Phonological Analysis A Functional Approach

Donald A. Burquest pdf download

https://ebookgate.com/product/phonological-analysis-a-functional-approach-donald-a-burquest/

★★★★★ 4.7/5.0 (33 reviews) ✓ 155 downloads ■ TOP RATED

"Excellent quality PDF, exactly what I needed!" - Sarah M.

DOWNLOAD EBOOK
 Phonological Analysis A Functional Approach Donald A.
Burquest pdf download

TEXTBOOK EBOOK EBOOK GATE

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Phonological Architecture A Biolinguistic Approach 1St

Edition Edition Bridget D. Samuels

https://ebookgate.com/product/phonological-architecture-a-
biolinguistic-approach-1st-edition-edition-bridget-d-samuels/

ebookgate.com

Handbook of Near Infrared Analysis 3rd Edition Donald A.

Burns

https://ebookgate.com/product/handbook-of-near-infrared-analysis-3rd-
edition-donald-a-burns/

ebookgate.com

Signs and Symptoms Analysis from a Functional Perspective

1st Edition Dicken Weatherby

https://ebookgate.com/product/signs-and-symptoms-analysis-from-a-
functional-perspective-1st-edition-dicken-weatherby/

ebookgate.com

A Microscale Approach to Organic Laboratory Techniques 5th

Edition Donald L. Pavia

https://ebookgate.com/product/a-microscale-approach-to-organic-
laboratory-techniques-5th-edition-donald-l-pavia/

ebookgate.com
Mathematical Analysis A Straightforward Approach Second
Edition K.G. Binmore

https://ebookgate.com/product/mathematical-analysis-a-straightforward-
approach-second-edition-k-g-binmore/

ebookgate.com

Economic Analysis of the Law Selected Readings 1st Edition

Donald A. Wittman

https://ebookgate.com/product/economic-analysis-of-the-law-selected-
readings-1st-edition-donald-a-wittman/

ebookgate.com

Speech and Thought Representation in English A Cognitive

Functional Approach 1st Edition Lieven Vandelanotte

https://ebookgate.com/product/speech-and-thought-representation-in-
english-a-cognitive-functional-approach-1st-edition-lieven-
vandelanotte/
ebookgate.com

Primer of genetic analysis A problems approach 3ed Edition

Thompson J.

https://ebookgate.com/product/primer-of-genetic-analysis-a-problems-
approach-3ed-edition-thompson-j/

ebookgate.com

Syntactic Analysis and Description A Constructional

Approach 1st Edition David Lockwood

https://ebookgate.com/product/syntactic-analysis-and-description-a-
constructional-approach-1st-edition-david-lockwood/

ebookgate.com
 - - - - 'S'" ~~ ; if:?
SIL
international ()~G/~GJ
~ ;.!
-'

PHONOLOGICAL
ANALYSIS
Phon ological Anal ysis
A F uncti un a) A ppr ()~lc h
Seco nd editi o n, revised
A FUNCT;ONAL APPROACH
H uman languagt: is a r~ll1arkabk phl;nOlll~nOll. It:;
study <.:ominllcs to bl.:! a source of f:ls<.:in<1lion ~l11d
delight. Dr. Donald Burquest. professor o flingliistiL's
at the University ofTt:.x "s al Arlington, developed this
foundation:ll h'::.l(tbouk du ring years ofhclping slUdt!IHS Second Edition
overCOIll(! tilt.:: feelings or dismay that new phonology i
:.;tlH.'knts experienc..:: whcn <.:onfrolllt!d by a tn ..1SS orraw t

( phom.:t1<.: data . While working through th~ matt:rial. th..:

student is kd through tile: steps of organi "ling data and ,i
i.s inlrudut.:(:d to panic..:ular theories ror [:ttt::r in-depth
-I
'""" I specializ'ltion .
",.' Thl.: i.tutht)[' c.'<pi.llllis un the prc:vilHIS t;dition ol"lllis
lext by adding inlroductions to AU1\)SLglllcI11;!I
II
PIHlIlok)gy and rvktri~~l1 Phonology, He l];,ls ;l1so i
indudcd a sl..'rics of problems al lhe end of 111 0S[ ,I ;
"',1
chapters thaI pru v id~ an opportuni ty I;')]' Ihe sludenl to
appl y thl: inti lrlll;ltiull in [hat ~haph,;r,
This tL'xlhlHA is intcmkd for usc in an Lipper
division inll'olillctury t.:IHII'Se in phonolog.y, prcp<lring ,J
lhe slmk'ill ['or further study ill aspeL:IS current or ,j
theury, Tht.:
t.:dilillll ,
prt.:s~ l l1 voluillt.: is a l\.' visioll 01' the II)l)X
J, ,
!
;

OJ ,j
-- ,- '- _.- -- - - .. - . -- --- -- C ; ,,
~ ,- , - -
;:u ',t

o
c •
m
ISI,N 1-:i:ii>71 -()i>7-4
en .\ . .
90000 ~ :\
-~
I , '
~
"

. ,! .• II I ; 11 11
,
"
Donald A. Burquest /"----
I
----~
 Phonological Analysis
A Functional Approach

Second Edition

Donald A. Burquest

Editors
Mary Ruth Wise
William R. Merrifield

Assistant Editors
Marilyn A. Mayers
Mary Huttar
Bonnie Brown
Martha Jakway

Production Staff
Laurie Nelson and Margaret Gonzalez, Compositors
Hazel Shorey. Graphic Artist

L
SIL International
Dallas, Texas --
" ,.
./
 © 1993, 1998,2001 by SIL International
First edition 1993
. Second edition 1998
Second edition, revised 2001

Library of Congress Catalog No: 93-84153

ISBN: 1-55671 -067-4

Printed in the United States of America

06 05 04 03 02 10 9 8 7 6 5 4 3 2

All Rights Reserved To

No pat1 of this publication may be reproduced, stored in a retrieval sys- Edie, Janie,
tem, or transmitted in any form or by any means-electronic, mechanical, and Jeff-thanks
photocopy, recording, or otherwise- without the express permission of SIL
International, with the exception of brief excerpts in journal articles or re-
views.

Copies of this and other publications of the SIL International may be ob-
tained from
International Academic Bookstore
7500 W Camp Wisdom Road
Dallas, TX 75236

Voice: 972-708-7404
Fax: 972-708-7363
E-mail: [email protected]
Internet: http://www.sil.org
·,

Contents

Preface to the Second Edition xi

Preface . . . xiii
....
1. Preliminaries to Analysis .
The place of phonology in linguistics . 1
Foundational principles . 3 ....
Phonological fearutes . 10
Excursus on redundancy. 21
Excursus on rule formalisms. 24
2. Phonemic Analysis. 31
Basic concepts in phonemic analysis 31
Procedures for phonemic analysis . 37
The procedure applied . . . . . . 40
Ph0nological universals: Vowel and consonant systems 45
Identifying phonetically similar segments 51
Free variation _ . . . . . . . . . . . 59
Phonotactics, neutralization, and underspecification . 60
Problems .. . .. ... . 71
3. Morphophonemics . . . . . . . . 81
Phonologically conditioned alternations 81
Lexical alternations. . . . . . 85
Morphologically conditioned alternations 86
Phonological representation and rules. . 88
Analytical procedure and morphophonemics . 91
Case srudies. . . . . . . . . . . . . . . 91
•
•
viii Contents Contents ix

• Positing underlying representations 102 Epenthesis. .......... . 287

Problems . . . . . . . 103 Interpretation of ambiguous segments. 289
...
~-
Reduplication . . . . . '",, ' 290
4-6 Natural Phonological Processes Problems 297
4. Conditioning by Surrounding Segments. 115 References. 30!
Sources of natural processes. 115
Assimilation. 117 Index . . . 311
D issimilation 129
Summary . . 130
Problems . . 131
5. Conditioning by Syllable Structure 147
Syllables and consonant strength . 147
Syllable structure. . . 149
Ambiguous segments ISS
Arobiguous sequences . 157
Processes conditioned by syllable structure 168
Length . . . . . . . . . . . . . . . . 180
Syllable-domain phenomena and processes. 185
Problems . . . . . . 197
6. Conditioning by Larger Units 205
'- Word-level phenomena , . . 206
Utterance-level phenomena 221
Problems . . . . . . . . 225

7-8 Three-Dimensional Models

7. Autosegmental Phonology . 231
Introduction .. _ _ . 231
The establishment of auto segmental tiers 233
Assimilation of one segment to another. '237
Processes affecting the phonological word. 241
Processes affecting larger units. 257
Excursus on the feature tree. 265
Problems . . 268
S. Metrical Phonology. 273
IntToduction. . . 273
Syllable stnlcture. . 275
Phonological feet. . 279
Phonological words. 281
Ex.trametricality 282
 Preface to the Second Edition
This revised edition continues the same emphasis of the first, presenting
pretheoretical principles useful for phonological analysis as a prelude to
study of current phonological theory. Still, after five years of use, most ev-
ery pedagogical work can profit from revision, and Phonological Analysis is
no exception. I extend my thanks to those students and instructors who have
,
made suggestions for improvement.
Sad to say, there were portions of the first edition in which the content or
its significance was not expressed clearly. I have rewritten those sections
that were reported to be difficult to understand or apply in an attempt to
ameliorate this difficulty.
In light of the continued focus on providing techniques for applying phono-
logical concepts to language data, for chapters 2 through 8 of this revised text
a series of problems has been included to provide an opportwifty for the stu-
dent to apply the information of that chapter in dealing with phonological data
from languages not included in the text itself. These problems are for the most
part taken from a compilation developed by David and Judy Payne and up to
now distributed only informally. Relevant phonology problems are at times
hard to come by, and the Paynes' generous willingness to allow their work to
be included in this volume is acknowledged with gratitude.
Finally, to assist the beginning student to make the transition to advanced
work more easily, two additional chapters have been added to introduce two
models of phonological theory that have now become standard,
Autosegmental Phonology and Metrical Phonology. Although the material
in these chapters is highly introductory, it is hoped that the infortnation will -,
provide the sort of overview of major concepts that will illustrate something
of the state of the art of phonological theory and analysis. As an aid to en-
able the reader to see more easily the sorts of contributions these proposals
 xiv Preface

• when faced with raw data. We see a focus on methods of analysis as essen-
tial to the beginning student.
The maj or point of the discussion throughout this book is that phonologi-
cal analysis must be based strongly upon phonetic plausibility. As in all of
sCience, the question to answer is: Why are the observable phenomena the
way they are? In some cases, academic honesty demands we admit that we
do not yet know. At this stage in the development of phonological theory
we are not able to give a solid phonetic reason for every phonological pro-
cess found in natural language. But the majority of the processes presented
in this text are' given a sound phonetic basis and represent to a large extent
the present state of understanding as represented in current literature on 1
phonology.
For most of the data in this book, we have used the 1989 International Preliminaries to. Analysis
Phonetic Alphabet. Data which were originally published with symbols
other than these have been modified in these pages, in an effort to Simplify ~
the task of the beginning student in understanding what we are illustrating Phonology is the study of the sound systems of language. Every lan-
with each example. In a few cases, however, Americanist symbols such as guage makes use of a vast number of sounds; the question we attempt to
[c] and m have been used. deal WIth m phonology 15 he'.': !h o!;t sounds are org~r:iz :::d i.n a given lau-
In addition to the students who have struggled through various versions guage to make communication possible.
of this work we are especially grateful to the fo llowing who have helped by
critically reading and evaluating earlier versions of all or parts of it: Doris
Bartholomew, Burt Bascom, Frase r Bennett, Alan Buseman, Karen The place of phonOlogy in linguistics
Buseman, Scott DeLancey, Ken Gregerson, Marv Mayers, Bob Mugele,
Joyce Overholt, Debbie Paulsen, Audra Phillips, David Silva, John Watters, Phonology is just one of several a.i?ccts vi" j;i ;l~ l i~~ l..: ;; . i ~ :~ ; \.:]::i:::d lu m ha
Kathie Watters, and Henry Whitney. We are particularly indebted to Judy aspects such as phonetics, morphology. syntax. a~d d iscourse.
Payne for significant input in both editing and content. Morphology, for example, is the study of word fonllation, i.e., how p r~­
We are also very grateful to Mary Huttar for producing the index. fIXes, roots, and suffixes adjoin to tonn words. In many lanouac~es ~ 0
the:
Surprisingly. in spite of all the study devoted to phonology, there are still SOun ds are affected when these parts of words come to!.!.ether. And. of
interesting discoveries to be made and controversies which remain. In fact. course, the words themselves, even where there :.tre 1..l0 prcfi~'\.es or ::;uf1ix.cs.
with three-dimensional models of phonology the subject has e,xperienced a are not mere abstract constructs. btlt nre actually pronounc:::J as sequence.s
resurgence of interest. Human language is a remarkable phenomenon. Its of sounds. The way the sounds of rhese words are org:lllized and are
study continues to be a soarce of fasc ination and delight. We hope these affected in a language is the substance of phollolo£!y. -
pag~s will be helpful in effe-ctively enabling more to join in the enterprise. Syntax, another aspect of langll .3g~. is concern::d"'~' ith the oi"[!3Iliz,nio:l of
words into phrases, clauses, and s ent~nc~s. [)isL"\)~!.r!) e study h,;$ to d ...) with
e ~e c. larger units such as paragraphs and text:.. \\'ht!n words ~n:: FronuunceJ
wlthm their larger units, there are often uiiler~nces in tbe way tbe wo rds
sound, as opposed to the way they are prOlwunced in isobt io!1. F',.K o.:~. X.:!r:1 -
pIe, there may be fast speech omiss ions or n ! odit1cari o:~s of cert:lio sOt: nus
at the beginning or end of words or utterances. and tht:rc arc- u~uallv SCll-
ren.ce intonation phenomena not evidt:ni. whet! words are pronounced ~d Gne .
This. too, is within the domain of phonology as i! relates to syn!Jx.
2 Pre~jmiltarjes to Ana(ysis
3
'; .:~. '.': .:;; : ; ; :
Because it deals purely with sound, phonetics is fundamentally tied to The orthographies of some other languages, like" Sparush';'are:'closer to
phon61o'gy:' so 'much so ,' thaf it~:i sometimes difficult for the beginning stu- this ideal. The Spanish onhography is not phonetic, though"if-is ' someiimes
dent to comprehend the distincti.on between the two. Their foci are distinct, heralded as such. For example, most dialects of Spanish distinguisp the two
ho wever. Pnonetics is principally concerned with the physical realization of d s?unds m a word such as hUlldido 'sunken', Sl'eakers pronoUnce ttie first
the sounds . of language, while phonology primarily concerns itself with the d somewhat like an English d and they pronounce the d between the vowels
mental (Qut subconscious) organization of the sounds of a particular like the English Ih in Ihe. These two sounds, which are phonetically dis-
language .. :..·.. ,. :.." . : . tmct, are perceived by native speakers of Spanish as phonologically the
. As. an ill~tratiqq of tills, conbder the pronunciation of the I in the nor- same. And they are rightly represented by the same alphabetic. character in
mal oL casu'iI 'proimnciaiion of the two related words below: Spanish. Most linguists agree that the optimal onhography for' it languaae is
!
. , phonological rather than phonetic. There are, in addition. many sociolo;ical
o
( 1) 3. Mlte -, . or .~olitical factors which enter into discussion Oll orthography, but -.te will
b. Writer . c: "'.: .'1lOt 'dISCUSS these here. . . . . ' . .'
. -'.
The two instances of what we write as I in these words are pho":etically. dis,- :; : ' .: ' .' I ;~. ... ~

tincL !nth!: lin (h) 'the outWard flow of air is completely stopped ~t the Foundational priiici"les .
pronunciation 'of the consonant.- The consonant' is, in fact, called a ,s:rcir.(pr
PLOSNE). !n (1.1,)); however; ··the air ·f low is never completely interr:up.ted. ,., -- There are two foundational principles which are panicul~rly imp-o~tant to
Rather, the tongue is quickly flapped up against the ridge behind the teeth. understanding the patterns exhibited by sound systems. ' : ..
This consonant is called a FLAP. The typical speaker of English does not
necessarily perceive the two t sounds as different, because they are orga-
nized or systematized as the same sound within the phonology. But their
1L/ Sounds are affected by the environments in which' they occur

