOla-ball
• • •
bos£n tired of bOsan-bos^n

• • « • • •
eleq bad Elaq-eleq

Eling remember Elang-Eling
IbU mother Iba-IbU
Iloq suitable Ilaq-Iloq
7.2.1 Looking only at examples like the above, wh:
it should be noted, illustrate the operation of Hab-Rep
Formation for the vast majority of stems (i.e. all those
which do not have /a/ as their final vowel), one can begin
to draw certain conclusions about the actual derivation
of Hab-Rep forms. First of all, it seems clear that the
232
process of Doubling does actually precede the vowel changes

described in Steps (2) and (3) above. If this were not the
case, there would of course be no way to get the member of
the stem which reflects the unaltered vocalism of the base.
Furthermore, it would seem very likely that this Doubling
process is precisely the same process which was responsible
for the other doubled forms of the language discussed in
chapter 6; not only do Hab-Reps and other forms involving
doubled stems share the characteristic manifestation of
Doubling, but they also have certain semantic features in
common: namely, the idea of plurality, which, in the case
of Hab-Reps, may be thought of as the plurality of actions
involved in repetitive or habitual behavior.
7.2.2 In addition, as one would expect to be the case
with most morphological processes, in the above examples,
Hab-Rep Formation seems to take place before certain phono
logical processes: specifically, those which assign mid
vowel tenseness. Thus, while the primary word bos^n 'tired
of' has a lax mid vowel because of the following ^ (2.6.3),
in the Hab-Rep form bOsan-bos^n, the first member has a
tense 0 as its first vowel— presumably because, at the point
where mid vowel tenseness is assigned, Hab-Rep Formation has
already changed the /^/ of the stem to /a/. Similarly, the
Hab-Rep Elaq-eleq, from eleq 'bad*, has a tense E as its
first vowel, even though the primary stem and the second
member of the Hab-Rep have lax e in this position. This may
233
be attributed to the fact that the first /E/ of the primary

stem, as well as the same vowel in the second member of the
Hab-Rep, is in the environment for laxing by Mid Vowel Har
mony (2.6.4.2.2), while the characteristic vowel change in
the first member of the Hab-Rep has destroyed this environ-
2
ment, so that MVH does not apply.
7.3 Taking only the examples discussed so far, Hab-
Rep Formation appears to be a very straightforward process,
whose operation might well be characterized as actually
involving the simple steps described in 7.2 above. The
formation of Hab-Reps for stems of the shape (C)V (C)V (C),
where V^ is not /a/, might thus be described as involving
the following processes:
Doubling
Vowel Change I (VC I): V ^ /a/ in the first
member
Vowel Change II (VC II): If V± Is /a/, V.J-} /0/
in the first member
Things become a bit more complicated, however, when the
Hab-Reps of stems which have surface a as their final vowel
are considered:
stem gloss Hab-Rep
dOlan engage in dOlan-dOlen
recreation
dj aran horse dj Oran-dj aran
dj aran-dj eren
djOran-djeren
234

lawas old lOwas-lawas
lawas-lewes
lOwas-lewes
Udan rain Udan-Uden
7.3.1 For the moment concentrating on stems like
dolan, which have /a/ only as the last vowel, it appears
that another vowel change must be added to those Just out
lined above, which would say that, for a stem of the shape
(C)V^(C)V2(C), where V,, is /a/, becomes /E/ in the second
member of the doubled form. For convenience, this will be
referred to as Vowel Change III or VC III.
7.3.1.1 It is worth noting that there appears to be a
good reason for the addition of this third vowel change
rule to those already posited for the derivation of Hab-Reps
from stems whose final vowel is not /a/. That is, because
the final vowel of stems like dOlan is already /a/, Vowel
Change I would be of no use as a marking device for Hab-Rep
forms (and VC II would of course affect only stems with two
/a/ vowels, like djaran, which are excluded from this dis
cussion) so that an input like /dOlan-dOlan/ would go through
the Hab-Rep marking rules with no alteration, to produce sur
face dOlan-dOlan. This form would, however, be in no way
distinguishable from, a simple, non-Hab-Rep, doubled form.
This undesirable development is of course avoided by the
335
addition of VC III, which changes the V^ in the second

member of the doubled stem to give Hab-Reps which are, like
dOlan-dOlen, for example, distinguishable from simple doub
lings .
7.3-1.2 Notice also that, if, as was concluded above,
Hab-Rep Formation takes place before the phonological rules
which assign mid vowel tenseness, the e which ultimately
occurs on the surface in the second member of Hab-Reps like
dOlan-dOlen can be originally inserted as /E/ and later laxed
by MVL 1 (2.6.1.2). It also would appear to be due to the
operation of the phonological rules that the Hab-Rep of
rEwang ’servant' is rEwang-reweng, with a tense penultimate
E in the first member, but lax penultimate e in the second
member; it can reasonably be assumed that the introduction
of final-syllable /E/ Into the second member by VC III and
the subsequent laxing of this vowel by MVL 1 creates the
environment for Mid Vowel Harmony (2.6.4.2.2) to lax the
penultimate /E/.
7.3.2 At this point, It Is worth digressing slightly
to note that, if the analysis suggested so far for Hab-Rep
forms Is correct, and If it is true that the vowel change
processes Involved apply before the rules which assign mid
vowel tenseness, one is provided with additional evidence
that the ordering approach to the analysis of doubled stems
with suffixes discussed above in chapter 6 Is not feasible.
236
Recall that In 6.3.4 It was suggested that an ordering

approach could be used to account for the deletion of h in
both members of doubled suffixed stems such as daJo-daJo-E,
from dajoh 'guest’, even though the intervocalic environment
for h-Deletion was met only in the second member. In order
to get the correct surface form, it was suggested, one would
only have to order Doubling to occur after h-Deletion. But
note that in 7.2.1 it was claimed that Doubling must precede
the vowel changes associated with formation of Hab-Reps, and
the latter take place before the operation of the mid vowel
tenseness assignment rules, and, specifically, before Mid
Vowel Harmony. However, In 3.3.2 it was claimed that the
mid vowel tenseness assignment rules must precede h-Deletion.
Thus, if one wants to treat the Doubling which occurs
In the formation of Hab-Reps and the Doubling which may take
place in the formation of other derived categories as being
one and the same process (and there can be little doubt that
both are accomplished by the same process) and still use an
ordering approach to the analysis of doubled suffixed forms
like daJo-dajo-E, an ordering conflict would arise: in Hab-
Rep forms, Doubling would have to precede MVH, which is known
to precede h-Deletion; in other doubled forms, Doubling would
have to follow h-Deletion. Such a conflict shows even more
conclusively that an ordering approach to the analysis of
Javanese doubled forms is not feasible.
237
7.3*3 Returning now to discussion of the examples