physical realizations are distinct.- They are thus distinct phonetic entities, The first principle to be considered is that SOUNDS TEND TO BE AFFECTED
but the same phonological entity, BY THE ENVIRONMENTS IN WInCH THEY OCCUR. The following data from
There are two branches of the srudy of phonetics which are relevant to Asheninka, an Arawakan language of Peru (data taken from Payne 1981 .
phonology: articulatory phonetics and acoustic phonetics. Articulatory pho- tranSCription modified) illustrate this tendency:
netics is the srudy of the production of the sounds of language focusing on
what' articulatory mechanisms . are used (tongue, lips, oral cavity, vocal (2) impoke ' he will come' ilaOle . he will chop'
cords, lung ' air, etc.). A'co\i;li2 phonetics is the srudy of the audible lan- imparje 'he will fall' intaje 'he will burn'
guage signal by means of .its measurement with electronic instruments. impanale
' . ... .
'he will plant' ill lowe 'he will fell (trees)'
AC9ustic phonetics is of\en'~iD.ybked in ,discussions of the perceptual distinc' . ,.
tiveness of the souqd~ of langUage. ·C"' .. · In Ash6ninka, the nasal consonants m and II are somewhat restricted when
Phonology has. a special relationship with orthography, the alphabetic ' followed by another consonant, such that we never fmd a sequence of ml or
sys tem used to represent the sounds of a language. It is a commonly held lip. Here the effect of the environment on sounds is readily seen. Nasal con-
misconceptioll that an ideal orthography is phonetic, If that were the case Sonants which precede p are pronounced at the same point of articulation
then the I sound in wrile would be represented differently from the I sound (the same place in the mouth) as the p, i.e., they are labial. Nasal conso-
in wriler. In fact, an optimal orthography should correspond not to the pho- nants which precede I are pronounced at the same point of ar£i.culation as
netic system, but to the phonological system as it is subconsciously stored the I, i.e., they are alveolar. Thus, the pronunciation of the nasal consonants
in tbe mind of the native speaker. ' This is the case with our English orthog- m or II is affected by the consonants which follow them,
rapby with .-espect to the two t sounds. The two kinds of I are written the /" . This principle has a narural explanation in the nature of sound produc-
same, even though tbey are really different sounds phonetically. Even so, in tion. When a sequence of sounds is produced, the sounds do not occur just
several other respects English onhography is rather remote from a phono- one after another as discrete units. Rather, they OCCur in a Single wave of
logically optimal writing system, \ sound, with each individual sound making its own contribution, . and they
 4 Preliminaries to Analysis Foundational principles 5

tend to blend into one another where they abut. Thus, in Asheninka, when mouth determines the degree of closure of the vowel.' The points on the hori-
articulating a nasal which immediately precedes a consonant, the lips and zontal parameter indicate point of articulation, specifically whether the closure
tongue anticipate the pronunciation of the following consonant, and the na- is forward or back in the mouth, and also whether the lips are rounded or not.
sal consonant 111 or n is pronounced at the same point of articulation as that These parameters generally define the space" in the oral cavity for vowel
consonant. This process of assimilation of a nasal consonant to the conso- sounds. Since the three vowels in (3) are spaced at extremes of the articulatory
nant which follows it is one of the most common phonological processes space available, the disrinctions between the three vowels are more easily pro-
found in the languages of the world. duced and perceived. We call this a symmetrical system.
The effect of the principle that sounds tend to be modified by their environ- A vowel system with more than three vowels may also be symmetrical,
ments is e~,phasized throughout this book. After dealing with the basic meth- with the vowels arranging themselves more or lessin a V-formation. The
ods of phonological analysis in chapters 2 and 3, in chapter 4 we examine five vowel system in (4) is the most conunon systel'; and is considered to
various wavs in which sounds are influenced by the sounds which surround be symmetrical. It is the vowel system found in Spanish:
them (the data in (2) is one example of this); in chapter 5 we examine ways in
which the strucrure of the syllable influences sounds; and in chapter 6 we ex- (4) Front/ Ba.c k/
amine ways in which phonological strucrures larger than the syllable affect Unrounded Rounded
sounds. In each case the principle is the same: SOUNDS TEND TO BE MODIFIED
BY THEtR ENVIRONMENTS. High i u
Mid e 0
Sound systems tend to be symmetrical Low a

The second principle to be considered here is that SOUND SYSTEMS TEND Notice that the mid front vowel e is shown on the chart as occurring more
TO BE SYMMETRICAL. (The hedging here is important because sound sys- towards the back position than the high front vowel i. The more open the
tems are not always symmetrical, at least not in every way.) The prinCiple mouth is ~ Le., for lower vowels , the less forward the tongue goes. Con-
.- of symmetry manifests itself in two ways: in the inventory of sounds and in versely, with high vowels there is more closur~ in the oral cavity and the
the way sounds are affected by their environment. These two ways are de- tongue moves farther forward.
scribed below. The American English vowel system shown in (5) is a rathor repkte one
? among: tbe world's languages. It is also considered io hs' SYlnmetrical, since
Symmetry in the inventory of sounds. First, the inventory of sounds in the vowels are more or less evenly distributed throu!lhout the spac~ avail-
a given language tends to be s)'TI1ffietrical. For example, if a language has -able to vowel sounds. r/_/",.-""..-/",..////

only three vowels. they are likely to be i. 1I. and a. as is the case for many
Quechua languages (spoken in the Andean region of South An1erica). (5) Front/ Bock!
Unrounded Rounded
(3 ) Front/ Back!
Unrounded Rounded High i 1I
J U

High j ZI e v
Low a Mid
E J
The system in (3) is characterized as a symmetrical system because the ~

three vowels are maximally spaced; i.e., they are near the extremes of the Low a
articulat( 'ry possibilities for y.~wels. The two parameters in a vowel chart such
IThroughout this volume, following the tradition within phonokl,!;ic:d th(;U0'. we: g.:neralJy u.'·;e
as (3) defIne the position of t1,e lips and tongue. The points on the vertical pa- the ternlS IIIGH, MID, and LOW to refer to the height characteristics of vo\vds . In phllnerics high
rameter indicate lOngue height: How close the tongue comes to the roof of the vowels are sometimes characterized as CLOSE and low VO\\ e!s :is OPEN .
6 Preliminaries to Analysis Foundaa"onal principles
7
(The begil)ping student should be aware that phonetic symbols vary some- Consonantal systems are affected by the sarne principle of symmetry. A
what in different texts on linguistics. For the vowels above, for example, I relatively symmetrical consonantal system like that in English is given in
is in some works transcribed I, and a as a.) (7).
Some current analyses of English vowels consider it to actually be a six
vowel system with contrastive vowel length. That is i, e. a, :J, 0, and u are
(7) Labial Dental Alveolar Retroflex (Alveo-) Velar
long vowels, and I, e, re, a, and u are corresponding short vowels. The long Glottal
vowels are indeed pronounced with measurably more duration. For the palatal
present introduction, however, the eleven vowel qualities given in (5) are Plosive
useful as an approximation of the vowel system, as an illustration of the
range of vowel sounds recognizable by the beginning student, and as a fur-
voiceless p t if k ?
voiced b d dj g
ther example of a symmetrical vowel inventory. .
In contrast to the symmetrical systems in (3), (4), and (5), we would not Fricative
expect to frod three vowel or five vowel systems like those in (6). These are voiceless f e s J h
asymmetri.cal and unexpected vowel systems since they do not use maxi- voiced v 0 z 3
mally opposed places of articulation or do not spread the vowels out
Nasal III 1/
proportionately. No known language has such a system.
Liquid
(6) a. Front/ Back!
Unrounded Rounded Glide w j (w)

High u For describing the pronunciation of consonants. positions along the hori-
Mid 0 zontal parameter are points of articulation from front to back in the mouth.
Low a Positions along the vertical parameter are manners of articulation. Moving
from top to bottom, the consonants generally go from those with the great-
b. Front/ Back! est degree of closure (plosives) to those with the least closure (semivowels
Unrounded Rounded or glides). The w occurs twice on the chart (once in parentheses) because its
point of articulation is actually labiovelar, that is to say, both LABIAL and
High VELAR..
Mid e 0 Observe thlt for each voiceless plosive or fricative, except glottals, there
re J is a corresponding voiced plosive or fricative. (Voiced glottal plosives or
Low a fricatives are articulatorily inapossible.) Thus, the system is symmetrical for
these manners of articulation. If there were arbitrary gaps such that only
c. Front/ Back! some voiceless plosives or fricatives had voiced counterparts, then the sys-
Unrounded Rounded tem would be asymmetrical.
As for nasals, most languages will have at least a labial and an alveolar
High i nasal consonant. To maintain symmetry they will also often have nasal con-
I sonants at some of the same pOints of articulation as other plosives, ,
Mid e fricatives, or affricates. For liquids (I and r sounds) and glides (semi vow- 1
e els), the symmetry with other manners of articulation is not as evident. That ""{
Low re is because there are fewer possible points of articulation for sounds made in
this manner. I
i
 Preliminaries to Analysis Foundational principles 9
8

An example of another symmetrical consonantal system, that of Spanish, b. Labial Alveolar (Alveo )palatal Velar
is given in (8),
~ .
Plosive
(Alveo )palatal Velar voiceless k
(8) Labial Dental
voiced b d3
Plosive Nasal m lj
p t k
voiceless
voiced b d g
Flap r

Fricative f s x
Glide w

Affricate /j c. Labial Alveolar (Alveo)palatal Velar

Nasal m II JI
Plosive
voiceless p k
Liquid voiced d u
~
Lateral
Flap r Nasal m
Trill r
Lateral
Glide w j
Glide j
in contrast to (7) and (8), nonsymmenical systems like those in (9) are not
expected, though occasionally a language may hove some sort of asymmetry. Synunetrical patterns* uch as those illustrated in (3), (4), (5), (7), and (8)
seem to have their basis in the use of language for conununication. Spe-
Labial Alveolar (Alveo)palatal Velar cifically, by distributing sounds uniformly in the pbonological spac" , and by
(9) a.
utilizing the extremes of such space (this is especially true for vo",els), tbe
Plosive differences between the sounds are most easil:, perceived. This !(:;sens the
voiceless p t burden on the language user in deci~ Dg the message. It is true that
voiced d g articulatory distance does not in every case have a direct correlation with
the acoustic distance upon which this perceptual interpretation of sJmmetry
Nasal is based. But to a large extent there is a reasoDJ.ble match which gives such
a notion credibility.
Fricative f s A second likely factor influenCing S)l1unetry 111:.1)' b~ th~ nature (If articu-
lation itself. There is a shared corn...1.1onality, for ex ample. in CO:lsonants
Liquid mad!! at the alveolar ridge or in those made w:th voicing or not. 11 may be
Lateral more efficient to utilize the same basic articulatory gt:stu re repeated ly. (This
Flap r
is symmetry from the viewpoint of sound production. )
Glide j
Symmetry in the way sounds are affected by their environm( nts, In
addition to its effect on the inventoty of sounds, the principle of symmetry has
its effect also on the ways in which sounds are affected by their environments.
'-
10 Preliminaries to Analysis Phonological features II

Specifically, sounds occur in set', and all members of a given set tend to be notions as syllables, units of stress placement, and s'!ntence intonation patterns.
affected in the same way in the same environment Similarly, all members of In fact this volume is organized specifically around that notion. Chapter 4
such sets tynd to have the same effect on other sounds which occur in their deals with phonolOgical patterns which are best characterized as beinab
due to
enviromnent. simple segmental influence; chapter 5 deals with pilonological patterns which
F or example, in Isthmus Zapotec, a language of Mexico (Merrifield et a1. are best understood as being determined by syllable structure; and chapter 6
1987:80, transcri ption modified) voiced plosives are realized as the corre- deals with phonological patterns which refer to larger units (phonological
sponding voiceless plosives when following the possessive prefix J-, as words or utterances).
indicated in (10). This is an example where the members of the set of A typical hierarchy proposed for phonology is the following (after Pike
voiced plosives are all affected in the same way in the same environment, 1967). That is, segments are grouped into syllables; syllables are grouped
i.e., following the voiceless consonant f into stress groups (also referred to as phonological words); and stress
groups are grouped into breath groups (also called utterances). There may,
(1 0) bere 'chicken' J-pere-be 'his chicken' of course, be phonological patterns making reference to even larger units,
biutlzaJ 'guest' J-piuuza -be 'his guest' such as paragraphs and even texts, but we do not discuss them here.
daa 'mat' J-taa-be 'his mat'
doo 'rope' J- too-be 'his rope' (12) A breath group (utterance)
geta 'tortilla' J-keta-be 'his tortilla' stress group (phonological word)
J-kami;ja-be 'his shirt' syllable
garni3aJ ' shirt' I
Y segment
An example where all members of a set have the same effect on another f ....!u<<-5
sound occurring in their environment is seen with the English plural, here The lower end of the hierarchy may be considered to extend beyond the
set off by square brackets. segment; that is, the phonological segment itself can be understood as being
a simultaneously occurring combination of phonological features.
(11) map!s} lab!z}
cllb! z} Features and the organization of sounds
ClIp!S}
lIIat!s} fad!z}
hil!s} bid!=} As with the higher levels, the notion that phonemes are combinations of
pack!s} rag!z} phonological features also has a lengthy tradition, as shown by the follow-
leg!z} ing quote from Bloomfield (1933 :79):
tick!;}
Among the gross acoustic features of any utterance, _certain ones.
In this case the English plural occurs as the voiceless fricative s follow- are distinctive, reCiirring in recognizable and relative ly constant
ing all membe~s of the set of voiceless plosives, while the voiced fricative z shape in successive utterances. These distinctive features occur in
occurs following all members of the set of voiced ploslVes. VOlceless lumps or bundles .. .
plos ivcs form a natural class and by the principle of symmetry have a com-
parable effect on voicing of the followmg frlcallve. VOiced plOSlves, too, Similarly, Trubetzkoy (1969:36) speaks of "the sum of the phonologi-
form a NATURAL CLASS and likewise have a comparable effect on the fol- cally relevant properties of a sound. " He goes on to identify such properties
as including the articulatory characteristics of oral closure, raising of the
lowing fricative.
velum, aspiration. voicing. rounding, and various points of articulation.
While the concept of the phonological feature has more recently been elab-
Phonological features orated and emphasized within the theory of phonology known as
GENERATIVE PHONOLOGY, it is an idea which has historically been widely
It has long been known that phonological patterns in language are orga- accepted in principle.
nized hierarchically. As far back as Swadesh 1934 we find reference to such
 12 Pre/imin3ries to Analysis Phonological features 13
The concept of phonological features has shown itself historically not only In the sarne way all vowels which are made with the same relative tongue
in explicit definitional statements like the above, but also in traditionally ac- height are on the sarne row, and those at the same point of articulation and
cepted phonological practice. Specifically, the idea of using articulatory charts with the same rounding are in the same column.
to display phonological segments is conceptually dependent upon the notion of
phonological features as well. That is, phonological segments are not posi- b. Vowels
tioned randomly on such charts, but rather they are organized in such a way as
to represent the complex of phonological features which makes up the seg- Front! Back/
ments in question. Thus, all consonants which manifest a specific manner of Unrounded Rounded
articulation are typically found on the sarne row, so that there is one row for
plosives, another for fricatives, another for nasals, and so forth. Similarly, the High i II
columns of a chart typically make reference to the approximate point of articu- I U
lation; there is thus a column for labials, another for alveolars, another for e 0
velars. Such a chart, which displays some of the most COmmon segments, is il- Mid a
lustrated in (13). e ~

;e
(13) a. Consonants Low a

Labial Labio- Dental Alveolar (Alveo-) Velar Glottal Ii Such charts make sense only inasmuch as phonological segments ~re
dental palatal considered to be divisible units made up of combinations of phonetic char-
!,,. acteristics. ~ h~Ee~ features are precisely what accollnt for the symmetry in
Plosive the inventory of sounds discussed in the previous section.
.
«
voiceless P I k ? !
1
Furthermore, the notion of features allows us to prt:cisely define each in-
voiced b d g dividual s egment. For example, the sound p ma y be charack riz l.! d
informally as the bundle of features in (14).
Affricate
voiceless rs i)' (14) p10sive 1
viJiced do Jj voiceless
[ labial
Fricative
voiceless f (J s J x h
voiced v d
- 3 )'
Features, natural classes, and phonetic plausibilit)·

Nasal 111 11 Jl V In addition to allowing a sys tematic organizati on of phonolugical m:!lri-

ces, the: notion of phonological features also provides t,.vo very s,gn:fic;;nt
Liquid controls on phonological analysis. First, through an und~rstandin~.: of tl:ese
Lateral features, we are able to recoh,rnize that pbonol')gicai segments o C:'J.r ir: 5(, [S
Flap r or classes. Segments which comprise one of these classes oftt.!n fU llet ::.)!} tu-
, getber as a set, either in modifying other sou ads, or in be illg I'!.l(·di!kJ in
Glide w j (w) ! some environments, as iilusrr~1ted in ( 10) :lDd (II ). These ci:;.:,;s~.:i . Gcc a us~
I
(hey are detiDt!d by phonetic features , are commonly called N.":"TUltAL
< i CLASSES. (Swadesb 1934 referred to them as PHONEME CLASSES .) SO pho"';=-

l logIcal features account for the symmetry discl'ssed in the pr e\'i olls Si.: (';!0 J'l.
14 Preliminaries to Ana(vsis Phonological features 15

For example, in Totontepec Mixe (Crawford 1963), a language of Mex- and similarly for alveolars, palatals, velars, and glottals. Further, we may
ico, voiceless plosives become voiced when they follow a voiced nasal expect all voiceless sounds to be affected in a comparable manner, arid sim-
consonant like m. ilarly for several types of voiced sounds. The charting of phonological
segments, then, which is one of the fIrst steps in phonological analysis, al-
(15) lahk 'house' mdahk 'your house' ready carries with it a set of hypotheses regarding phonological patterns
pahk 'bone' nIbahk 'your bone' which may be expected to occur in the language in question, as well as pro-
kak 'chest' mgak 'your chest' viding clues for possible conditioning.

It is not the case here that one specifIc plosive is voiced in the appropriate Representing phonological features
environment. Rather, all plosives are voiced. The plosives form a class, and
the rule accounting for their voicing (stated informally above) makes refer- There is some question as to exactly how phonological features are to be
eace to that class. The voiced nasals tbat constitute the conditioning represented, and thus how natural classes and phonetic plausibility are to be
environment are also a natural class, of COl!I"se, as is the class of voiced captured. It would be possible, and has indeed been the practice in some
stops that results. . traditions, to simply use as features the labels which are given on a chaI1
Note how tbe concept of phonological features allows us to recogmze like that in (13). Linguists have observed, however, that certain sets of seg-
this. The concept of phonological features allows natural classes to be ments which are not grouped together by the traditional articulatory
readily captured, as we see later. features in (13) do, in fact. tend to modify other segments and be modifIed
Second, note that the notion of phonological features potentially serves a by them in comparable ways. These sets of segments form natural classes
valuable function also by pointing out the PHONETIC PLAUSmlLITY of pho- which a feature system should capture.
nological patterns which are found in a language. In (15) it is quite apparent For example, in Asheninka (peru) there is a process whereby some word
that it is the voiced quality of !m! that is affecting the voicing of the follow- initial plosives become semivowels when prefixes are adjoined (D. Payne
ing plosive. As we have already noted, sounds in a sequence do not occur 1981, transcription modified).
as discrete units; rather there is a single wave of sound, WIth each mdlvld-
ual sound making its own contribution. The sounds have a tendency to (16) piiri 'bat' i-wiiri-te 'his bat'
blend into one another where they abut. Thus, the vibration of the vocal toniro 'palm species' i-toniro-te 'his palm'
cords, i.e., voicing, in the pronunciation of a consonant in Totontepec Mixe kapeji 'coati' i-japeji-le 'his coati'
tends to carry over to the following consonant. It is the notion that pho-
nemes are combinations of such features which makes the plauslbIl1ty Observe that this process applies to labial and velar plosives but not to
evident. alveolars. Labials and velars thus form a class whose members are affected
In summary, then, not only is it the case that phonological features pro- in the same way by the same environment. Conversely, sounds made with
vide the motivation for charting and recognizing phonetic content, but they the blade of the tongue, i.e., the area between the labial and velar points of
are also useful as a check on phonological analysis. They give the phonolo- articulation, commonly function together as a class in other languages.
gist a clue as to what sorts of patterns might be expected in a language. We These facts suggest that traditional articulatory features are not sufficient
expect such patterns to manifest phonetic plausibility and to make reference in themselves-that there are other phonetic features which are needed to
to natural classes. Both of these notions find ready expreSSIOn through the capture the natural classes of languages. The feature proposed to capture the
use of phonological features. . class of sounds made with the blade of the tongue, e.g., alveolars and
Looking back at the chart in (13a), we may make some observatIons. alveopalatals, is CORONAL. Labials and velars are grouped together, as they
First, regarding manner of articulation, we expec.t all ploslVes In. a gIven function in Asheninka , in relation to this feature by both being
language to be affected in a comparable manner ~ th~ same envlf~nIn:ent NONCORONAL.
(the data in (15) illustrate this), and similarly for fricatIves, nasals, lIqUIds, Just as sounds at different points of articulation may group together into
and glides. Secondly, looking at point of artIculatIOn, we expect that all natural classes, sounds with different manners of articulation may also func-
labials will be affected in a comparable manner in the same envIronment, tion together. In many languages plosives, affricates, and fricatives behave
 16 Prelimina.ries to An3/ysis Phonological ieawres 17

similarly. "lbe term OBSTRUENT defines this natural class, all the members of of the sound. Soon no more air can pass through the glottis without some
which exhibit significant closure in the oral cavity. Conversely nasals, liq- further adjusnnent being made, such as increased pressure from the lungs
uids, and semivowels (which also behave similarly) are referred to as or lowering of the larynx. With no further air passing through the glottis,
SONORANTS. voicing cannot be maintained. This characteristic of being difficult to
As the theory of Generative Phonology developed, with its principal ex- voice results in sucb sounds being classified as [- sonorant]. (Sounds
pression in Chomsky and Halle 's 1968 work The SOl/nd Pattern of English, whicb are [ - sonorant] are often referred to as OBSTRUENTS.) Liquids,
one set of phonetically motivated features came into widespread use among
linguists. Though some aspects of this set of features are not universally ac-
knowledged, there is a high degree of uniformity in their use. Thus, with
·'If
nasals. semivoweis, and voweis by contrast have no such interrupnoll of
flow and are thus commonlyvoiced. They are characterized as [ + son-
orant] (nasals, in fact, have complete obstruction in the oral cavity, but be-
only a few exceptions, we present below the basic set of features used cause the velum is lowered dwing their articulation air flow is Dot
within this theory. impeded).
These features are BINARY: that is, they are used with only two values:
plus (+) or minus (-). For example, voiced consonants are specitled as Point of articulation features. The following features arc those which
(+ voice] while voiceless consonants are specified as (- voice]. Distinctive define the point of articulation of the sound involved.
features in their binary specifications are commonly enclosed within square
brackets. ANTERIOR. Sounds which are [ + anterior] are made in the..forward Jl;lf\ o.f..the"
mouth, i.e., at or in front of the alveolar ridge. _ abjals, de.n and
Major class features. There are three features which are commonly re- l lveolars are + anterior}! wbile alveo alatali, alatals and sOULds made
ferred to as MAJOR CLASS FEATURES , because they .divide the set of fartner'back are [- anterior] . ea!Ufe' used to spe.cify: only- conso-
-- phonological segments of a language into the most Significant classes. ts, Current theory uses this feature only 10 classify sounds made with
the tongue; under this interpretation labials are not specified at all for [a~­
SYcLABIC. Sounds which are [+syllabic] are those which are found in the terior] (see the discllssion of tbe feature tree in chapler 7, .• Autosegmcmal
eak;::of<thec-syUable. Thl" is the only feature for which no direct phonetic Pbonology")_
l"b,\J. ~
correlate has yet been found. In general, vOR els:ar4 +-=syUabicJ:;and_con- ~~<!/ . CORONAL. 'Sounds wbich are [ + coronal] are made with the
[ ,N~ I"" / ';1f- . tip and blade of l.h.bd;..iJ.
sonants' ar~[,,",",syllabic}. Nasals:or iquids which occur as syllable peak in ?;?i: ·· ~ the tongue (dentals, alveolars, alveopalatals). Utber sounds are
a particular language may. be...-F-+"SYllabic:]::aso:wel~, as the Iml in prism or ::::.- CorTt:..,.J....
: U!.::..-'P lif.,) [-coronal]. This feature is ·also used to specify only consonants.
the 1.:1 ill weather in English. It is:common:to see:the :abbre",atiorr-& for ! BACK. SOUllds whicb are [ + back] are made witb the back part of tbe tongue.
[- syllabic] and 'i for.:[-,t-:.-syllalJiC'j , and some tbeories have replaced the i while those whicb are [- back] are not. Velar sounds IU", Ii: and \I' and
feature SYLLA...BIC with a fonnal notation that uses only these abbreviations . those made farther back are thus [ + back]. Central aod back vowels are
(see ohapter S. Metrical Pbonology). also specified as [+ back]_ All other sounds are [ - back].
CONSOK-\NTAL. Sounds whicb are [+ consonantal] are made with a signifi- HIGH. Sounds whicb are [+ high] are made with the body of the tongue
cant stricture in the oral cavity. By " significan smcture:' is meant any raised above its nonnal resting pOSition. The resting position is approxi-
stricture which is more than that found in vowels, semivowels, or glottal mately the b(!ighr of the: vowel E. The (+ hi ~h] sounds are alvenpa!:!t;d.
sounds. AllKTRl:J consonants are-.-Ihus [+consonanta:!], while owels, palataL and velar consonants, and high vowel;. Otbc-f consonants and ndd
cSeonvowe , d lru:)(llgeal-glides::Z,and::h ar.e:-[-"Cons.ouantal] . Note that or low vowels are [ - high]_
this ''<ature is distinct from the feature [syllabic] in that it groups seon- LOW. Conversely. sounds which are [+ low] are made with the tongue ootly
vowds and donal sounds with the vowels. lowc:red bdow it s Donnal resting position. The [ + low] sOlmds :lre ph~i­
SONOIVNT·. So~ds wbicb are [ + sonorant] are those whicb bave only the de- rynge.:11 consonants. anti open vowels. O[h~r ::onsonants and mid or high
gree of closure whicb d')es not prevent voicing from taliing place rela- vowels are specified [ -Iow]_ This feature is a necessary one to speciry
tively easily. There is an explanation for this: if the closure is extreme mid vowels, which are [-high] and [ - low].
enough--as with plosive", affricates, and fricatives-the cavity behind tbe
closure in the mouth quickly fills with the air involved in the production
'. i
L
18 Preliminaries to Arul/ysis Phonological features 19

The relationship of semivowels (glides) to [ + high] vowels is not obvious -'

feature [distributed] to distinguish labiodentals from labials. Uvulars are
with only traditional articulatory features. It is easily captured., however, with distinguished from velars in these recent treanneots with the feature
these distinctive features. For example, a feature specification which can desig- [high].
nate both} and j in many languages is [- consonantal, + high, - back], while LATERAL. Sounds which are [+ lateral] are made with air passing oyer the
IV and !I would be designated by [-consonantal, +high, + back]. sides of the tongue; this is not true for sounds which are [ -lateral]. Thus,
Note that hand? (glottal plosive) are typically interpreted as not speci- I and related sounds are [ + lateral], while sounds which are not I-like are
fied fo r any of these features, since they involve no oral point of [ -lateral]. This feature is relevant only for distinguishing among liquids.
articulation. Thus r-like sounds are the important [ -lateral I segments.
ADVANCED TONGUE ROOT. Sounds which are [+ ATR] are made by drawing
Manner of articulation features. The following are features which the root of the tongue forward. This feature distinguishes between tense
characterize the manner of articulation of a sound: vowels as [+ATR] and lax vowels as [ - ATR].

NASAL. Sounds which are [+ nasal] are made with the velum open, those Laryngeal features. The following features are used to characterize the
which are [ - nasal] are made with the velum closed. All nasal consonants state of tbe glottis:
and nasalized vowels are thus [ + nasal], while all sounds which have no
nasal component are [ - nasal]. VOICE. Sounds which are [ + voice] are made with the glottis vibrating, while
CONTINUANT. Sounds which are [+ continuant] are those which do not have those which are [ - voice] are made with the glottis not vibrating. Sounds
a closure in the ORAL cavity sufficient to stop the air flow through that like b, III, \ ', I, and w, and voiced vowels are thus [+voice], while sounds
cavity. Among the [+ continuant] sounds are fricatives, liquids, semivow-:; like p. f. voiceless glides, and voiceless vowels are [ - voice].
els, and vowels . Plosives, affricates, and nasals are considered to be SPREAD. Sounds which are [+ spread) are those made with the glottis com-
[ - continuant].2 pletely open at some point in their articulation. Thus [h] and all aspirated
ROUND. Sounds which are made with the lips rounded are considered to be sounds are [ + spread], all other sounds are [ - spread].
[ + round) , while those which are not are considered to be [ - round]. CONSTRICTED. Sounds which are [+ constricted] are made with the glottis
Round vowels and all labialized sounds (wand sounds that have a completely closed, either in the articulation of glottal stop or sounds made
w-()ffglide) are thus [+ round] , while all other sounds are [ - round). with the glottalic air stream mechanism. Thus, [1) , [p'], [6), and other
STRIDENT. Fricatives and affricates which involve a considerable amount of such sounds are [ + constricted] , all other sounds are [ - constricted].
noise in their articulation are [+ strident], while those which are
[ _ strident) are relatively quiet. This definition is based upon acoustic fac- The features just discussed will adequately describe the majority of sounds,
tors which are clear when measured with instruments but which are not so but they do not account for secondary articulations such as palatalization,
apparent auditorily. This feature is relevant only to fricatives and labialization, prenasalization, or postoasalization. 10 some works, features such
affricates. Those fricatives and affricates which are labiodental, alveolar, as [palatalization] are used in a binary way; but, in most current theories, there
alveopalatal, and uvular are [+strident], while others are [ - strident] . For is no need to have extra features like these. This is because the model contains
the labiodental and uvular fricatives and affricates, the stridency is caused a mechanism which more naturally accomodates phonetic sequences which
by the closeness of the articulator to the point of articulation. 10 the case function as a single pbonological segment. This is explained more fully in
of alveolar and alveopalatal sounds, stridency is caused by GROOVING chapters 4 and 7.
the tongue blade. These strident alveolar and alveopalatal sounds are in Various other features have been proposed from time to time, but tbose
fact called GROOVED FRlCATIVES. Some recent treatments of feature described above are the most important ones from Generative Phonology
specification use [strident) only for grooved fricatives, and use the whose usage has continued. For further discussion of various proposals
which have been made regarding phonological features, the reader is re-
:!Thc specification of nasals as [- continuant] in Generative Phonology ~ff:rs fro~ so~e ferred to Ladefoged 1975, Kenstowicz and Kisseberth 1979, Clements
phonetic classifications (Pike 1985) which specify nasals as COtmNUA.'lTS. ~ike ~ S~CdlcatlOn
1985, Sagey 1986, McCarthy 1988, Goldsmith 1990, and Kenstowicz 1994.
for continuants considers overall impedaoce of the airstream, while the specification m Genera-
tive Phonology considers closure only within the oral cavity.
 Preliminaries to Analysis Excursus on redundancy 21
20

Using distindive features (19) [- sonorant] - [+voice] / [ - syllabic]

+ vQlce
By means of these features all of the segments given in (13), and many
more as well, can be distinguished from one another. For example, p, m, J, This more generalized expression of the process specifIes that obstruents,
and i may now be specified by means of the following bundles of features: i.e., nonsonorants, become voiced following another voiced consonant. It
makes predictions which go beyond the data; that is, it assumes that all
(17) m fricatives and affricates in the language become voiced following any
P
- syllabic voiced consonant. Rules are usually expressed in the most general way pos-
- syllabic
sible that is not contradicted by the data. Such formulations make explicit
+ consonantal + consonantal
what we know about symmetry-that all members of a natural class tend to
- sonorant + SODoraot be affected in the same way in the same environment.
+ anterior + anterior
- coronal - coronal
- voice + voice Excursus on redundancy
- continuant - continuant
+ nasal It is characteristic of human language to exhibit redundancy. and this is true
- nasal
no less of phonology than other parts of language. No language uses phono-
logical features in a mathematically efficient way because not all combinations
f i
of features occur in any language. Indeed, although [high] and [low] are sepa-
- syllabic + syllabic rate features, it is physically impossible to articulate a segment that is both
+ consonantal - consonantal [ + high] (the body of the tongue is raised) and [ + lOW] (the body of the tongue
- sonorant + sonorant is lowered). Nor is it possible to articulate a sound in which the glottis is both
- anterior + high [ + spread] and [+ constricted]. Even ignoring such impossible combinations,
+ coronal - low however, languages manifest a considerable amount of redundancy because of
- back the specific facts of their own phonological systems.
- voice
- round Consider a language with a typical fIve-vowel system, for exampk 3
. + continuant
Using the standard features to distinguish the vowels according to their
- nasal + ATR
phonetic characteristics yields the following: .
- nasal
(20) e Q o u
At fIrst glance this may appear to be a rather bulky way to refer to a par-
ticular phonological segment. The utility of binary. features IS more high + +
apparent in defining natural classes. For example, we rrught informally ex- low +
press the phonological process illustrated in (15) as in (18). back + + +
(18) p, c, k - b, d, g / voiced consonant _ _ Three features are enough to distinguish eight segments if used with maxi-
mum efficiency. Two features are enough to distinguish only four segments ,
In more formal terms, using binary features, the process may be ex-
pressed a:; in (20) (see below for discussion of rule formalisms). 3A more elaborate system o f segments wo uld yield more redundancy. but thjs simple system
is sufficient to iUustrate the point being mad!! here.