given in 7-3, the Hab-Rep forms of stems like djaran 'horse1,
which have /a/ in both syllables, must be considered. Stems
of this type stand out from those like dOlan by virtue of
the fact that they have three permissible Hab-Rep forms.
Thus, in 7.3, djaran is listed with dJOran-dJaran, dJaran-
djeren, and djOran-dJeren as possible Hab-Reps. My informant
will accept all of these variants as being good Hab-Reps if
asked specifically; however, he himself usually will only
produce one or two variations for each stem voluntarily.
It is important to note also that all three variants do not
occur with equal frequency. The frequency of those like
dj Oran-dJ aran seems to be greatest, with those like dj aran-
djeren being produced almost as often; those like djOran-
djeren occur least often and are always used alongside one
of the other variants.
7.3.^ The derivation of all these variant Hab-Reps
and, to some degree, their frequency of occurrence, may be
reasonably accounted for in terms of the analysis already
suggested, with only one small change in Vowel Change III.
That is, looking first of all at Hab-Reps like dJaran-djeren,
the change from /a/ to /E/ in both syllables of the second
member would seem to necessitate an extension of VC III to
Include stems with /a/ In both syllables. One might thus
alter VC III to read:
238
Vowel Change III (VC III): If Is /a/, V ^ > / E / in

the second member; If V^ is also /a/, /E/
the second member.
At the same time, Hab-Rep forms like djOran-djaran would
seem to be the result of applying VC II to the doubled stem,
while those like dJOran-dJeren could arise through the appli
cation of both VC II and VC III to the doubled stem. The
existence of all three variations and the relative frequency
of occurrence of each may be accounted for by considering a
few facts about the formation of Hab-Reps in general.
7.3.^.1 It has already been noted that It is Vowel
Change I (with or without accompanying operation of VC II)
which applies to the doubled stem in the derivation of the
great majority of Javanese Hab-Rep forms. Specifically,
VC I has been seen to apply with any stem whose is not
/a/, with VC II also applying, Just in case V^ is /a/.
Accordingly, nearly all Hab-Reps are characterized by, In
addition to Doubling, a change of the final vowel of the
leftmost member to /a/, with or without an accompanying
change of the penultimate vowel to /0/ In the same member.
Thus, it is almost always the case that in a Hab-Rep, the
leftmost member is the one which contains the characteristic
vowel changes, while the rightmost member reflects the
unaltered vocalism of the basic stem. For convenience, I
will refer to this as being the "general pattern" for Hab-
Rep forms.
239
Notice, however, that in Hab-Reps which have undergone

VC III, such as dOlan-dOlen, from dOlan, or djaran-djeren,
from dj aran, this characteristic relationship between the
members of the doubled stem is reversed: in Hab-Rep forms
of this type, the leftmost member reflects the vocalism of
the basic stem, while the rightmost has undergone the charac
teristic vowel change. Even though such forms have under
gone ah acceptable Hab-Rep vowel change, dictated by the
fact that they have /a/ as their V^, looking at them in terms
of the general pattern for Hab-Reps, one might regard them
as somehow being misfits in the general scheme of things.
As was mentioned earlier (7.3.1.1), with stems like
dOlan, speakers have no choice but to apply VC III in the
derivation of Hab-Reps; neither of the other vowel changes
can apply to such bases to produce a recognizable Hab-Rep
form, since VC I only operates on stems whose is not
/a/, and VC II only operates when V^ is_ /a/. With stems
like dJ aran, however, it seems that they do have a choice,
since such forms, having /a/ as V , are eligible to undergo
VC II, as well as meeting the environment (i.e. V^. is /a/)
for VC III. Thus, taking a stem like dj aran, one may,
instead of applying VC I' S. to produce dj aran-dj eren, apply
VC II to derive the variant Hab-Rep djOran-djaran. This
form is recognizable as a Hab-Rep and, at the same time,
conforms to the general Hab-Rep pattern in that the altered
vocalism appears in the leftmost member, while the rightmost
240
member reflects the original stem vocalism.

Thus, it could reasonably be claimed that Hab-Reps like
djaran-djeren arise through the regular application of VC III,
which is necessitated in the grammar in order to deal specif
ically with stems like dOlan, but whose environment is met
equally well by stems like djaran; and it might also be
claimed that speakers produce Hab-Reps like dj aran-dj eren
in spite of the fact that such forms are different from the
great majority of Hab-Reps by virtue of their reflection of
the basic stem in the vocalism of the leftmost member. On
the other hand, Hab-Reps like dJOran-dJaran do conform to
the general pattern, reflecting the basic stem vocalism in
their rightmost members. Such Hab-Reps can be produced only
by ignoring the fact that a stem like djaran meets the
specific environment for VC III and allowing it to undergo
VC II exclusively. The result Is an adequately marked
derived form which conforms to the general pattern for
Hab-Reps.
7.3.4.2 As was noted earlier, my informant actually
volunteers Hab-Reps like dj Oran-dJ aran a bit more frequently
than those like dJ aran-dj eren, although he always seems to
accept both if asked specifically. This preference is, it
seems to me, wholly understandable on the basis of the
comments in the preceding paragraph. That is, since Hab-
Reps like dj Oran-dj aran may be seen to conform completely
to the general pattern for Hab-Reps in that the rightmost
2*11
member reflects the original vocalism of the stem; and since