I
1
4Tbe fea.ture [round] is DOt needed in this system because it i~ · totaUy predictable (redu ndancy
rules could be written to account for its values); only Donlow back vowels are round. and tbey
always are.
22 Prelimi1l3ries to Analysis Excursus on redundancy 23

however, so at least three are needed. Capturing the redundancy in such a

system is helpful, however, not only by facilitating rule writing (redundant
(24) [ + high] [+ low] .-
j j
features need not be specified, see below), but also by allowing natural
[ - low] [ - high]
classes to be more easily and e>:plicitiy recognized.
A suitable technique for capnuing redundancy is to compare each feature
Further, comparing [low] and [back] yields all additional redundancy
specification one by one with all the others. For example, it is evident in the
(any [+ low] segment is [ + back]):
chart in (20) that if a vowel is specified as [ + high] it is always [ -low] also
(this is one of the universal redundancies mentioned above that has its source
(25) e a 0 u
in impOSSible competing articulations). We may capture this fact by circling
the redundant features in the chart and writing a rule to express the redundancy high + 8 +
(by convention, redundancy rules are commonly written vertically): low 8 + 8
back EEl + +
(21 ) e a o u
high + + (26) [ + high] [+ low]
low 8 + 8 j j

back + + + [ - low] [ - high]

+ back
(22) [+ high]
The converse relationships are not valid, however, because [ - low] seg-
j
ments can be eitber [ + back] (/01 and lu I) or [ - back] (Iii and leI).
[ -low] Consider now the feature [back]. If a segment is [- back] it is necessar-
ily [-low]:
Note that it is not the case, however, that [ - high] involves a redundancy as
regards [low]; there are segments that are both [- high, - low] (l eI and 101)
(27) e a 0 u
and [ - high, + low], e.g. , l a/ . Further, comparing [high] and [back] yields
no redundancy either; there are segments that are [+ high, + back] (luI), high + 8 +
segments that are [+ high, - back] (Ii!), segments that [ - high, + back] (101 low 8 8 + 8
and l al) , and segments that are [ - high. - back] (leI). back EEl + +
Now conSider the next feature [low]. While there is no redundancy in-
v ol v ed between [-low] and [high] (there are segments that are (28) [ + high] [+ low] [- back] .,~:
[ -low. - high] (leI and 101) and segments that are [ - low, + high] (I iI and j j j 1
lu!)) . there is a redundancy involving [+ low] aad [high] ill that [+ low] [ - [ow] [ -low]
[ - high]
segments are necessarily [- high] (the counterpart to the universal redun- + back
dan cy above). Thus, adding to the above:
Again, there is no redundancy related to segments that are [ + back].
(23 ) e a o u Extracting redundancies in this manner is significant in a number of
high + 8 + ways. First, the non-redundant features indicate the minimal features needed
low 8 + 8 to specify a segment in any rule making reference to it. Thus, for lal the
back + + + only specification needed is [+ low]. Further, defining natural classes is .
Simplified also once the redundancies have been extracted. For example, if
palatalization takes place before front vowels (a not uncommon pattern, see
 24 Preliminaries to Analysis f Excursus on rule formalisms
[ 25
chapter 3), it is evident from the chart in (27) that only [- back] is needed (29) A ~ B/ _C
to-specifY such vowels, with [ -low] unnecessary. But that very fact reveals
also a si~ficant result from the extraction of redundancIes ; each redun-
dancy rule reveals a true fact about the language. So with reference to the
!I where A, B, and C are sets of distinctive features referring to natural
classes and the underline indicates the pOSition of the segment affected by
redundancy rules in (28), it is clear that there are no low front vowels (the
the rule, i.e., the class A is manifested as the class B in the environment (in
last pair of redundancy rules). This is a true fact of the language that should this case preceding) C.
be stated overtiy somewhere rather ihan just being left to the observer to ID- f
Consider a ruie of final devoicing;. such as exhibited in the following
tuit from the chart of phonemes, and it is the redundancy rules that make
Russian data (data from Brett Benham, personal communication):
such statements explicit.
The same principle of extracting redundancies can be applied also to sys· (30) NOMINATIVE GENITIVE
terns of consonants, with comparable results.5 SINGULAR SINGULAR
snop slloba 'snob'
snop snopa 'bale (of hay)'
Excursus on rule formalisms
kot kada 'code'
kot kota 'cat'
It may be useful at this point to introduce some of the formal devices mak nzaga 'magician'
that are used to express the facts of phonological systems.6 mak maka 'poppy seed'
Phonological rules typically have a characteristic form:
Because both voiced and voiceless obstruents occur intervocalically, it cannot
5If the redundant features, circled in (27). are omitted from the chart, the result is referre~ to be the case that the rule is one of intervocalic voicing; rather, it must be the
in currer.t tbeory as CONTRASTIVE UNDERSPECLFICATlON. Often an additional step is ~en . t~ .Y'el~ case that obstruenrs are devoiced finally_ We might write the rule like this:
a represmtation that is r eferred tC> as RADICAL UNDERSPECLF1CATlON. as follows . In ( _7) Jt I~ eVI-
dent th.u there is an additional redundancy that can be captured: because the features an: bmary. (31 )
it i::; ]ogi-;ally necessary to specify only a single value for each fe~rure . with.the otho:r. va~e p~o:.
f - son ] -
[ +VOIce
[- voice]/_# (where # marks word boundary)

I!
dicmblc as the default. that value which occurs if the ftrSt value IS not speCified. So 10 (_7), tor
example, even ;lfter tbt! redundanci es have been extracted and the redundancy ~les s~( ed., the
fearure [high] stiU shows both + ;lnd - Yalues in the chart. In radical underspeclfi c;ltl~n , then.
That is, for all parts of the rul e, we say as little as possible, making the nat.
for each fearure ooly a single value is specified, with additional redW1dancy rules wnlle!D . So
fo r [high], for example. it might be proposed that only [+ high] is specified. with any [ - hIgh] ural <.:iasst!s referred to in the rule as large as the data will allow, Thojfe is
spccific~ltjon accoIUltcd fo r by th-! foll owing rule: still a redundancy here. however, such that it makl!S no ditTerl!nce whtther
[ ) the input to the rule is voiced or Dot; the important thing is thar the ou('put is
I voiceless . Thus, we may simplity the rule: '
[ - high) _ _. _ .
Such a n.:dundaOl.:y rul e! is [0 be interpreted as indi cating that if there IS no specl hc:ltlOO tor the
feature !hig;h), supply the fcarure ::.-pecification { - high]. Typically, arguments can ~e so:t forth.to
support whether it is lhc + or tbe - value for such a feature that should be spe.clfio:d . The 10·
tent is t·) kave one segment to[a 'iy devoid of specified features. that segment bemg tcnned the
R,·\OIC'Al.LY UND.l::RSrECIFLED SEGMeNT, in this case the radically underspecitied vowel. In the
I! (32) [ - son] - [ - voice]/_-_#

If the input obstruent is voiced, this rule will make it voiceless: if it is

tivc:·yowei system of Spmish. it is in fac t It which IS the radically underspeciticd VOW !!!. A sup· ! v~iceless to begin with, the rule will apply, but it will not change th·, speci.
poning argu~ent is the fact that It is e th3t is inserted in borrowed words to~brin£ them ~lO a~. ficatIon for [voIce] whIch IS already [ - voice] (this is re f"rre rl to as
cord with pemlissible syllable structures of Spanish (Hooper 1976:23.Jfi): e'\'!l()b . snob. f VACUOUS APPLlCATlON)_ FUrther, if in a given rule there is a speCification
c:;m()kil. ~ 'smoking Uacket)" esl.IVo 'Slav', The idea is that the insertion rule need speedy only
for a feature on the n ght SIde of the arrow (here, [ - voice]), it is almost al.

I
a vowe! pos ition, with all the fe.ltures characterizing tbat vowel predic~bl~ on the: basi~ of t~ e ays
redundo·. Dcy rules. Chapters 4 and 8 discuss syllable structure in sO[J)e detaiL For extens Ive dIS· :v possible to omjl any specification for that saIne" feature on the Jeft side
cussion of underspecification lbl'ory see Archangel..i 1984, 1988. tsee ~n example of an excc:ption:.d C<ISe below); this is pieferred.
6S 0fD !! o f the examples cited bere are discussed in roore detail later in this book. They are in- I~ IS often the ca~e rhat a given phonological nile applies in more: thm One
-. cluded ben~ in an abbreviated f('·rm only to illusrrate the formalisms of rule formation.

L envlJonment; conSIder a language which has ·the same devoicillg rule as

26 Prelimin3ries to Analysis
Excursus on rule formalisms
27
Russian, but which also devoices obstruents when they precede voiceless ob-
strucnts in clusters. This second pattern of devoicing could be written as fol- Note, however, that the same pattern occurs also when a consonant inter-
lows (making use of the same simplification notions mentioned above): venes between the [s] and the [iJ, as shown in the following compo~ds:

(37) ?;irip + -su!"u ?;irip,{uru

(33) [ - son] - [ - voice]/_ [ - son ] ' to cut a strip' ' off' 'to cut a strip off'
- VOIce
pahi:p + suruk pahip-fiiruk
As is commonly the case, the form of the rule itself demonstrates the 'pepperwood ' 'under' 'under the pepperwood' (place name)
assimilatory nature of the process involved (note the pairing of [ -voice]
and [ - voice]). Now, when we compare (33) with (32), we see that they are The rule for the data in (37) would be as follows:
very similar in form; in fact, the input and the output of the rules are identi-
cal, only the environment is different. We can combine the two rules into
one using the braces notation to indicate that either environment suffices) as
(38)
- son]
+ cor _
[ -ant
+ high
JI [ +SYll
+ high
J [+cons] _
follows: [
+cont - back
;
.!
(34) [ - son] - [- voice]/_ Again, we can see a similarity between (38) and (36); if there are no
rules which must apply between the two in the grammar of the lan!!Uaae
o 0'
we can abbreviate the two rules by using the parenthesis notation to indi-
-; cate the optionality of the intervening consonant:
Providing there are no rules which must be ordered between (32) and (33)
in the grammar, the theory encourages us to write the process as (34) in-
stead (there is some debate as to whether technically it is now a single rule
(39)
- son ]
+ COr _
[ -ant
+ high
JI [ + sYIl ] ([+cons])_
+ high
or still a pair of rules). [
+ cant -back
A second abbreviatory device used in the theory is that of indicating an
optionality in the environment. For example, Karok shows a pattern of
Consider now a pattern in which there are two variables which must be
palatalization of [s] to ill when following [iJ, as shown in the following associated. For example, in Hausa (a Chadic language widely spoken in
pairs of words (Bright 1957): West Africa) vowels may be contrastively short or long, but they are always
short in closed syllables (tone omitted in these examples):
(35) nzu-spuka 'his money' lJpuka 'money'
?zi-skak 'he jumps ' ijk.ak ' to jump' (40) ci: ' to eat' cm 'eating'
IeEe: 'lip' lel5ellsa ' his lip '
Again, the aSSimilatory nature of the change is evident, the high front
tongue position for the vowel carrying over onto the following sibilant. The The rule for this pattern (assuming that it is completely general and not
n:le can be formalized as follows (this is ooly an approximation, pending a grammatically defmed) would be: 7
complete analysis, but it does illustrate the formalism in question):
(41) [+syll] - [-longj /_ [ + cons]$
(36)
-son] [ -ant ] I [ +SYll ] _ There is a complication, however, in that if the vowel in question is /0: /,
+ cor _ + high + high
[ + cant it is not ooly shortened, but also unrounded, approximating [,,]: .
-back
f 7Here [longJ is used to capture in as simple a manner as possible the contrast beween long
,
l and sbort vowels, and $ represents sylJable boundary. For more discussion of syUable strucrure,
and a more accurate formal account of such facts, see chapters 4 and 8. ,
-~
28 Pre/imiTl3ries to An3/ysis Excursus on role formalisms 29

(42) ci.wo: 'illness' ci:wansa 'his illness' (46) [+ cons] ~ [+voice] /_

ido: 'eye' idansa ' his eye' +cons]
[ + voice
The rule for the data in (42) can be written:

(43) +SYil l i -long l 1_ [+cons]$

Comparing the two rules, we can see a formal sim ilarity: in each instance,
the specifIcation for [voice] which is the output for the rule matches the
_. high
[ + back
~ l-roundJ speCification for [voice] carried by the conditioning consonant. Further,
J conceptually the process is the same: two consonants in a cluster must
agree in voicing, conditioned by the second. In such an instance we are
Now, however, tbe rules in (41) and (43) can be combined through the an- better served to coilapse the two rules together if we can_ It is the notion of
gled braclcet notation: Greek variables which allows us to do so, as follows:

(44) (47) [+ cons] - [avoice] /_ +cons]

+syil ] [ -lOng ] I _ [+cons]$
- high > _ < - round> [ a.voice
[
<+back
Here the a is a cover specification indicating that whatever the speCification
This formalism represents co-occurring variables; in a given reading of the for the fearuee so indicated (in this case, [voice]) in the conditioning envi-
rule to test for its application, in the first instance all information within the ronment, whether + or - , the specifIcation which results from application
angled brackets is included in the ru le fIrst (= (43», then in the second in_ J of the rule must match it; i.e" in any given application of the rule, the two
stance all information within the angled brackets is excluded ( = (41». a 's must be identical.
Thus, if the vowel in question is [ - high, + back], the result will be Note that the same sort of formalism is available also for the less com-
[ - long. -- round] in the environment in question; otherwise, i.e. , the vowel mon pattern of dissimilation. For example, Huamelultec Chontal shows an
is not [ -- high, + back], the result will be [ - long] only. Note that no rule alternation for the imperative morpheme -13? such tha t the lateral is voice-
contains only a single set of angled brackets; there are always at least two less after a voiced segment, and voiced after a voiceless sC!!...'lnent (data from
sucb set s in any given rule. Waterhouse 1949 as reported in Kenstowicz and Kisseber;h 1979:364):
Thert: are also phonological patterns which show assimilation that works
botb ways, such that there is assimilation (or dissimilation, see below) (48) kola? 'say it!' pandi1 ? 'sit down !'
whatever the specifIcation (+ or - ) for the fearuee involved_ For example, miHYa? 'tell him !' fuF'a? 'blow it!'
consider a language which has voicing assimilation in consonant ciusters ,
such that whatever the specifIcation of voicing for the second consonant of Hen:: we want the voicing of the lateral to be exactly the opposite of the
the cluster. the fIrst must Ircatch it. In the case in which the second conso- voicing of the preceding segment. not the same as ill th.e assimilatory pat-
nant is voiceless, the rule might be as follows (in that the example is tern captured in the fOnTI3lism above. For such a tlissimi latary parter:), V; •..!
hypothetical, the details ma y be slightly different in a genuine instance): can mark one of the Greek variables as - to indicate that the cpposi[~
specificat ion of the conditioning environment rr~u st result (this fom :ulation
(45) [+ cons] - [ - voiceJ/_ of the rule assumes that the p3.[[crn is completely gen t:ral for latenls, n Ol
+cons]
[ - voice
, limited to this specific morpheme):
i
i (49) [+latJ - [ - avoiceJ /[avoice]_
I
Similarly. if the second consonant is voiced, the rule would be as follows: }

r Taking a page from mathematics, if th¢ cOllditi011inQ L'-n '.'i;onm ent i:,

l [ +voice], then the output of the ruk will be [ - + voice] = [ - voic e] ;

 I
,I
30 Preliminaries to Analysis i
I
conversely, if the conditioning environment is ( - voice], then the output of
the rule will be (- - voice] = ( + voice].
There are also instances in which two or more fearures vary independ- f
ently as to their + or - value, but such that there is a related assimilation !
(or dissimilation) in relation to the fealUres involved. Consider, for exam-
ple, the phenomenon of nasal assimilation; it commonly happens that (in
the general case) the nasal in question is bilabial before a bilabial, alveolar
before an alveolar, alveopalatal before an alveopalatal, velar before a velar,
Ii
e.g., consider the classic case of English morpheme in-, in words like im-
possible, intolerant, injustice, incongruous (in some dialects) . As in the case
I
above, the fealUres must match, but now there is more than one fearure in- 2
volved, and each is independent. As a first approximation of the rule,
consider the following: Phonemic Analysis
(50) (+nas] - + cons
[am, 1
I3cor
yhigh
1-
nan!
I3cor
Linguists often engage in firsthand investigation of a language that has
been studied very little or not at all. In these cases, the first data they obtain
they transcribe phonetically, but they soon find that the phonetic transcrip-
oback yhigh tIon does not correspond to native speaker intuitions about which sounds
oback are the same or different. So a very early step in the process is to analyze
the data systematically to ascertain the phonological system which the pho-
Here each Greek letter has its counterpart, and the counterparts must netIc transcnptlOn reflects. Of course, phonological analysis is of interest in
match in any given application of the rule. But note that the use of CJ. and 13 its own right. In this chapter we present analytical principles which will aid
here is different from the use of CJ. and - CJ. above; CJ. and - CJ. must always the IOvestigator in getting at the phonological system reflected in phonetic
be opposite in result, whereas CJ. and 13 are independent of one another and data.
thus in result may match or differ, depending on the specifications in the I ~n the previous chapter we pointed out that different segments can share
environnient. Note that, as was the case with angled brackets, no Greek many features. When two segments have a number of fearures in common,
variable occurs but a single time in any given rule; each always occurs at
least twice (if the pattern is such that the specification for a given fearure
I
'I' (
we call them PHONETICALLY SIMILAR SEGMENTS. The basic goal of a sys-
tematlc phonologIcal . analysis is to determine whether those phonetically
for a single segment can be either + or - , which is what the Greek vari- ~ slID1lar segments are 10 fact the same phonological unit or not.
able notation indicates, it can be omitted from the rule). Note in (50) that if ~'sr,,·bl1-f'e,......
in fact the rule applies only to n, not for example to m, then the input to the
rule must specify that information: I n- CA;rt- .

Basic concepts .in phonemic analysis "(" c..c->-vt"~"I6-t;

f'r.e..<- v"",-<' a.bl.--
., There are three relationships which phonetically similar segments may
(51) +nas ] ... have to one another. They may be CONTRASTIVE, in which case they are
[ +cor dIfferent phonological units. Or they may be in COMPLEMENTARY
DISTRIBUTION, in which case they are merely different superficial variants
This would then be an instance in which the same fealUre (here, (cor]) of the same phonological unit. Or they may be in FREE VARIATION, in which
appears on both sides of the arrow, a fact that (as mentioned above) is not case it will turn out that they may be either in contrast or complementary
the typical case. I distribution, depending upon the specific details involved. We turn to the

f
32

notion of complementary distribution first, then contrast, and take up free

variation later in the chapter.
Phonemic Analysis
I Basic concepts ill phonemic analysis

(53) [s] [z]

33

_apo 1111'-mo
t.
po_o a_no
Complementary distribution
p"- Lla
If two phonetically similar segments are each consistently found in diS~ U_to ko_mos
tinct contexts in the phonetic data from some language; the investigator kozmo_
should suspect thatjthey are in complementary distributionJ Consider the
phonetically similar segments [s] and [z] in the data in (22)nom certain di- (54) [z] always precedes a voiced sonorant consonant, i.e., [m], [aJ, [I];
alects of Spanish.8 A quick perusal of the data (remembering that in a real [s] never precedes a voiced sonorant consonant.
language the number of words manifesting such patterns would likely be
vast) shows that the contexts in which [s] occurs are quite distinct from We can hypothesize, then, subject to examination of further dato, that the
those in which [z] occurs. voicing of [z] is in fact a reflection of the influence of the voicing of the
consonant that follows it. We conclude that in these dialects of Spanish [s]
(52) [sapo] 'toad' [mizmo] 'same' and [z] are different forms of the same phonological unit; such a pair of
[poso] 'well' [azno] 'donkey' sounds we term A.LLOPHONES. The allophonic statement can be written as in
[pas] 'peace' [izla] 'island' (55); that is, the phonological segment /s/ is realized as the phone [z] when
[listo] 'ready' [kozmos] 'universe' occurring before a voiced SODorant consonant. And it is realized as the
phone [s] in all other environments .
Oue technique which is particularly useful in examining the distribution of
phoneti·;ally similar segments is charting. In application, what the analyst does .t-j ? (55) /s/ - [z] / _ voiced sonorant consonant
is list the environments in which each of the sounds in question occurs, group- / [s] / _ elsewhere
ing the envirorunents as well as possible in terms of phonetic sirnilarity, and as
much as possible organizing the data so as to bring into relief the similarities After having stated that [s] and [z] were different fomls of the same phol
ari10:1g the: c:nvironments in which the:: pholles in question OCL:Uf. For the data in nological unit. wby did we then go on to choose /s/ as the basic phoneme \
(52). for exa.mpk, we would draw up charts, as in (53), where all instances of rather than IzJ? Thae are sc:veral reasons. First, the: process as stated in (55)
a given sound are lined up to facilitate comparing the environments. is simpler than the reverse process would have been. Examining the dat~
In the chart in (53), we can see easily that the distribution of [s] and [z] (53), we can see that the rule for the reverse process would have to include /'
call be characterized in the two simple statements of (54). Notice that these three different environments-preceding vowels, word finally, and preced-
cnvirorments are MUTUALLY EXCLUSIVE. that is, [s] Dever occurs in the en- ing voiceless consonants. The rule would be stated as in (56).
vironment in which [z] occurs, and vice versa. The statements in (54) have
a pbont:!iC: plausibility abot:! them: the difference between [s] and [z] is one
of voi~ing. ~!ld the diffaence in their distribution, as statc!d in (54), is pre-
cisely thot [z] occurs only when followed by voiced sonorant consonants,
(56)
/z/ - [s] / _ r1#vowel
(word finally)
voiceless consonant
I
\\.'hlle [s] n!!ver occurs in that environment. They are in complementary dis-
tribution with each other. [z] / elsewhere

Rule (56) is less satisfactolY than rule (55). First. (56) is more complex than
l:ITo thi:: point in the text. sound,: have been represented without rd'erence to their phonologi· the broad statement in (55), where only one environment is needed. Second.
cal station. It is common practict·, however, to enclose phonetic material in square brackets, rule (55) is phonetically more plaUSible than rule (56); in (55), the process is
e.g., [I], while material that is con·.idered to be a phonological unit within a particular system is from voicelessJlt!ss to voicing in an envirollJl.leot prc:ceding voiCing; in (56),
enclosed in :.Ianted lines (e.g .• I i I ). This will be the practice throughout the remainder of lhe text
wherever the level of represent.3tinn of a sound is not otherwise clear.
34 Phonemic Analysis Basic concepts in phonemic 3na{vsis
35
the process· is from voicing to voicelessness preceding both voicelessness (the
The easiest sott of contrast to recognize is CONTRAST IN IDENTICAL
end of a word and voiceless consonant) and voicing (vowels).
ENVIRONMENTS . Note the distribution of [s] and [zl in the following English
The importance of phonetic plausibility in the explanation of complemen- data.
tary distribution cannot be overemphasized. Unfortunately, phonologists have

I
not yet corne to universal agreement as to the details of what constitutes pho- (57) [SIp] 'sip' [ZIp] ' zip'
netic plausibilty; but there is vridespread agreement on the principle: If it is [suw ] 'sue' [zu w ] 'zoo'
claimed that two sounds are manifestations of the same phonological unit, it is [hIS] ' hiss ' [hIZ] 'his'
the analyst's job to demonstrate that the differences between the two sounds
can be attributed to the different environments in which they occur. Although Each pair of words in (57) differs semantically, and has a slight phonetic
there is no exhaustive compilation of plausible environmental conditioning, a difference. Specifically, [s] and [z] differ only in voiCing, both of them be-
number of patterns appear repeatedly in natural language. As an aid to devel- ing alveolar grooved fricatives. When we look at the data, we might first
oping awareness in this regard, chapters 4, 5, and 6 present some of the most look to see if the difference between them (voicing) can be attributed to the
common patterns which are found. More important to us at this point than the
details is that the emphasis is on explanation. The interpretation of what con-
different contexts or environments in which they occur. Jor the data in (57). y
however, It IS clearly Dot the case that the voicing difference can be attrib-
stitutes an adequate explanation will become more precise as the analyst has
more experience in analyzing phonological data.
uted to environment. for in each pair there is no environmental difference. _0.
In these pairs the two segments contrast in identical environments, as (58)
It is often difficult for the beginning student to sunnise whether a rule is shows.
phonetically plausible. However, if the evidence is overwhelming that the
environments are munmlly exclusive, and the corresponding rule is simp~-? (58) [SIp] versus
this should normally indicate that he or she is dealing with a valid rule! [suw ] versus
Chapters 4, 5, and (; deal with processes that are considered to be natural. [lus] versus
For now, we list here a few of the most common ones:
Since both of the sounds in question occur in the same environment. it is
NASAL ASSIMILATION: A nasal consonant can assimilate to the same point of not possible to attribute the difference between them to the environment.
articulation as fue consonant which follows it, as seen in (2). We must account for the difference on other grounds, specifically that they . .iJ
VOlClNG ASSIMILATION: A consonant can take on the same voicing as the are different honeticall because they are different phonologically. These
·sounds either before or after it, as seen in ( 10), (11), and (15). different phonological units are called PHONEMES. A P oneme is a distinc-
VOWEL NASALlZATlON: A vowel can be nasalized contiguous to nasal tive phonological unit in a given language which can be shown to contrast
consonants. with other phonetically similar units. Pairs of words such as those in (58),
WEAKENING: A consonant can become more vowel-like when surrounded by which differ in meaning and which show the two units to be contrasting in
vowels, especially more open vowels. That is to say, it is produced w ith
less closure. For example, a plosive can become a fricative, or a fricative
can become a glide because the vowel next to it is produced with so much
r identical environments, are called MlNlMAL PAIRS.
It is sometimes the case that for a particular pair of phonetically similar
segments in a certain language there are no pairs of words which differ only
less closure than the consonant. This process is often called WEAKENING in terms of the sounds in question. ~s.equ~y, no minimal pairs can..he_
or LENlTION. It is illustrated in (16). found. In such cases, however, the ·sounds may still be contrastive. Often,
VERY SIMILAR environments can be found to show that the two sounds are
Contrast "..,,;,"" J. P'"i f S contrastive, even though the environments are not IDENTICAL . If no phoneti-

?- When two phonetically similar sounds stand in contrast to one another,

t cally plausible rationale can be found to show how one of the sounds has

r
been modified by its environment, then the two sounds are still considered
they are distinct phonological units; that is, the difference between them to be contrastive,]...
~ be attributed to the contexts in which they occur. For example, Oa!a from Kaiwa, a Guarani language of Brazil (Bridgeman
1961), give no minimal pairs contrasting [p] and [b]. As we examine these
36 Phonemic Analysis Procedures for phonemic analysis 37
data regarding the context in wlllch [P] and [b] occur, we see that there are If we chart these data as we did in (53), we get results like those in (61).
no minimal pairs; there are no words which differ only in that one has [p]
where the other has [b]. .;, .. (61) [P] [b]
0_ a a_ a
(59) [opa.] 'it is fInished' Cabal 'place'
[ipo] 'Ills hand' [bo?i] 'necklace' i_o #_01j

[Piki] 'fish' [oboko] 'his own purse' #- ikj o_ oko

[api] 'here' [bia] 'other people' a_ j #- j

Though. one can usually fmd minimal pairs to contrast a pair of phoneti- Such a chart shows clearly that even though there are no words which
cally similar segments, cases like (59) are not unheard of. If we were to try differ only in that one has [P] while the other has [b], the environments in
to account for the difference between the sounds in terms of the different which the two occur are very similar. In particular, there is notlllng in their
environments in which they occur, for the data in (59) we would have to respective environments which could plausibly be interpreted as accounting
say sometlllng like the following: for the difference in voicing.
Furthermore, other Kaiwa data show minimal pairs which contrasl [k]
(60) [P] occurs before [i] and [g], as in [ka?u] 'intoxicate' versus [ga?u] 'desire'. The principle of
following [i] symmetry adds further confmnation to the notion that Ipl and I bl are indeed
before [a] only if it also follows [0] contrasting phonemes of the language. This is so because [p] and (b] difrer
before [i] only if there is not a vowel following the [it in exactly the same way as [k] and [g] do, wlllch do contrast in identical
'
. environments .
[b] occurs before [a] only if it also follows [a] A conmlent about analyzing real language data is perhaps in order here.
before [0] only if it does not follow [i] In working with restricted data in the classroom, it will almost always be
before [i] only if there is a vowel following the [i] possible to make some sort of statement of complementary distribution if
no minimal pairs exist. But given the extensive data of real language. it
We can clearly make a statement regarding the distribution of [P] and should quickly become evident that when the lack of minimal pairs is acci-
[b], for we can identify specific environments in which each one occurs, dental and tbe phones are in contrast in the language. allY such statement
and in nOlle of those environments does the other occur. But note that the will be so complex, considerably more so than in (60), that it could not pos-
differen: e between [p] and [b] is one of voicing, and so it is that specific sibly be a correct representation of [he patterns actuall y opl.:rative in a
differen:e which must be: anributable to the contexts in which each occurs. natural language.
If we examine the enviror.ments stated in (60), there is no phonetically
plausibl, explanation why hapbazard particulors of the surrounding vowels
should influence voicing. Furthermore, both [P] and [b] do occur before Procedures fo r phonemic analysis
somt! or the same vowels. It is only the wida contexts which differ. and
r:uht:r insignificantly. Even though we ftnd no minimal pairs, we are unable . An effic!enr syst~ma[ic way to approach pbon(ti ogi c.- a: a r.;.-~lysis is to ~X 3:;!­
[0 ma.k( a plausible statement of environmental conditioning either. Such me each parr of phonetically similar segments to d~termitlC' which C0nt-USt and
data we classify as representing CONTRAST IN ANALOGOUS ENVIRONMENTS. r whicb are in complementary distribution. The best daul fvr an il.lit.iai ?Jluivsis
This type of contrast can b~ : taken to be as conclusive as contrast in identi- are. short w~rd~. Once the tlG31yst is sure that he 'J[ she bJS an <.Iccur:..te pbo-
cal environments, which is based on minimal pairs. I netIC transcnpnon for these data, the follOWing $ t~·ps can be tak¢n:

I !. Chart of phones. Make an articulatory chart of each of tbo phonetic seg-

ments (pbones), along the lines of that in (13). This should involve a chart

1 of consonants and a chart of vowels. In most cases it will be nece,:sary to

38 Ph onemic Analysis
r
1
Procedures for phonemic analysis 39
make some preliminary decisions as to whether some segments are func-
the native speaker about sameness and differentness, and is therefore usu-
tioning as vowels or consonants. Semivowels, liquids, and nasals may
ally more suited as a basis upon which to construct an orthography. As we
function as either, depending on context. Similarly, some sounds, e.g.,
shall see in chapter 3, however, morphophonemic considerations can
affricates, may function as single units or as sequences. Again, the con- }
come to bear on decisions regarding phonemes. :,
texts in which such sounds are found in a particular language justify the
interpretation as consonant versus vowel, or unit versus sequence. In
6. Distribution of phonemes. Determine whether all phonemes in the in-
some cases this may be a fairly complex issue, so we postpone a more
ventory have complete distribution. @ften there wiJl:be.cce ltainopnonemes
thorough discussion of it to cbapters 4 and 8 when syllable sffilcrure is
which do not::occur:::in ll~possible environments. For example, in many
discussed in detail. For tbe present, the charting of the segments as they
languages there are certain phonemes which do not occur at the begin-
are in tbe transcription will suffice for this step.
nings or ends of words, but only between vowels. In Asheninka, the pho-
nemes Ir/ and Iyl occur only between vowels, never word initially.
2-. Phonetically similar segments. Determine pairs of phonetically similar Another parameter along which distribution may be checked is wtretl'ieF OF
segments. These pairs of segments are the ones for which either contrast f ot:e" eI¥ consonant:can- b"e:followed by every. ,",owel-m the=inventory. It
or complementary distribution should be established. At the stage of de-
termining each of these pairs, it is usually helpful to connect each pair
l may be helpful to make a chart with all consonant phonemes along the
with lines on the phone chart made in step I, or draw a circle around a
group of sounds each of which must be compared with every other sound
II vertical parameter and all vowels along the horizontal parameter, and to
note examples at the appropriate intersections of each combination of
consonant and vowel. (Chart 90 is an example of such a chart with xs
in the circle. S.ounds::al:e.comparea::"based::on theic:sjroilarjry to-ea9h:other
because of:shared:featnres. A knowledge of the kinds of processes that are
found in languages also helps in these decisions. Since this presents ;!
i rather than examples.) In some cases this sort of charting may reveal that
two segments really should have been analyzed as allophones. This may
-'

difficulty for the beginning student, we give more helps in this area on the
following pages.
I happen if two segments were not originally considered to share enough
phonetic similarity to have been compared in step 2. Another distribu-
tional check could be the phoneme's position with regard to tbe syllable.
-i
-~
For example, d= eve1'J"'<:'onsonant o"Ccur-in both syllable-..initial-and sy.1Ia, "

';
3. Contrast. ook:io mnjnjma,l- . S-OF near- minimal::paU:s.,."f0r each set of 'ble-filial:-pesition? On a higher level, this same type of check can be made .,;,
nholletically similar segments. Finding these minimal pairs establishes with regard to the phoneme's position in a word. These are only some of
~hat the t'W'o sounds contrast in identical or analogous environments and the various kinds of distribution to investigate. The discussion throughout ,
are therefore phonemically distinct from each other. the book regarding the kinds of phenomena relevant to different levels of
the phonological hierarchy should indicate the range of distributional con-
4. Complementary distribution. Chart environments of similar segments siderations which should be investigated in a thorough analysis of the
for which contrast cannot be easily demonstrated. If the segments are phonology of a given language.
found to be in complementary distribution, an allophonic rule like that in
(55) may be written. If complementary distribution cannot be easily dem- Phonemic analysis is an iterative process. A later stage in the analysis
onstrated or if any potential allophonic rules are not phonetlcally plaUSI- will frequently cause the analyst to go back to an earlier step and reevaluate
ble, however, search for the best examples to establish contrast in decisions made then. It must be kept in mind that all through the analysis,
analogous environnlents. each decision is a hypothesis which needs to be rechecked and confirmed
before it can be considered fact. The nature of phonological processes is es-
5. Phonemes and phonemic representation. Make a chart of the pho- pecially relevant to these earlier hypotheses.
nemes. At this stage the data can be rewritten in a phonemic transcription.
Such a transcription represents only the phonemes, not allophones. This
step assures that the analysis is thorough. As we pointed out in chapter l ,
this phonological or phonemic representation, as dlStUlct from the pho-
nctic representation, should more closely correspond to the Ultmtlons of
·- Phonel'mc Analysis
~
40 The procedure applied 41

The procedu re applied (63) a. CONSONANTS

Labial Alveolar (Alveo-) Velar/
We will now illustrate these procedures by applying them to data from palatal Glottal
Cashinahua, a Panoan language of Peru and Brazil. The data (from Ken-
Plosive vI P k
singer 1963 , with additions from Montag 1981) are given in (62)
(transcription slightly modified).9 Note vd b d
Affricate rs if
(62) [paka] 'bamboo' [misu] 'swollen hand' Fricative jJ s J h
[taka] 'liver' [m;Su] 'dark, black' Nasal m 11
[kaka] 'type of basket' [b;tu] 'spotted face' Flap r
[baka] ' fish' [bifsu] 'to be squeezed'
GJide w j
[daka] 'to rest' [bis~] 'face'
[fsaka] ' to kill' [kutSa] 'type of arrow'
b. VOWELS
[i}aka] 'bad' [kuSa] 'to hit'
[maka] 'rat' [naka] 'to chew' Front Central Back
High j j Ii

[tapa] 'floor' [tapa] 'washboard' Low a

[tarnal 'peanut' [tawa] 'sugar cane'
[bim] 'eye' [kuja] 'to have pus' In the second step, pairs of phonetically similar segments to be compared
• [kana] 'type of macaw' [hana] ' type of bird' are connected by lines in (64). In the section concerning phonetically simi-
[isi] 'unending' lisa] 'bird (generic)' lar segments we will say more about which pairs were sl!lectcd for
[isu] 'spider monkey' [is;) 'head painting ' comparison in (64).'0
~
[dani] 'body hair' [bari] 'sun'
(64) k j - j-- II
[ba] 'friend' [napu] 'people'
/ Ia
[iwiriwi] 'bring quickly!' [dasipi] 'all ' ij'
[dar;] 'medicinal plant' [binu] ' duck-like bird'
Uana] 'type of fruit' [sipi] 'weaving design' J--h
[hipi] 'type of palm' [jani] ' quickly'

[0
As a first s"tep. we chan all of the phonetic segments in the data in order
set::. which segmt:nts must be compared, shown in (63). C' w
i,
J

Contrasts can be demonstroted for most of these sets of pho!lt:tica lly sim-
9The phone ~~] uOc:s not occur in any of the Cash..inahua data presented in this volume. ilar se,gments with minimal (or ne:!r minimal) pairs, fo llow ing step 3, as in
Kensinger (190:208) gives the followi.lIg: (65).
[nllsu] . swollen hand'
[wIu] ',jade black'
lmi~u] ' :0 creep. crawl'
This and similar data should be consid,:red in a more complete analysis of Cashinabua. IOSee panicularly the discussion preceding (86).
42 Phonemic Analysis The procedure applied 43
;,-
(65) [P] - [b] [paka] 'bamboo' [baka] 'fish' (66) [b] [Il] [d] [r]
[P] - [Il] [tapa] ' floor ' [tapa] ' washboard' itu ta_a aka bLu
[b] - [Il] aka I1Q_U ani ba_i
[b] - [m] [baka] 'fish' [maka] ' rat' itsu dasLi aslbi iwi_iwi
[Il] - [w] [tapa] 'washboard' [towa] 'sugar cane' isu ari da_i
[m] - [n] [maka] 'rat' [naka] ' to chew' _iru
[w] - [k] [tawa] 'sugar cane' [taka] ' liver' ari
[t] - [d] [taka] 'liver' [daka] ' to rest' a
[t] - [is] [biro] 'spolled face' [biiSu] ' to be squeezed' inu
[d] - [n] [daka] 'to rest' [naka] 'to chew'
[d] - [r] From this type of charting it becomes apparent that [b] occurs only at the
'to kill ' [tSaka] 'bad' beginning of words, while [Il] is between vowels. Similarly, [d] occurs only
[fs]- ltD[fsaka]
word initially, while [r] is between vowels. Since these segments are in
[fs]- [5] [biiSu] 'to be squeezed' [bisu] 'face'
[5] - [fJ
[s] - [h]
[misu]
[sipi]
'swollen hand'
'weaving design'
[miSu]
[hipi]
'dark, black'
'type of palm' ! complementary distribution we may write an allophonic rule as in (67).

[fJ - [h]
[n] - [r]
[Jana]
[birlU]
'type of fruit'
'body hair'
[hana]
[biru]
' type of bird '
'eye'
[
ltD - [k] [tSaka] 'bad' [kaka] 'type of basket' !
I·
ltD - [fJ [kutSa] 'type of arrow' [k:uSa] ' to hit'
'type of arrow' [kuja] 'to have pus'
f This process involves some symmetry, but it is not completely symmetri-
ltD - [jJ [kutSa]
[fJ-[jJ [kuSa] 'to hit' [kuja] 'to have pus' cal. There is symmetry in that both voiced plosives are modified to have
'type of macaw' [hana] 'type of bird' less stricrure when they occur between vowels, resulting in different conso-
[k] - [h] [kana]
nants with less closure at the same point of articulation. But there is
[i] - [t] - [u] - [a] [isi] 'unending'
asymmetry as well, in that there is no voiced velar plosive corresponding to
[isi] 'head painting'
the other voiced plosives. In addition, Idl does not have a fricative
[isu] 'spider monkey' allophone [0] like Ibl does; instead it has a flap [r] allophone as a counter-

Il
[isa] 'bird' (generic) part to the [Il] allophone of I b/. This sort of partial symmetry is not
unco=on in language.
In step 4, the pairs of phonetically siroilar segments for which contrast The fifth step involves establishing the chart of phonemes, as in (68) . In
could not be demonstrated can be charted to determine if there may be enVl- this case it is identical to the chart of phones in (63) with the exception that
roruneutal conditioning, as in (66). [ll] and [r] are Omitted, since they are allophones and not phonemes.
 Phonemic Analysis Phonological wu·versals: Vowel and consonant systems 45
44
The final step with these data is to determine if tbere are any limitations
(68) a. CONSONANTS
on the distribution of phonemes. We would need considerably more data to
Labial Alveolar (Alveo-)
palatal
Velarl
Glottal
, make most of these determinations. At this stage, however, We might ob-
l' .
serve that Iwf does not occur word initially and f hl does not occur
Plosive vI
vd
P
b
I
d
rs if
k
I! intervocalically. All other consonants occur in both of these pOSitions. This
might alert tbe analyst to search for furtber data tbat migbt fill in these dis-
Affricate tributional gaps . If such data cannot be found, the anaiyst may want to
s J h explore tbe possibility tbat Iwf and f hf may in fact be allopbones of the
Fricative
Nasal m n same phoneme. This could be done with some psycho linguistic orthography
testing. For tbe present, however, we will tentatively consider f bf and fwl to
Flap
j be pbonemes with limited distribution since we cannot formulate a pboneti-
Glide w
cally plausible allophonic rule wbich relates the two to eacb anotber.

b. VOWELS
Front Central Back Phonological universals: Vowel and consonant systems
High i j u
Low a To some dc:gree, knowing what to expect in a phonological system can
be a considerable help in the process of analysis. Chapters 4, 5, and 6 are
In these data, the only allophones which differ from the phonemes which written to aid tbe beginning student in knowing wbat sorts of pbonological
they manifest are [Ill and [r]. Therefore, only words with these two sounds PROCESSES are considered to be natural and widely occurring ones. In this
will have a phonemic representation that differs from therr phonetic represen- section, we give a few introductory comments on phonological SYSTEMS
tation. The phonemic representation of these particular words IS presented ill wbicb are in accord with wbat is known about phonological universals, as
(69). an aid to the beginning student in knowing what sorts of phonological in-
ventories to expect. It should also be a belp for tbe beginning student to
(69) l tabal 'washboard ' make hypotheses regarding the directionality of rules .
I bidul 'eye'
I bodif ' sun' Vowel systems
fnobuf 'people'
fiwKiiwif 'bring quickly!' Crothers (1978) proposc:d a number of phonologica l universals regarding
fdosibif 'all' vowel sys tems based on a cross-13ngu;:gc: study of 209 languJges, :.l.$ well as
!dodi! 'medicinal plant' a review of the previous literunm! on universals affecring vowel systems.
Most of tbe following is from Crothers' study.
In d(ltng ana lysis. it is ll...lportant to show the phonemic repre~entation of
all fom,s . For forms that de not have any allophonic rules applymg to them,
,-. The minimJI \'owd system indlldcs Ii a ul , and 311 l:!r.gtl~lg::S are s;] id to
have these three vowds . Some languages may have one or more aT Illese
the phom:mic representation will be identical to the phonet~c re~resentatlOu. vowels varying slightly from these phonetic nom1" but :.dw:lys clUSk:-iflg
Showir.g a phonemic repre:;entation is one way of conflr~lll~g tn~t the anal- around theSe:! thr(!1! vowel SDa~C$. Some !alH!U~!.!.e~ have: bt:~!l reponed to
ysis accounts for all of tbe data. More importantly, It IS this level . of have only two vo'.Vd s. Cror!;c:rs suggests tum-the;\! ?rubably rcAec'! abs tr~c t
repc<sentation that is c\aim,d to correspond to a mental realtty of the native , analyses. which may be reana !yz(!d in a more CODCrct~ fashion to . yic'ld a
speake!". Sociopolitical facors aside (andas we sball see m the next chapter, standard three-vowd, four-vowel , or five-vowel sy:;tem.
if there are not relevant m0rphopbonemlc factors), this phonerruc level ap- Four-vowel systems are almost always t=ithcr Ii E a ul or Ii ~ Q u/ . Asheniuka
proxin' ates what a practicrJ orthography would optimally encode. is an example of the fITSt of these four-vowel systems, and Cashinahua the seo-
and (both languages of Peru). Languages with three-vow:;! or four-vov./el
'0
 !
t
46 Phonemic Analysis Phonological universals: Vowel and consonant systems 47

systems teq.d to show considerable allophonic variation of the vowels. In It is not simply fortuitous that these common vowel systems do not ha ve
Asheninka, for example, Ii i has allophones [1] and [i], and l al has allophones many back unrounded vowels, and that there are DO front rounded OIi~s.
[0] and [e]. The norm for the back rounded vowel in AshOninka is [0], but an Such vowels are considered to be unusual, although there are, of course~ a
[u] variant occurs as well. The system therefore quite naturally fits the norm number of languages which have them. Crothers again offers a functiona.I
for a four-vowel system. explanation for the normal correlation of front vowels with..unrounding aud
The five-vowel system I i e Q 0 ul is by far the most common, perhaps orb-ck vowels with rounding. With acoustic measurements done with a
representing the optimal system in terms of most efficient use of the vowel ~ spectrograph for front unrounded and rounded vowels, these two types
space to keep contrasting vowels perceptually distinct. The vowels lei and show little distinctiveness. For the hearer, then, they are perceptually rela-
/01 are often transcribed as lei and 10/, respectively, in descriptions, as in the tively difficult to distinguish. Likewise for back vowels, rounded versus
Spanish examples in (52). Crothers asserts that lei and /01 more usually re- unrounded variants are relatively indistinct perceptually. In fact, acollstic
flect the norm. Some ftve-vowel languages appear to have unrounded the measurements show front unrounded and back rounded to be at the ex- :1
normal lui , reSUlting in Ii e a a w i. tremes of the vowel space and the more unusual front rounded and bac :1
Conunon SLx-vQwel, seven-vowel, and nine-vowel systems are given in unroun e vowels to be more toward the center of the vowel space. This j
(70). led Crothers (1978:100) to distinguish peripheral and interior voweIS, as in
(71) (symbols have been modified).
.,•
;~

-'
(70) Six-vowel systems:
a. i b. j j (71 ) Peripheral Interior Peripheral
"
o a "
~
e £ ~ j u
£ ~
e 0
a
Seven-vowel systems:
c. j u d. j j

e o e a "o The functional explanation for the common vowel systems, as opposed
to the relatively unusual vowel systems, then, has to do with perceptual dis-
£ a ~ a
tinctness of the acoustic signal. Peripheral vowels, being at the extremes of
the acoustic vowel space, are easier to distinguish perceptually, and are,
Nine-vowcl system: therefore, more likely to be employed in natural language. Interior vowels
e. j j u
are less distinct and, therefore, come at a higber price. Languages are less
e o
£
"a ~
likely to llse them (or at least very many of them) because they are rela-
tively more difficult for the hearer to distinguish.
,.,
i
I
Crothers (1978:137) proposes three additional universals: (a) Languages
Eight-vowel systems, and systems with more than nine vowels are signif- with two or more interior vowels always have a high one. (b) The number
icantly less commonly attested, according to Crothers. of height distinctions in interior vowels will not exceed the number in pe-
Crothers suggests a functional explanation for why these are the predom- ripheral vowels. (c) The number of height distinctions in front vowels is
inant patterns. The basic notion is that the most common vowel systems equal to or greater than the number in back vowels. Each of these three uni-
utilize vowels which have the greatest perceptual distinctness from one an- versals is based on the notion of maximal use and distinctiveness of the
other. Starting with the five-vowel system as optimal, languages which have vowel space.
more (or less) vowels disperse the additional vowels in the vowel space, al-
lowing each additional vowel 10 occupy the greatest available space which Consonant systems
will allow it to be most distinct from the other vowels. Vowel space IS a
perceptual-acoustic notion, one measurable with instruments, e.g., sound Consonant systems, being more complex than vowel systems, are less
spectrographs, not a merely articulatory one. Even so, the notion of vowel conducive to yielding succinct universals. Two works that offer some
space is not unrelated to the traditional notion of articulatory vowel charts.
 !