such forms are readily distinguishable as Hab-Reps by virtue
of the change of from /a/ to /0/; and, finally, since
such Hab-Reps may be easily derived by applying Vowel Change
II to the doubled stem, it is quite reasonable that speakers
would prefer to use this easy means of derivation instead of
producing Hab-Reps like dj aran-dj eren. The derivation of
the latter may be thought of as doing things the hard way in
all respects, since it involves applying VC III, which
alters two vowels of the stem and results in a form which
deviates from the general pattern for Hab-Reps by having the
characteristic changes occur in the rightmost member of the
doubled stem. Nevertheless, since the grammar has VC III
available, and stems like djaran, with two /a/ vowels, meet
the environment for that change exactly, it is reasonable
that forms like djaran-djeren should turn up very frequently.
7.3.^-3 There is one other variant type of Hab-Rep for
stems with two /a/ vowels which remains to be discussed.
This type is represented by dJOran-djeren and, as was noted
earlier (7.3.3), is the least frequently occurring of the
three possible variations, always occurring alongside at
least one of the other two variants. Hab-Reps of this type
appear to be derived by applying three vowel changes to the
doubled stem: V^ of the first member becomes /0/ by VC II;
and both vowels of the second member become /E/ by VC III.
242
There can be little doubt why such Hab-Reps occur only

infrequently, since their production Involves all the compli
cations inherent in deriving forms like dj aran-dj eren by
application of VC III and, consequently, any advantages
gained by taking the easy way out and applying VC II are
negated. That is, in a Hab-Rep like djOran-dJeren, not only
does a vowel change take place In the rightmost member,
violating the general pattern for Hab-Reps, but there Is also
a change In the vocalism of the leftmost member. The result
is a derived form in which the original stem vocalism Is not
reflected in either member.
Clearly, forms like d.1Oran-dJeren may be considered to
be "over-marked" as Hab-Reps. The existence of such forms is
explainable partly as a result of the fact that stems like
djaran meet the environments for both VC II and VC III; at
the same time, such over-marked Hab-Reps are sometimes used
to indicate a stronger or more emphatic Hab-Rep meaning than
either of the two other variations. Thus, at least in some
cases, speakers may associate the greater degree of Hab-Rep
marking with a greater degree of Hab-Rep connotation.
7.3.4.4 In summary, then, it seems that the analysis
of the formation of Hab-Reps suggested so far is able to
account reasonably for the derivation of the Hab-Reps for
a great number of different stems In the language. For
convenience, this analysis, for a hypothetical stem of the
shape (C)V^(C)V (C), is summarized as follows:
243
Doubling
VC I: If V2 / /a/, V2“^ the f’*1’3*' niember.

VC II: If V^ = /a/, V^—^ /0/ in the first member.
VC III: If V^ = /a/, /E/ in the second mem
ber; and, if V1 is also /a/, V^ -> /E/
in the second member.
7.4 The reader may have noticed that, so far, only the
Hab-Reps for stems with /a/ in final closed syllables have
been discussed. No mention has been made of those which are
built on stems like mEdjo ’table’ or loro ’ill, painful',
which have surface lax o derived from basic /a/ by the rule
of /a/ to o# (2.3.1.3) or by a-Vowel Harmony (2.3.3.3 and
2/3/4/2). It Is such stems which pose the most interesting
problems for the analysis of Javanese Hab-Rep forms.
7.4.1 Solely on the basis of the shared underlying
segments, one might expect all stems which have basic /a/ in
their final syllables or in both syllables to produce the
same types of Hab-Rep forms, regardless of whether the basic
final-syllable /a/ is in an open or closed syllable. Thus,
on the basis of Hab-Rep dOlan-dOlen, from /.dOlan/, one might
naturally expect to find *mEdja-mEdjE from /mEdja/. Simi
larly, on the basis of dJ Oran-dJ aran, dj aran-dj eren, and
djOran-dJ eren, from /djaran/, one might well expect lOra-loro,
^lara-lErE, and s10ra-lErE from /lara/. In fact, the Hab-Rep
of mEdjo is mEdj a-mEdjo, while the only acceptable Hab-Rep
for loro is lOra-loro. Other examples of Hab-Reps built on
stem s w i t h f i n a l b a s ic / a / i 5* I n c lu d e t h e fo llo w in g :^
stem gloss Hab-Rep

dOngo prayer dOnga-dOngo
dowo long dOwa-dowo
rEko plan, intention rEka-rEko
roso (the)' taste (of) rOsa-roso
sldo succeed in doing slda-sldo
In all of these examples, what seems to be happening is
that stems in final /a/ simply do not ever undergo the type
of vowel change (VC III) which changes /a/ to /E/ in the
second member of the doubled form; rather, those like mEdja-
mEdjo, from /mEdja/, appear to have undergone no vowel change
at all, while those like lOra-loro, from /lara/, appear to
have undergone VC II exclusively.
7.^.2 I can think of three alternative ways to account
for the surface shape of the Hab-Rep forms of words like the
above, all three of which are quite plausible, given the
derivation already suggested for Hab-Reps built on other
types of stems (7.3.4.4) and the phonological framework
discussed in other parts of this dissertation.
7.4.2.1 The first alternative would involve a slight
adjustment to VC III. The environment for the application
of this process would have to be altered so that it would
affect stems with final-syllable /a/ only if that syllable
were closed. Taking this tack, an Input form like /mEdJa-
mEdja/ would not meet the environment for VC III (or any of
245
the other vowel change rules either) and would thus come out
on the surface as Hab-Rep mEdJa-mEdJ o , with only the phono
logical rule of /a/ to o§ applying. At the same time, Vowel
Change II would be the only vowel change rule which could
operate on an input form like /lara-lara/, and would thus
do so to produce /10ra-lara/, which would then undergo the
regular phonological rules (cf. footnote 3) to give surface
lOra-loro. Thus, with only a small adjustment to VC III,
the suggested analysis could be used to derive the correct
surface Hab-Reps for all types of stems in the language,
including those in final /a/.
There are, however, problems with such an analysis
involving its general lack of explanatory powers. One
disturbing thing is that the proposed adjustment to VC III
involves the incorporation into the rule of a distinction
concerning the shape of syllables containing /a/, even
though such distinctions are totally irrelevant to the
formation of Hab-Reps from stems containing all other vowels.
Thus, one might ask, why should both bOdO 'stupid* and
bodol 'come out (hair)' both have Hab-Reps derived by VC I,
even though one has /0/ in an open final syllable and the
other has /0/ in a closed final syllable, while, just In
case the same forms had /a/ instead of /0/ In their final
syllables, their respective Hab-Reps would have to undergo
two separate derivational procedures.
246
At the same time, one might wonder why stems with final
closed syllables containing /a/ should be allowed to undergo
VC III, while stems with final /a/# either undergo VC II
exclusively or are the only stems in the language to undergo
no change at all. Stated differently, there would seem to
be no good ^sason why the language should use VC III to
produce dOlan-dOlen from /dOlan/ but exclude VC III from
applying to produce *mEd Ja-mEdJE from /mEdja/ and instead
allow only mEdja-mEdJo, with no morphological vowel change;
or, similarly, why it should allow VC III to apply to pro
duce Hab-Reps like djaran-dj eren as alternatives to those
derived by VC II (i.e. dJOran-dJaran) but at the same time
exclude the application of VC III to stems like djoro /drill'
(/djara/) to produce *djara-dJErE as an alternative to dJOra-
dj oro.
7.4.2.2 A second alternative approach which has a
certain amount of explanatory power suggests itself if one
considers the derivation of the Hab-Reps for stems like
/mEdja/ and /lara/ in terms of what appears to be the min
imum requirement for a recognizable Hab-Rep form: namely,
the requirement that a vowel change must take place so that
the left and rightmost members of the doubled stem are not
identical. Failure to meet this requirement would of course
result in Hab-Rep forms which would be indistinguishable
from other categories involving Doubling.
247
Thus, It might be claimed, a surface Hab-Rep like mEdja-