48 Phonemic Afl3iysis

universals regarding consonant systems are Gamkrelidze 1978 and

II Phonological universals: Vowel and consonant systems 49

whereas if there is a gap in the voiced losive series, it will most likely lack
Maddieson 1984. Some of the generalizations given in what follows are r Ig!. The data from Cashinahua in (68) illustrates this ten ency bYilie asymmet-
from these two works. while others are based on our own informal observa- i n cal voiced plosive series whicb lacks Ig!. The same implied correlation ap-
tions of a variety of consonant systems. --P pears to hold also for modifications of the plosives such as aspiration or
All languages are expected to have at least the following consonant pho- \ glottalizatioD.
nemes: (a) voiceless plosives, (bj nasals, (c) a grooved fricative such as lsi, Gamlaelidze notes a similar tendency among noncoronal fricatives. For
and (d) a lary"Ugeal glide, usuaUf7Il/i,Some languages var'j from these the segments in (73), if a fricative is missing from the series, it will most
norms in m.inor ways. For example, some languages may have [fl, [ts), or likely be the one in parentheses.
[h] instead of [s] , and some may have [x] or [?] in place of [h].
An example of a language which has a rather minimal consonant system (73) (tj/f) x/II
is Pirahii. an unclassified Brazilian language (Everett 1986). jJ/v/w (}~

(72) P k ? Thus. for the voiceless fricatives. labiaJ.s...ar.~re likely to be absent, and
s II for the voiced series, the velar is more likely to be absent if there are asym-
m n metries in tbe system. --0

I
One final observation on consonant systems: When complex consonants, \
As with minimal vowel systems, langtll!ges with minimal consonant sys- those with aspiration, glottalization, prenasalization, palatalization,
.tems may be expected to have a fair amount of allo honic vana.!'on. Accord- labialization, etc., are found in a language, tbey usually imply the existence
ingly, Pira bas a number of fairly far-ranging consonant allophones. of the simpler consonantal counterpart at the same point of articulatioJ
The nom1 for obstruents js v oicelessness while the norm for sonorants is to There appear never to be more complex consonants than tbe series of sim-
be-Voiced. This follows narurally from the definition of the fearure [sonorant] ple consonant counterparts.
~ven earlier. Consequently, a language would not be expected to have voiced The inventory of consonants in Hausa (widely spoken in West Africa)
plosives unless it also has voiceless ones, and it would not be expected to have demonstrates some of these asymmetries (bere the apostropbe represents a
voiceless nasals or liquids unless it also has voiced ones. nonimplosive glottalized articulation).
Fric3tives are somewhat ambiguous in this regard.. however. Though thc:y
are usually c1assitied as obstruents, it does not necessarily follow that (74) If k ?
voiceless fricatives represent the norm. Tbis is likely due to the pOSition of
fricatives as being midway between prototypical obstruents (plosives) and
b d J5 g

trut!sonv:·ants. For grooved fricatives and affricates tbe obstruent/voiceless

6 a k'

correlati on does app-ear to bold: the presence of Iz 3 & 031 implies the exis- s }

lsi in the language. On the otber f s J

tence of tho: voiceless counterparts Is S fs
hand. voked fricatives which are not grooved, e.g .. l{3 v 0 y/, are more corn-
mOll (or .t kast no less common) than tbeir voiceless counterparts 14> f e xl .
v "
m II
Tht:se t~: cts arc: likedy due to a greater constriction with grooved fricatives. i
Gamkrolidze (1973) and Maddieson (1984) provide evidence for another r
correlatio"1 be[ween point of articulation and voicing of obstruents. Spe-
}
cifically. :here is a high correlation between anterior obstruents (especially
labial,) and voicing; and thert is a high correlation between posterior obstru-
Here lhl! "'missing" vuicdess plosive is bilabial Ip/ ; among the glottalized
ents (especially velars) and vcicelessness. For voiceless plosives. for example,
consonants only anterior 161 and 1& are voiced; there are no glottalized con-
the bierar,;b):/pl < ItI < Iki generally bolds; and for voiced plosives the op-
sonants that lack a nonglottalized counterpart; th~re are no voiced fricatives
posite hie,arcby Ibl > Idl > Ig! generally bolds. What this means is that if a
that lack a voiceless counterpart.
language bas a gap in its voiceless plosive series. It Will most likely lack Ipl,
1
 f
!
r
50 Phonemic Anl1/ysis
I identifying phonetically similaT segments 51

Nasality The explanation for this runs somewhat along the same lines as the
In this section, we present some universals and tendencies relating to na-
sality, a feature which commonly interacts with both consonant and vowel
systems. Ferguson (1963,1975) and Ruhlen (1978) give the best statements
I above. Vowels in a nasal context, either nasalized vowels or preceding a
syllable final nasal, are susceptible to losing height contrast. For example,
some dialects of American English no longer maintain a contrast between
the vowels in pin and pen. Again, with nonopen vowels, nasalization is
concerning nasality universals of which we are aware. more difficult to produce and perceptually . less distinct. The solution opted
Nearly all languages have .nasal consonants. As is the case with for in languages like Dakota is to utilize the extremes of the vowel space to
Cashinahua (68) and Hausa (74), if there are gaps in the series of nasal con- maintain perceptual distinctiveness among nasalized vowels. In such sys-
sonants, it is most likely velar Inl or other nonanterior nasals that are tems, the mid nasalized vowels are therefore the most likely ones to be
missing. There are rare examples of languages which are described as hav- miSSing.
ing only one nasal consonant.
Slightly more than one-fifth of the 700 languages in Ruhlen's sample
have nasaliz weIs which contrast with oral vowels. The normal case is Identifying phonetically similar segments
for the set of nasal vowels to be exact nasa ized counterparts of the oral
vowels. It is not uncommon, however, for a nasal vowel system to be In this section, we give some helps in identifying which pairs of phoneti-
smaller than the oral vowel sY$tem, in which case it is usually mid nasal cally similar segments should be compared in order to determine whether
vowels which are miSSing. Les,; commonly, some languages lack high na- they contrast or are in complementary distribution. Note that there are no
salized vowd counterparts. hard-and-fast rules; that what we offer here are guidelines reflecting the
There are natural explanatior.,; for both of these tendencies, where ~ _kinds of patterns often found in Ian a~. By expandmg the range of
vowel systems are sma than the---"OLtesRonding Q[~. phones to be compared beyond what we suggest here. the analyst does not
Frenc is an example of a language which has only relatively open nasal, err, for the analysis will reveal the nature of the phonological system in any
ized vowels. case. The only cost is that the amount of work involved is expanded.
We compare ~uJ:1os~ents whic sh.a~ the greatest number of fea-
(75) French Oral Vowels French Nasal Vowels rures. If a phoneme has more than one allophone, we would expect these
iy u ---.rrophones to have a good number of features in common. The variant
eo
Ere a ,o allophone should differ from the phoneme of which it is a submember only
in the features which are being affected by the environment. .So we would
a re d compare [p] and [Ph], for example, since they differ only in aspiration. We
would not compare [p1and 1:3] since the only thing they have in common is
Open vowels, pronounced with the tongue lowered and the jaw open, alJ that they are both obstruents. Accordingly, we would not expect to encoun-
low for a greater degree of velie ope~g. Nasalizat~on is thus easier to ter a language in which Ipl had an allophone [3] . Thus, the primary
produce with the more open vowels and IS also more likely to be percepru- consideration in isolating pairs of phonetically similar segments is ihe shar-
ally distinct. ing of phonetic features. This is particularly applicable when segments are
Dakota, an indigenous language of North America, is a clear example of at .the same point of articulation and only vary somewhat in their manner of
a language lacking mid nasalized vowels: articulation. I he pnnciple here is that in languages such sounds are some-
times found to function as allophones of the same phoneme. J
(76) Dakota Oral Vowels Dakota Nasal Vowels
i u j II Phonetically similar consonants
e o
a ii For example, if a number of bilabial consonant phonetic segments are
found in a set of data, we would compare the pairs connected in (77).
 52 Phonemic ATUJ./ysis
![ Identifying phoneu'cally similar segments 53

(77) ph [4» contrasts with [p] , and (P) contrasts with [hl, it would not be expected
I tbat [4>] and [b] be related allophones, since they differ by more features
tban either of tbe first two pairs. If in this scenario [<p) and [p) were found

(It to be allopbones, but (P] and [b] in contrast, our comparison of [p] and [b]
would still suffice, i.e., it would not be necessary to compare [ep) and [b],
since the PHONEMES Ipl and I bl would have already been contrast ~Ulti­

e)
mately. it is only necessary to demonstrate the contrast between phonemes
using wbatever allophones are most conducive to show the contrast. It is
not necessary to demonstrate contrast between all the allophones of differ-

(m- (another nasal consonant)

ent phonemes.
At the velar and glottal points of articulation, we would compare rhe
connected pairs in (78).
w- (velar)
(78)
(P] and [b] differ by voicing
[4>] and [J3) (alveopalatal) - - - ?
(P] and [~] differ by stricture
[b] and [J3)
[J3] and [w)
[b] and [m) differ by position of the velum
(P] and (Ph) differ by aspiration
[4>1 and (Ph) differ by stricture and aspiration

We assume that [w) will be conapared with sonae velar sound in the system (another nasal) V
(since it is technically a labiovelar). while [na] will be compared with an-
other na,;aL Seven of the sets given above differ by only one feature, while [k) and [g) differ by voicing
the last differs by two features, as indicated. -0 [x) and [y)
Notic that we have cOalpared [ph) with [ep] but not with [b], even 1 [k] and [x) differ by stricture
thougb each pair differs by two features «(Ph] and [b) differ by voicing and [g] and [y)
aspiration). Bc:cause there are still some theoretical questions regarding the [?] and [h)
universal set of pbonc:tic features. counting features is not always accurate [k] and [?) differ by point of aniculation
in detennining which segm<!nts are the most phonetically similar. That is, [x) and [b]
features do Dot nil carry eq!livalent weight. For example, in some respects [g) and [ry) differ by position of the velum
- voicing is a more crucIal or pervasive difference than a modifying feature [k) and [kh) diffa by aspiration
like aspiration. This says, then, that [ph] is more T@y to ave an [x) and [kh) differ by stricture and aspiration
allopboDlcrelationship !"itl:· anotber voiceless bilabial tban with a voiced
one. Therefore. it is compared with [4>] rather than [b) in this panicular set Tbe sound [ry] w ill generally be compared widl another DOsa!. As with
of sounds. If there were no unaspirated (P) in the data, bowever. we would tbe bilabials discussed earlier. the voiceless "spir ited p[osive [kh] will be
want to compare (Ph) with [b). --b . ~ compared with the voiceless tricarive [x) rather tkn the voiced plusive [g).
We would compare (P] with [4>] since they differ only in stricture, and In addition to this, if the language bas alveopalatcJ sounds. at least olle of
we would compare (P] with [b) since tbey differ only in VOicing. But we tbe velar obstruents should be compared with an alveopalata!.
would Dot necessarily compare [ep) with [b). The assumption bere is that if
~
54 Phonemic Analysis Identifying phonetically similar segments 55

It is not as, crucial to look for proof of contrast or complementation with [t] and [d] differ by voicing
some pairs of segments that differ only in point of articulation. The seg- eel and [0]
ments [p] and [t], for example, are not expected to be in an allophonic
relationship ir) any language, even though they share a good many features.
Languages do not commonly have processes whereby a labial oral conso-
r [s]
[JJ
and
and
[z]
[3]
[fs ] and [tD differ by point of articulation
nant becomes a corresponding alveolar consonant in some environment, or eel and [s]
vice versa. The articulatory distinction between sounds made with the lips [s] and [JJ
versus sounds made with the tongue is greater than the differences among [0] and [z]
sounds made with the tongue. [z] and [3]
Sounds made w ith the tongue (interdental, alveolar, alveopalatal, and [n] and []I] "
palatal points of articulation), however, are quite susceptible to being in- [d] and [r] differ by manner of articulation
volved together in allophonic processes. Somewhat less commonly, velar [r] and [I]
sounds may be in an allophonic relationship to a counterpart farther forward [fs ] and [s]
in the mouth. For example, [Ie] has an allophone [tD in some languages, [tD and [JJ
usually affected by proximity to high front vowels. The fronlness of the [d] and [I] differ by stricrure and manner of closure
vowel in this case causes the stricture to move forward in the mouth. Since [d] and [n] differ by position of the velum
interdentals, alveolars, alveopalatals, palatals, and velars are all made with [th] and [t] differ by aspiration
the tongue, as opposed to the lips, there is a greater likelihood that they will [t] and [fs ] differ by fricative release
be involved together in some sort of allophonic relationship, and very little [th] and [fs ] differ by aspiration and fricative release
possibility of their having such a relationship to labials. [t] and [e] differ by stricture and point of articulation
At the interdental, alveolar, and alveopalatal points of articulation we [r] and [n] differ by manner and velic position
would compare the sets connected in (79).
Nasal consonants are an exception to the notion that it is unnecessary to
(79) th compare labials with other points of articulation. In example (2), we drew
I attention to the extremely common process of assimilation of a nasal to a
following obstruent. Nasal consonants differ rather radically from all other

1) manners of articulation for consonants. In general, moving downward on an

e
J fS -
I I
s - - J
I I
-
, /(Velar)

tJ
articulatory chart corresponds to less closure in the oral tract. With nasals
there is complete closure in the oral tract, but sonority in the nasal cavity
due to the lowering of the velum. This peculiar property of nasals tends to
result in their commonly being involved in clusters with other consonants,
and in assimilating to their point of articulation. Because of this, nasals are
usually compared across the board. With sets like those in (80), it is usually
a Z --3 advantageous to compare [n] with every other nasal, since the alveolar is
more commonly the phoneme from which the other nasal allophones might
be derived in this type of assimilation.

;) (80) _________
m - - Il - - J' - - Y
11 - - Jl - - (another nasal)
Another anomaly concerns the semivowel [wJ, which is classified as
labiovelar. Since it involves both labial and velar articulation, it should be
56 Phonemic Analysis Identifying phonetically similar segments 57

compared with its closest labial and velar neighbors. There are other less (84) a. I" b.
common labiovelar consonants such as [q,] which should also be compared
with both labial and velar counterparts. (The labiovelar plosive is not un-
I~ ~"tJ
" t h
common in African languages and is found less frequently in other parts of
the world.) I
J
(81)
t----g
w
c. k - - k"
I
w
d.

"p
e.
~ .
e - - e'
The glonal plosive and fricative-are also somewhat outside of the norms I
m
for comparing sounds made with the lips and tongue. The glottal plosive [1]
should be compared with all other voiceless plosives, and the glottal frica- Phonetically similar vowels
tive to all other voiceless fricatives. In addition, [1] and [h] should be
compared to each other. These comparisons are due more to acoustic simi- F or vowels, the notion of feature similari~s, in general, a sufficient
larilies than articulatory similarities. The glottal sounds involve no specific guide as to which vowels should be comparedI It is not usually necessary to
articulations of the tongue or lips. But the stopping of air at the glottis may compare vowel sounds that are at opposing extremes of the vower
sound enough like the stopping of air in [P], [t], or [k] for any of these to be chart-such as [i], [a], and [u]-since these are expected to occur in all lan-
in an allophonic relationship. Similarly, the friction sound of [h] may have guag~Otherwise, and especially if there are a large number of vowel
acoustic properties similar to other fricatives. Some languages have an segments in the phonetic inventory, it should be helplul to compare all.pai.t:s.
allophonic process called DEBUCCALIZATION whereby a consonant like It! of vowels which differ by one, two, or, at the most, three f"a!ures. The ma-
has an allophone [1] in an environment where a weaker consonant might be jority of the pairs listed in (85) differ by only one feature. In the case of the
expected. Thus: last seven pairs. more than one feature is called in.