mEdjo meets this requirement without having undergone any of
the Hab-Rep vowel changes, simply by virtue of the difference
between the vowels of the left and right members which results
from the application of the phonological /a/ to o# rule In the
latter. Accordingly, one might claim that the vowel change
rules (specifically, VC III) have the power to look ahead
Into the derivation of an input stem whose is /a/ to see
If that form will eventually undergo /a/ to o#; if it will,
then no morphological Hab-Rep vowel change needs to take
place. Note that, here again, the identity constraint which
normally cuases /a/ to o§ to apply to both members of doubled
forms would have to be suspended. This suspension would not
be unnatural, however, if it specifically functions to pro
duce a Hab-Rep which is identifiable mainly by virtue of the
non-identity between the final segments of each member of the
doubled form.
The reader will note that this approach is nearly iden
tical to the first alternative discussed above, which involved
adjusting VC III to apply only with stems whose V^ Is not
stem-final /a/. In the second approach, stems of that shape
would also be excluded from undergoing VC III, but not, how
ever, simply because the rule’s environment does not allow
it to apply to stems In final /a/, but specifically because
stems which end in /a/ will eventually undergo '/a/ to o#,
making any other marking unnecessary. Thus, it seems to me,
248
the second approach offers some explanation for why there

should be a difference in the Hab-Reps of stems with final
/a/ versus those built on stems like /dOlan/, for example,
with final /aC/, while the first approach simply produces
the correct surface forms mechanically without explaining
anything.
'Although the second approach seems to be in this
respect preferable to the first, difficulties arise in using
it to analyze Hab-Reps like lOra-loro, built on stems,
which, like /lara/, have /a/ in the position and stem-
final /a/ as the second vowel. The problem is that, if one
claims that Hab-Reps like mEdja-mEdJo have gone through
Hab-Rep Formation without undergoing any vowel change simply
because their surface Hab-Reps will be recognizable as Hab-
Reps anyway (because of the operation of /a/ to o# in the
second member), then there is no good reason why the deri
vation of Hab-Reps for stems like /lara/ should not follow
the same course to become surface *lara-loro, which would
seem to be Just as recognizable as a Hab-Rep as a form like
mEdJa-mEdJo. In other words, why should the operation of
VC III be governed by such considerations of necessity, while
that of VC II (which apparently changes penultimate /a/ to
/0/ In the first member of lOra-loro) seems to be oblivious
to the same considerations? At the very least, one would
expect to find slara-loro as a permissible variant alongside
lOra-loro.
2*19
7.*1.2.3 There is a third analysis, which,<although it

too has certain failings, seems to offer a better explana-*
tion for the existence of the Hab-Rep forms in question and
the exclusion of certain others. This approach would involve
ordering the vowel changes suggested for Hab-Rep Formation
to apply after the phonological rule of /a/ to o# has already
applied. The Immediate effect of this ordering would be to
provide a distinction between stems with absolute final /a/
and those which have /a/ in final closed syllables. That is,
the former would undergo /a/ to ojt and enter the vowel change
rules, not with /a/ as their final vowel, but with the o
produced by /a/ to o# occurring in both members of the doub
ling, in accordance with the identity constraint; the latter
would of course not meet the environment for /a/ to ojt and
would thus enter the vowel change rules with their original
vocalism.
7.*1.2.3.1 Although it might at first glance appear
that this ordering is merely an indirect way of getting at
the distinction between /a/ in closed final syllables versus
stem-final /a/, and thus would be comparable to the first
alternative discussed above (7.*1.2.1), the ordering approach
can explain facts about Hab-Rep forms which were left unac
counted for by the earlier approach.
7.*1.2.3.2 Recall that the first approach was criticized
for not giving any explanation for why stems in final /a/
should either exhibit no apparent vowel change at all, as
250
with mEd Ja-mEdJo, or undergo VC II, as with lOra-loro, but

should never undergo VC III; while those in final /aC/
either undergo this change exclusively (as in dOlan-dOlen)
or as one premissible Hab-Rep variant alongside others
(i.e. djaran-djeren, dj Oran-dJ aran, and dj Oran-dJ eren).
If one assumes that /a/ to off applies before the Hab-Rep
vowel changes take place, a stem like /mEdja/, for example,
would enter the vowel change rules in the shape /mEdJo-mEdJo/,
so that, at this point, its final vowel would be /o/ and not
/a/.** In this respect, then, such a form could naturally
fall together with all other stems whose final vowel is not
/a/, such as /bOdO/ 'stupid* and /Iloq/ 'suitable', and thus
quite regularly be expected to undergo VC I. This process,
it will be recalled, produces a change of the final vowelof
the first member of the input stem to /a/, so that, with an
input like /mEdJo-mEdJo/, VC I would give as its regular out
put mEdJa-mEdJo, Just as it produces bOda-bOdO and Ilaq-Iloq
• •
from /bOdO/ and /IlOq/, respectively. Thus, although Hab-

Rep forms like mEdJa-mEdJ o appear to have undergone no_ vowel
change, they would, using this analysis, like the vast major
ity of stems in the language, have undergone VC I. At the
same time, a doubled stem like /mEdJo-mEdJo/, which has
undergone /a/ to off, would of course not ever be expectedto
fall together with those like /dOlan-dOlan/, which meet the
requirements for VC III (i.e. V£ Is /a/) and thus would not
ever be expected to end up with a Hab-Rep comparable to '
251
dOlan-dOlen.
7.4.2.3-3 Another point of criticism for the approach
which simply altered the environment for VC III so that it
could not apply to stems with final /a/ was that it gave
no explanation for why stems like djoro 'drill*, from
/djara/, have only the single permissible Hab-Rep form
dJOra-dJoro, apparently derived by application of VC II,
m
while those like dJ aran have three possible Hab-Reps, with

one of them resulting from the operation of VC III. Given
the above explanation for Hab-Reps like mEdj a-mEdj o , it is
easy to see how the ordering of /a/ to of before the Hab-
Rep vowel change rules can account for this difference.
If a stem like /djara/ has undergone both Doubling and
/a/ to of by the time it gets to the vowel change rules, it
will of course have the shape /djaro-djaro/ and be eligible
to undergo VC I but not VC III, which operates only on stems
with final-syllable /a/. At the same time, such a form
would regularly meet the environment for VC II, which would
accordingly change its V^ to /0/, with resulting /djOra-
djaro/. The rule of a-Vowel Harmony would subsequently
5
apply to give surface djOra-dJoro.
7.4.2.3.4 Mote also that the difficulty which arose
with the second alternative approach, discussed in 7.4°2.2,
does not arise here. Taking the ordering approach, all types
of stems in the language (including those like /mEdja/)
252
undergo some morphological vowel change in conjunction
with Hab-Rep Formation, so that there would be no reason
to question the application of VC II In forms like lOra-
loro.
7-5 Thus, it would seem that the ordering of /a/ to

o# before the vowel change rules connected with the forma
tion of Hab-Reps can succeed In areas where the other
approaches suggested earlier failed. The ordering analysis
does, however, suffer from at least one major shortcoming,
resulting from the fact that languages normally do not order
phonological processes to apply before morphological pro
cesses unless some purpose Is served by this ordering.
Thus, for example, in chapter 5 , 1 argue that Javanese may
avoid homonymy in Elatives by applying Elative Formation
after certain phonological rules have already applied.
7.5.1 In the case of Hab-Rep forms, however, there
does not at first glance appear to be much gained by the
suggested ordering. In fact, it might be claimed that the
ordering of /a/ to of before Hab-Rep Formation makes Hab-Rep
derivation more complicated by causing stems which have
basic /a/ as their undergo two different vowel change
processes, when it might seem more natural to allow Hab-Rep
Formation to treat all stems with basic /a/ in their final
syllables identically. That Is, presumably, if /a/ to of
did not apply before the Hab-Rep vowel change rules, the
Hab-Reps of stems of the shape /CaCa/ as well as those like
253
/CaCaC/ would undergo VC III to give Hab-Rep forms like

CaCa-CECE and CaCaC-CECEC.
7.5.2 On the other hand, if one looks at it from a
slightly different angle, this: result of ordering /a/ to o#
before Hab-Rep Formation could be viewed as being beneficial.
Recall that, as has been noted before, the vast majority of
bases fit into the category of stems undergoing VC I, with
or without accompanying operation of VC II (I.e. they do not
have /a/ in their final syllables), and thus they have Hab-
Reps with altered vowels in their leftmost members with the
original vocalism of the stem reflected in the rightmost
member. Accordingly, speakers probably have some feeling
for this particular type of Hab-Rep formation as being the
normal one, while VC III, applying to only a very specific
type of stem to alter the vowel In the rightmost member of
the doubled stem, might be looked at as being somewhat unu
sual. One would in factprobably expect to find that, in
the course of time, VC I would become generalized and VC III
might be eliminated altogether, JLf It were not for the fact
that stems like /dOlan/, for example, which do not have a
penultimate /a/ which can become /0/ by VC II, would come
through Hab-Rep derivation unmarked and indistinguishable
from other doubled stems, if there were not some alternative
marking process available. And, looking at the language in
general, it Is difficult to see how this particular problem
could be handled in any way except by leaving VC III as a
254
part of Hab-Rep Formation.

7.5.3 Note, however, that the grammar of Javanese does
offer a way to enable stems in /a/ to undergo the prevalent
method of Hab-Rep formation (i.e. VC I, with or without VC
II) and to avoid having them undergo the Hab-Rep marking
associated with VC III. Since /a/ to ojt is in the grammar
anyway, speakers would only need to apply this rule before
the Hab-Rep vowel change rules to enable stems like /lara/
or /mEdja/ to Join the vast majority of stems in the lan
guage in undergoing VC I. In other words, I am suggesting
that perhaps the unusual ordering of /a/ to ojt before Hab-
Rep Formation does indeed serve a purpose within the grammar
of Javanese: that of regularizing as far as possible the
method used for deriving and marking Hab-Rep forms by allowing
as many stems as possible to undergo VC I.
7.6 By this time, the reader may be beginning to wonder
whether there actually is a process of /a/ to ojt in the lan
guage, as is argued in chapter 2, or whether it might not be
the case that stems like loro are, after all, derived from
/#loro/ or /*laro/. If the latter were true, the Hab-Reps
of stems of this type would present no problems whatsoever;
their derivation would be entirely regular and predictable,
without the use of any unusual orderings.
I have found myself thinking along these lines at various
points In the course of writing this dissertation; however, I
find the arguments given in chapter 2 for the derivation of
255
of forms like loro from basic /lara/ by means of /a/ to o#

and a-Vowel Harmony to be so convincing that there is little
room for doubt that Javanese speakers have some feeling for
stem-final /a/ as a basic segment. At the same time, however,
it is possible that the language is at a transitional stage
on the way toward accepting /o/ as a basic segment which is
realized as a in certain environments. The likelihood of
such a development is supported by the fact that, in surface
Javanese forms, the effects of /a/ to o# show up (almost—
cf. 2.9.1.1) exclusively in the non-derived form of the pri
mary stem, since the addition of any suffix (except Imperative
-o) destroys the word-final environment for the rule’s opera
tion. Thus, one might eventually expect to find speakers
setting up a correlation between "primary stem" and "primary
(I.e. basic) phonological stem", which could lead to the
acceptance of surface shapes like loro or mEdJo as basic
shapes /loro/ or /mEdJo/.
Although such speculations are interesting and can form
the basis for future research, the question of the degree to
which /o/ may be basic is unfortunately beyond the scope of
this dissertation; for the present, the analysis of /a/ as
being the basic final vowel in forms like mEdJo and loro
will have to suffice. At the same time, considering all the
advantages and problems involved with the alternative
approaches to the analysis of the Hab-Rep forms of stems like
mEdjo and loro, I would say that the explanatory powers of
256
the ordering approach are greatest and seem to outweigh any

disadvantages it may possess.
7.7 At this point, as a conclusion to this final chap
ter, some" attention will be devoted to certain forms which
show the effects of both Hab-Rep Formation and other deriva
tional processes. Such forms are interesting both for their
complexity and becuase they may serve as a sort of testing
ground for analyses which have been posited in this chapter
and in those preceding.
Consider the following examples, where, for each stem,
the Hab-Rep.(including variant pronunciations), doubled Causa
tive, and Hab-Rep Causative (again, including variant pronunci
ations) forms are given
stem gloss derived
adoh far Odah-adoh
ngadoh-ngadoh-akE
ngOdah-ngadoh-akE
mbodoq-mbodo-qakE
mbOdaq-mbodo-qakE
gawE job, task gOwa-gawE
nggaweq-nggawe-qakE
nggOwaq-nggawe-qakE
ombo large Omba-ombo
ngambaq-ngamba-qakE
257
ngOmbaq-ngamba-qakE or
ngambaq-ngembe-qakE
roso (the) taste (of) rOsa-roso
ngrasaq-ngrasa-qakE
ngrOsaq-ngrasa-qakE or
ngrasaq-ngrese-qakE
salah mistaken sOlah-salah or
salah-seleh or
sOlah-seleh
nj alah-nj alah-akE
njOlah-njalah-akE or
nj alah-nj eleh-akE
tjObo try tjOba-tjObo
njObaq-nJOba-qakE or
njObaq-njObe-qakE
7.7.1 It Is striking that, as the examples given above
illustrate, as long as a given stem is not one of thoselike
tjObo or ombo, derived by the operation of /a/ to o# ona
basic stem whose final segment Is /a/, the method used In
forming the simple Hab-Rep and the Hab-Rep Causative is appar
ently the same. That. Is, from adoh, for example, the simple
Hab-Rep Is Odah-adoh, with VC I and VC II having applied;
and the Hab-Rep Causative Is, accordingly, ngOdah-ngadoh-akE,
again with the effects of VC I and VC II being evident. Simi
larly, salah has several different Hab-Reps, and, again. Its
Hab-Rep Causative also has corresponding variants. 7
258
7.7.2 On the other hand, with stems like ombo, which

have two basic /a/ vowels, with the last in stem-final posi
tion, while there is only one acceptable form for the simple
Hab-Rep, the Hab-Rep Causatives show variations— the same
variations which occurred for stems like salah (i.e. those
with /a/ in a final closed syllable). A similar situation
exists for tjObo: although the simple Hab-Rep tJOba-tJObo
appears to have undergone no vowel change, the Hab-Rep
Causative has clearly undergone VC III to become njObaq-
nJObe-qakE.
7.7-3 What would seem to be happening with stems like
the latter, which have basic /a/#, is that in simple Hab-Reps
one finds the expected outcomes discussed in earlier parts
of this chapter for stems of the shape /CV^Ca/, where V may
or may not be /a/; but in Hab-Rep Causatives, such stems
appear to act exactly like those which are of the shape
/CV^CaC/. This change in behavior is not at all unusual if
one makes the plausible assumption that, at the point in the
derivation where the Hab-Rep vowel changes take place. Caus
ative Formation has already applied, adding its characteris
tic suffixes (chapter 4). As was noted earlier (7.2.1), by
the time Hab-Rep formation takes place, Doubling must have
already applied, so that the assumption would have to be
made that Hab-Rep Formation takes as its input doubled Causa
tives.
7.7.^ The effect of this ordering on the derivation can
259
be seen by considering the possible derivation of the Hab-

Rep Causatives ngOmbaq-ngamba-qakE or ngambaq-ngembe-qakE.
Beginning with the basic stem /amba/, Causative Formation
would apply to produce /amba-qakE/. This intermediate form
could then undergo Doubling to become /ambaq-amba-qakE/.
Notice that for Causatives, as well as for Locatives, as
was noted in 3.3.1, Doubling may be thought of as copying
not only the basic stem, but also taking along the initial
consonant of the Suffix. Alternatively, as was suggested in
footnote 3 to chapter 4, the Causative or Locative stem-form
ing process for adding ^ or n to vowel-final stems could, in
keeping with the identity constraints proposed in chapter 6,
be formulated as applying to both members of doubled stems.
At this point, I would probably want to apply ng-Pre-
fixation, as discussed in 6.6.2.3, which is a necessary part
of well-formed Causatives; however, the exact point at which
the ng-Preflx is added is not especially relevant to the
present discussion, since it does not interact in any way with
Hab-Rep Formation and thus could be added at any time in
the derivation after Doubling has taken place (7.2.1). Thus,
at this point in the derivation of Hab-Rep Causatives
ngOmbaq-ngamba-qakE and ngambaq-ngembe-qakE, one would have
the intermediate form /ngambaq-ngamba-qakE/. The Hab-Rep
vowel change rules would scan this form and see that the
first member of the doubled stem does not end in /a/#, but
instead ends in /aq/ so that it is just like non-derived
260
inputs like /salah/, for example, and not like non-derived

/amba/. Thus, the form would be eligible to undergo VC III
to produce ngambaq-ngembe-qakE. Alternatively, the speaker
could take the easy way out and simply apply VC II to change
penultimate /a/ to /0/, producing the variant form -
ngOmbaq-ngamba-qakE.
7.7.5 It is interesting that, while three alternative
forms were given for simple Hab-Reps based on stems like
salah (7*7), one of which was of the type described earlier
as being over-marked in that it had the effects of both VC II
and VC III, only two variants show up for Hab-Rep Causatives.
That is, while there is a simple Hab-Rep sOlah-seleh, there
does not appear to be a Hab-Rep Causative *njOlah-nJeleh-akE♦
One way to account for this gap would be to suggest that,
perhaps, with all the complicated marking that necessarily
goes on in the derivation of complex forms like the Hab-Rep
Causatives, speakers would tend to avoid complicating things
further by over-marking, which would result In a form whose
basic stem is reflected In neither member of the doubling.
While this is only speculation, it Is perhaps supported by
the fact that, if pushed into It, my Informant will accept
a form like *njOlah-njeleh-akE as being possible, but unlikely,
instead of rejecting it outright.
7.7.6 Returning now to the derivation of Hab-Rep Causa
tives for stems like /amba/, given above, one can see how
any of the approaches to the analysis of simple Hab-Reps
261
discussed earlier in this chapter can be used to derive the

more complex forms. Recall that the first approach involved
reformulating VC III so that it would operate only on stems
with /a/ in closed final syllables. Using this approach,
the derivation of Hab-Rep Causative ngambaq-ngembe-qakE
would proceed as follows:
/amba/
Causative amba-qakE
Doubling ambaq-amba-qakE
ng-Prefixation ng-ambaq-ng-amba-qakE
Hab-Rep
VC I---------------- --- -
VC II (ngOmbaq-ngamba-qakE)
VC III (ngambaq-ngembe-qakE)
7.7.6.1 The second approach discussed above (7.4.2.2)
involved allowing VC III to look ahead into the derivation
of input forms to see if they would eventually undergo /a/
to o#; if this were the case, VC III would not be allowed to
apply. Note, however, that as long as a suffix is present,
the word-final environment for /a/ to of can never be met.
Thus, for Hab-Rep Causatives, VC III would be allowed to
apply to derive forms like ngambaq-ngembe-qakE, In precisely
the manner illustrated in the preceding derivation.
7.7.6.2 At the same time, the third alternative - ■ - •
approach, applying /a/ to of before Hab-Rep formationtakes
place, would achieve precisely the same effect, since, once
262
the Causative suffix had been added to the stem, the form
would not be eligible to undergo /a/ to o# anyway, because
of the rule's word-final environment. Thus, the actual
derivation would be exactly like that given above, with
/a/ to o# being given a chance to apply at some point before
the Hab-Rep vowel change processes, but not applying in the
derivation of Hab-Rep Causatives because its environment
can not be met.
7.7*7 It appears, then, that, interesting as they may
be otherwise, Hab-Rep Causatives do not give one a basis for
deciding between the various suggested approaches to Hab-Rep
Formation for stems like loro or mEdjo, with all three appar-
Q
ently working equally well. At the same time, however, it
appears that the various analyses discussed in earlier chap
ters lend themselves to the derivation of even such complex
forms as Hab-Rep Causatives.
263
FOOTNOTES
^Note that, at this point at least, I do not neces
sarily wish to claim that the actual derivation of Hab-Reps
consists of the above steps in the order given.
2
It is interesting to compare this ordering of Hab-
Rep Formation with respect to the phonological rules to
that of Elative Formation with respect to the same rules.
With Elative Formation, it was claimed, the morphological
process may be ordered after the phonological rules in order
to avoid homonymy which might result from the neutralization
of underlying contrasts inherent in Elative Formation. With
Hab-Rep forms, however, even though the vowel changing pro
cesses result in even more neutralization, this ordering does
not seem to obtain. The difference may, it seems to me, be
attributed to the fact that, in Hab-Reps, because the charac
teristic vowel changes take place (usually) in only one
member of the doubled stem, the basic stem is always reflected
on the surface in the other member, so that it does not matter
to what extent underlying contrasts are destroyed.
^Notice that in Hab-Rep forms of this type, the oper
ation of the identity constraints suggested in chapter 6 seems
to be itself somehow constrained to keep the final vowel of
the first member of slda-sldo, for example, from becoming o
along with the final vowel of the second member to produce
*sIdo-sIdo. The suspension of the tendency to preserve
identity between the parts of doubled forms would, however,
be wholly natural in such cases. That is, since non-identity
between the two members seems to be the distinguishing charac
teristic of Hab-Reps, and since the a of the first member of
slda-sldo is the only thing that distinguishes the Hab-Rep
from the simple doubled form, one would not expect the iden
tity constraint to operate to destroy this distinction. (In
7.4.2.3 below, however, an analysis of forms like slda-sldo
will be suggested in which the identity constraint does
operate to change both occurrences of /a/ to o, but the Hab-
Rep vowel change rules subsequently operate to undo the
effects of /a/ to off in the leftmost member.) .It.should dlso
be pointed out that, in general, the Identity constraints do
seem to work, even In Hab-Rep forms, as long as their opera
tion does not interfere with the proper marking of such
forms. The usual operation of the Identity constraints can
be seen in the examples in 7.7 below, as well as In Hab-Rep
Locatives (cf. footnote 8) such as ngOda-ngado-I , from adoh.
4
In keeping with the identity constraint analysis,
/a/ to off would presumably apply to both members of the
doubled stem.
264
5
Note that, if this analysis is accepted, it repre
sents evidence that the effects of /a/ to o# and a-Vowel
Harmony can not be attributed to the operation of a single
process (cf. 2.7.2).
Âlthough it is not indicated in the examples, in

certain Causatives the realization of the nasal prefix in
the second member is optional (cf. chapter 6).
7
For discussion of the fact that this stem has three
Hab-Rep variants but only two Hab-Rep Causative forms, cf.
7.7.5 below.
O
Note that Hab-Rep Locatives behave exactly like
Hab-Rep Causatives, the only difference being in the presence
of the Locative suffix and the application of certain phono
logical rules in Locatives (which are not relevant to the
present discussion). Hab-Rep Causatives and Hab-Rep Loca
tives were the only complex Hab-Rep forms my informant would
regularly construct; although he would accept a few Hab-Reps
which had also undergone other types of derivation, these
were only sporadic.
265
REFERENCES
Dudas, Karen and Margie O'Bryan. 1972. A wider perspective
on apparent phonological exceptions. Presented at the
Summer Meeting of the Linguistic Society of America,
Chapel Hill, N.C.
Dudas, Karen. 1974. A case of functional phonological
opacity: Javanese elative formations. Studies in
the Linguistic Sciences 4:2.91-111.
Hock, Hans Henrich. 1973. Exceptions and synchronic
analogy in Sanskrit. Studies in the Linguistic
Sciences 3:2.81-101.
Horne, Elinore C. 1961. Beginning Javanese. New Haven:
Yale University Press.
. 1963. Intermediate Javanese. New Haven: Yale
University Press.
. 1974. Javanese-English Dictionary. New Haven:
Yale University Press.
Kiparsky, R. V. P. 1971. Historical linguistics. A survey
of linguistic science, ed. by William Orr Dingwall,
576-649. University of Maryland: Linguistics Program.
Kisseberth, Charles W. 1973. The interaction of phonological
rules and the polarity of language. Presented at the
Indiana Conference on Rule Ordering, Bloomington.
and Mohammad Imam Abasheikh. 1974. A case of
systematic avoidance of homonyms. Studies in the
Linguistic Sciences 4:1.107-124.
266
Pigeaud, Theodore. 1938. Javaans-Nederlands Handwoordenboek.

Batavia: Groningen.
. 1948. Nederlands-Javaans Handwoordenboek. Batavia:
Groningen.
Sumukti, Rukmantoro Hadi. 1971. Javanese morphology and
morphophonemics. Unpublished doctoral dissertation,
Cornell University.
Uhlenbeck, E. M. 1949* De structuur van het Javaansche
morpheem. Bandoeng: A. C. Nix.
. 1953a. Woordverdubbeling in het Javaans. Bijdragen
tot de Taal-, Land- en Volkenkunde 109.52-61.
. 1953b. The study of word classes in Javanese.
Lingua 3-322-354.
. 1954. Verdubbelingsprocedes bij het Javaanse
werkwoord. Bijdragen tot de Taal-, Land- en Volkenkunde
110.369-387.
. 1955. Over woordvorming in het Javaans. Bijdragen
tot de Taal-, Land- en Volkenkunde 111.286-307.
. 1963. Review of Beginning Javanese, by Elinore
C. Horne. Lingua 12.69-86.
. 1964. A critical survey of studies on the languages
of Java and Madura. Koninklijk Instituut vor Taal-,
Land- en Volkenkunde Bibliographical Series 7.
The Hague: 's-Gravenhage.
Wilbur, Ronnie Bring. 1973. The phonology of reduplication.
Unpublished doctoral dissertation. University of Illinois.
267
VITA
The author was born In Chicago, Illinois on May 14, 1946.
She attended public schools in Downers Grove, Illinois,
graduating from Downers Grove Community High School In 1964.
In 1968 she received a B.A cum laude from Cornell College
in Mt. Vernon, Iowa, having majored in German and spending
a semester In Germany with the Experiment in International
Living. During her senior year, she was elected to Delta
Phi Alpha, the German honorary fraternity. She entered the
Department of Linguistics at the University of Illinois In
1969. Prom Fall, 1969 until Pall, 1971, she held a National
Science Foundation Traineeship in Linguistics. Prom 1971
until 1973, she was employed as a Research Assistant by the
Institute for Research on Exceptional Children. During the
school years 1973-74 and 1974-75 she held a Research Assist
ant ship with the Department of Linguistics, funded by a grant
from the University of Illinois Research Board to Professor
Charles Kisseberth. In 1973, the author was elected to the
Honor Society of Phi Kappa Phi.

Phonology of Modern Javanese

Uploaded by

Copyright:

Available Formats

Phonology of Modern Javanese

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Phonology of Modern Javanese

Uploaded by

Copyright:

Available Formats

INFORMATION TO USERS

The following explanation of techniques is provided to help you understand

5. PLEASE NOTE: Some pages may have indistinct print. Filmed as

Xerox University Mcrofsllms

DUDAS, Karen Marie, 1946-

Xerox University Microfilms, Ann Arbor, Michigan 48106

_______________ KAREN, TIARJE..DUDAS______ ____________

E X T FT IT '.!)_._...... THE PHONOLOGY AND MORPHOLOGY OF MODERN JAVANESE

BE A C C E P T E D IN P A R T IA L EC I.F I 1.I..M E N T O F T H E R E Q U IR E M E N T S FOR

T H E DEG R EE DOCTOR OF PHILOSOPHY

Ifirecto r of T li/iis R ese arch

involves the degrees to which the areas of phonology and

of E. M. Uhlenbeck. What I would claim for the present

Javanese scholars translated Sanskrit works Into their

origin came Into Javanese. As the island developed

a person by lowering the status of another person who

Throughout this dissertation, in citing Javanese forms,

1.5 Within morphemes, the following syllable types

Notice that, although syllables may begin with more than

Javanese primary morphemes begin with the following types

as well as £ and w plus 1

as though the nasal and consonant are

1.6 Within words, stem-penultimate open syllables are

(i.e. part of the archaic Kawi literary language), and

There are, however, some logical exceptions to this

is the only morphological process in the language which

Javanese, the latter being the marking of a given category

However, for the purposes of this dissertation, I will

stem gloss Active

tUlis write nUlis

baj ar pay,salary mbaj ar

djogo guard, watch ndj ogo

gOleq get, seek nggOleq

adj ar learn ngadJ ar

IbU mother nglbU

llwafc go past, by way of ngllwat

1.8.1.2 A second very commonly occurring prefix is

OmbE drink dl-OmbE

first consonant of the stem. There is in fact little doubt

primary gloss Elative

this unorthodox approach, I am of course making no claims

dOjan like, enjoy dOj an-dOJ en

1.8.3 There are numerous processes of suffixatlon used

primary stem ends in -o, the suffixed form of the word

stem-final consonant. In such cases, the derived stem

klUwung rainbow klUwUng-E

stem gloss Causative

lUngguh sit nglUngguh-akE

primary root vowel

1.8.3.2.3 Locatives represent the third and final

djjrfrO deep ndj^ro-nl

stem gl03S Locative

1.8.3.3.1 The suffix -o may be added to vowel-final

stem gloss Substantive

without accompanying vowel changes. For clarity I will

morphological process in Modern Javanese. By the use of

doubled stem and a Causative or Locative suffix, If the

appear after the final vowel of the first member of the

been made here to provide a general description of Javanese

_ KAREN, TIARJE..DUDAS