(82) _-~ (85)

P I k - -?
[ i]
[ I]
and
and
[e]
[e] illff<, >, d~"" (
"\
, ")
I
I I I I
.
[e] and

\\T- " ~t
['" ]
,p s x - -h [u] and
. ------ -----~ [0] and
[0]
[0]

~ 1"',, /
[i] and [0]
Finally, [k] and [h] should be compared if there is no [x] to compare to [ I] and [e]
both of these. since it is possible for [h] to function as a weakened back fri-
cative in the absence of a true velar or uvular fricative.

(83) k
(e]
(x]
[u]
[0]
aod
and
and
and
['" ]
[a]
[0]
[0]
,\/pJa
~h [i]
(e]
and
aod
[I]
[f]
differ by closure, possibly AfR

[u] and [u]

When segments involve two articulations, the complex segment should [i] and [ i] diller by backness
be compared to both of the members that compose it. Some examples of [e] aod [0]
this ar< aspirated stops, affricates , palatalized or labialized sounds, [a] and [A]
prenasalized stops, and diphthongs.
 i

58 Phonemic Analysis
I,t Free variation
59
I
rot·and
[£ ]
[i]
and [A]
and [u]
[A] differ by backness, closure
"
differ by backness, rounding
I,I [Is)
[k)
[n)
and
and
and
[k)
[h]
[c]
differ by release, point of articulation
differ by stricture, point of articulation

[0) and [u] , .

[i) and [i]
differ by velic pOSition, manner of articulation
differ by backness
[a] and [0] " !
[i] and [a] differ by closure
[A] and [0] differ by closure, rounding ,I [i] and [u] differ by backness, rounding
[A] and [0] [u] and [a] differ by closure, backness, rounding

As mentioned above, the informal guidelines given in this section should ;. ",,,,,,, ""'" c.lr h c:;..e--">
not be taken as an absolute authority. Which segments to compare fIrst will
vary with each particular constellation of phonetic segments encountered. 1£.
Free variation <: b~~ .rJ,~.e-"
be absolutely certain not to miss any pairs of segments which might be in At tbebeginning of this chapter, we stated tbat there were three relation-
comp lementary distribution, it would be necessary to compare all pairs that ships which phonetically similar segments may bave in relation to one
sbare any feature in Common. In some contexts this may indeed be appropri- another-contrast. complementation, and free variation. We have discussed
ate, e.g., actual field investigation of some previously unstudied language with the fIrst two above and now turn to free variation. There are two types, the
an abundance of data. In the pedagogical setting of this text, the helps above fIrst of whicb can be illustrated by use of the Englisb data in (87).
sbould suffice. If the student follows these guidelines, all of the allophones in
"the data will usually be discovered. Charting the distribution of phonemes in (87) [kbreCJ - [kbret'] ' cat'
relation to one another can serve as a safeguard against nllssing some pairs [biit' ] - [biit'] 'beet'
that may be in complementruy distribution as well. ,- [Wp '] - [liipb] ' leap '
To summarize, look again at Cashinahua. The pairs given in (65) are [kbijp'] - [kbiipb] 'keep' .
relisted below based on their minimal differences. [soWk' ] - [sowkb] ' soak'
[tbeik'] - [tbeikb] 'take '
(86) [p] and [b] differ by voicing
[t] and [d] " It is a fact of English that, utterance ftnally (or here in single words pro-
[b] and EJ3] differ by stricture nounced in isolation), plosives occur as either unreleased or aspirated. It makes
EJ3) and [w] " no difference to the speaker of the language:-- r he sarne word is being pro-
[fs ] and [s) nounced, whichever form is used. In some cases there may be some dialects or
[t] and [fs) differ by release social groups which tend to use one pronunciation over another, but usually
[d] and [c] differ by manner of articulation there is little or no perception that the pronunciations are distinct. We may
[b) and [m) differ by position of tbe velum conclude that plosives in English have at least two allophones, and that the al-
[d] and [n) lophones in question vary freely, i.e., without known conditioning, in utterance
[m] and [n) differ by point of articulation final position. This fact is representea'6y the informal rule (88).
[s) and [SJ "
[SJ and [b] (88) t - t' - t' / _ # (utterance tinal)
[s) and [h]
The concept of symmetry suggests tbat we would fInd the same for Ik!
[rD and UJ and /p/, and the data in (87) show tbat as well . Because [t'] and [t'] never
[fs] and ltD
[P] and EJ3] differ by voicing, stricture contrast in the language, we may describe this pattern more precisely as
[w] and [k] FREE VARIATION BETWEEN ALLOPHONES.
[tD and [j] Now consider the English data in (89).
[SJ and [j]
 60 Phonemic Analysis Phonotactics, neutralization, and wlderspecification 61

(89) [iikan=lks] [Ekon=lks] 'economics' sequence of Iskl is permitted at the beginning of a word, while I ks /, in this
..H<';r~] [drer~] 'data' environment, is not.
[tb~meirowl 'tomato'
Distribution of phonemes in relation to surrounding segments

t
If we gather data from different speakers of English, or perhaps from the
same speaker at different times, we are likely to get data like that in (89). Phoneme distribution can be restricted at any level of the phonological
Similar patterns are found in vh-rually every language. Further, when we hierarchy sketched in (12). At the lOWest leVel of the hierarchy there may
ask speakers which is CORRECT, we are likely to get the response that either be limitations on which segments can occur contiguous to other segments.
is accept,ble. Again, there may be dialects or social groups that tend to use In some languages, the semivowel [j] does not occur before or after the
one pronunciation over another. Speakers mayor may not be able to verbal- vowel [i] and the semivowel [w] does not occur before or after the vowel
ize these factors. Whatever the case, they interpret the alternative forms in [ul The semivowel in each case is very similar to the vowel. When both se-
(89) as being variants of the same word. quences of Iji/ and IiI are present and contrastive in a language. there are
These data differ from the examples in (87) in an important way, how- perceptual difficulties for the hearer to distinguish them. So sequences like
ever. The variation found in the alternatives in (89) is variation between Ij i/ are avoided in many languages. Cashinahua is an example of such a lan-
phonemes which are contrastive elsewhere in the language. (VVe can readily guage (though the data in (62) do not fully illustrate these facts).
produce minimal pairs to demonstrate these contrasts for English, as in l! is helpful to chart language data to determine if there are such distribu-
bead vs. bed, bait vs. bat, and hate vs. hot.) A native speaker may be unable tional limitations. The simplest way to do this is to list. for example, all
to even recognize any difference between the variants in the data in (87) . consonants along one parameter and all vowels along another, and register
For those in (89), however, the difference will be readily recognized; but examples of each particular consonant-vowel sequence to fill in the matrix .
•
the response will be that the two forms are simply variant ways of pro- In (90), we illustrate this sort of chart for Cashinahua (based Oil Kensinger
nouncing the same word, although the speaker may associate the difference 1963), though the chart below relies on a more inclusive set of Cashinahua
with some regional or social dialectal factors. Such free variation we de- data than that found in (6~).
scribe as FREE VARJATION BETWEEN PHONEMES because the sounds are
contrastive elsewhere and thus are differ.enLph..o.nemes. (90) i j 11 a
To surrunarize the three relationships discussed: Two phonetically similar
segments may be in contrast, in complementation, or in free variation; the p x x x x
~e variation itself may be between sounds which are elsewhere in contrast, x x x x
or which are not. k x x x x
b x x x x

. \?~~-r~
-/ Pbonotactics, neutralization, and underspecification
d
rs
x
x
x
x
x
x
x
x
), .......-<'- . (f' ~
.& • l:,.....· if x x x x
v A final step in phonemic analysis is to detennine what limitations there
S x x x x
mi;ht 1: e in the distributior of phonemes. Such characterizations of the dis-
tribution of phonemes is sometimes referred to as PHONOTACTICS. Most .f x x x
Ii x x x x
languaf'es have some of these sorts of constraints which define what is a
1/1 X X X X
possibk word in the langm'ge. For example, speakers of English could con-
II X X X x
ceive of skap as a possible English word. A new dog food could be
markett:d with this name alld it would not strike speakers of English as for- r x x x x
eih'Il soelIlding. A name lik" ksap, on the other hand, would immediately be w x x
perceived as a non-English-sounding word. In English, a consonant j x x x
 r
62 Phonemic Analysis
I Phono/aeries. neutralization, and underspecific8rioll 63

In CashInahua the gaps in the distribution of phonemes in sequences of Pistribution of phonemes in the syllable
consonant plus vowel are few. In addition to not allowing sequences of Ij iI
and lwu /, the only other disallowed sequences are ISil and Iwi /. In general, In that segments are combined into syllables, syllables may properly be
we expect the distribution of phonemes in a language to be complete. Gaps considered to be the phonological level just above the segment. In fact, syl-
in their distribution are normally due to some articulatory or perceptual dif- lables are the minimal pronounceable pbonological wlits. No segmental
ficu lty with such a sequence. We have already suggested that the difficulty phone can be PRONOUNCED without taking on syllable qualities and dynam-
involved with Iji l and lwul is a perceptual one. The difficulty with ISil and ics. The syllable is a structural unit into which segmental pbonemes are
Iw il is more likely to be an articulatory one. The segment lSi is specified as distributed. We will deal with tbe syllable in more detail in chapter 5. In
[ + high] and [- back] while Ii i is [+ high] and [+ back]. With the tongue this section we merely point out a few respects in whicb the syllable is rele-
already in tbe high front position for lSI, if the following sound is to be an vant to phoneme distribution.
unrounded high vowel, it is easier to keep it in that position and produce the All languages are thougbt to have the syllable type CV, and some lan-
vowel sound Iii rather than Ii i . So if there is to be a high unrounded vowel guages allow only this type. In such cases, the distribution of pbonemes
following lSI in Cashinahua, it will be the one that shares the [ - back] fea- within the syllable is a straightforward matter: A simple matrix like that in
ture with IS/. (90) will sbow the distribution of segments into the syllable pOSitions . For
Along the same lines, Iwl is [ + high] and [ + back] while Iii is [ + high] and more complex syllable types, distribution of phonemes is not a mere matter
[ - back]. With the tongue already in the high back position for Iwl , if a high of sequence distribution. In many languages witb CVC syllable types, tbere
Wlfounded vowel follows , it is easier to keep it in that pOSition and produce the are restrictions on which consonants can OCCur at the end of syllables. In
vowel sound Iii rather than Iii . If there is to be a high Wlfounded vowel fol- Cashinahua (Kensinger 1963) only fricatives can occur in this position, and
lowing Iwl , it will be the one that shares the [ + back] feature with Iw/. These in Asheninka only nasals can. In CashInahua, the fact that l si and lSi can oc-
distributional constraints are specific to CashInabua, just as allophonic pro- ( cur as the first member of a CC sequence where two syllables come
cesses can be specific to individual languages. together, and word fmally as well, is due to their being allowed in sylla-
Since the class ification of segments into consonants versus vowels is the ble-fmal position. This fact would, therefore, be accounted for in describing
most important one in phonology, our concern for determining the distribution tbe distribution of pbonemes into the syllable.
of phonemes follows that basic distinction. At this level we will want to check
not only all sequences of consonant-vowel (CV) as in (90), but also Distribution of phonemes in the word or phrase
vowel-consonant (VC). as well as all sequences of vowel-vowel (VV) and con-
sonant-consonant (CC) to determine what distributional limitations are in In some languages there are limitations on the distribution of phonemes
effect. Of course, if there are clusters of three consonants or three vowels, or that are relevant only in relation to levels higher tban the syllable. For ex-
longer still, we will want to know what limitations there may be on these as ample, in Asbeninka the syllable-final nasal consonant never Occurs at the
well. It requires a large body of data to determine accurately whether such lim- end of a word. In a number of languages, only voiceless consonants are al-
itations are genuine properties of the language. If the analyst attempts to deter- lowed word fmally or utterance fmally. In Arabela, a Zaparoan language of
mine these limitations with too little data, the limitations will really only Peru, the phoneme /11 only OCCurS contrasting with its absence at the end of
reflect the fact that not enough data were examined, not that the language truly a breath group. In a question asked with hesitation or uncertainty, as in
bas tbe limitations. (9 Ia), the glottal plosive at the end of tbe breath group does not occur,
Some languages bave restrictions on the env ironments whicb both pre- while in a question asked with surprise, as in (9 1b) it does (Ricb 1963, tran-
cede and follow certain pbonemes. For example, in Asbeninka the velar scription modified).
fricative Iyl must be both preceded and followed by l a/. In this position it
contrasts witb all other consonants, but Iyl OCCurS in no other contexts. (9 1) a. [taan niislxinii] ' How did he say it?' (uncertain)
Charts like those in (90) demonstrate wbether any phoneme bas this sort of b. [bsaan heeYlnl'l] 'What is that?' (surprise)
limited distribution.
At the level of the breath group, then, I?/ is a contrastive (intonational)
pboneme.
 Phonemic Analysis Phonoracrics, neutralization, and underspeciJicatiol1 65
64
7 either of the informal phonological statements in (93) is possible for the
Neutralization
plosives involved.
It is commonly the case that even though a contrast between two phonet-
ically similar segments can be established in a language, that contrast
cannot be found in every environment. We may say that the contrast is
(93) a. h,d,g - p,t,k 1 _ #

-
b,d,g elsewhere
NEUTRALIZED in those environments in which it does Dot occur, and we p,t,k p,t,k (occurs in nonfinal position only)
may refer to the pattern itself as NEUTRALIZATION. Limited distribution
which does not involve phonetically similar segments is not neutralization,
since for dissimilar segments there is no concern in the first place to dem-
b. h,d,g
p,t,k
-- b,d,g (occurs in nonfInal position only)
p,t,k (in all environments)
onstrate contrast.
Russian is one example of the commonly found restriction where voiced Languages might occasionally have rules like (93a). All things being
plosives are found word initially and word medially, but not word finally, as equal, however, we would normally prefer option (93b) because it is sim-
indicated :in (92) (data from Brett Benham, personal communication). pler than the other analysis. In fact, under that analysis the rules in (93) do
not really need to be stated. Instead, we need only state the general condi-
(92) NOMINATIVE GENITIVE tion that "only voiceless plosives occur in word fInal position." (The fact
SINGULAR SINGULAR that in the examples above it is evident that morphemes show variation as
to whether their final consonants are voiced or voiceless , it is clear that '
snop snobi 'snob' there is more involved here than simply phonology. We reconsider this
...,
,~
snop S1l0pa 'bale (of hay)' analysis when discussing a comparable set of facts regarding German in
chapter 3 where we discuss morphophonemics.)
kot k6da 'code'
Now consider the data in (94) from Taiwanese (with tone and some pho-
kot kOlEi 'cat' netic detail omitted), which illustrate a sligbtly different type of neutralization.
:: nwk maga 'magician'
mak nuika . poppy seed' (94) p'ak' 'hang clothes' kak' 'born
gnef gntiva 'anger' kne 'stream' pIl~ 'pen'
ser sela 'chief k'a 'foot' tap 'answer'
(hay , insect' kUl ' 'bone'
glas g/;iza "eye'
(h 8 k' 'read ' bk' 'nation'
glas gl:isa 'voice' (of reason)
p';)k' 'respected' pet' 'others'
etiI Eca,}-i 'story, floor '
fnlaI JalaJii ' hut' For Taiwanese the siru3tion is more complex than for Ru ssian. Tilere is a
contrast bt!rweeo aspirated and unaspirated plosive:; in initi3! position, so it
In Rllssian. voiced and voiceless obstruents are paired in phonetically is clear that at each point of articulation there are r-.vo plosive phonemt:s as
similar :;egments. and they are contrastive with one another in nonfinal po- in the Russian data in (92). But in Dnal positi on only urrreleaseJ plo:;ives
sition. Thus, in these data, we find the obstruents Ip t k f s SI and Ib d g v z occur at each point of atticulation.
31 nontlnally. Because the phone occurring in final position matches lll:!ither of the
But now note that in final position only voiceless obstruents occur, There phones occurring contrastively in initial position. i1 is not qu it ~ as stra :ght-
is no coatrast between voic(!d and voiceless obstruems in final position, and forward in Taiwanese to detettnin e which phonemic plos ive the Dnol
thus the contrast is neutralized in that context. The analytical question we unreleased plosive is manifesting. It may be equally plausible 10 say that as-
face (and the theoretical controversy involved) is that of ,determining the pirat~d plosiv es are unreleased word finally, in whic h C3se the tinal plm: i \' ~ s
phonological nature of the obstruents which occur in fInal position. Spe- are treated as allophones of the aspirated phonemes ', or to say tbat voicdess
cifically, because there is no contrast and only a single phone occurs there,
.,
66 Phonemic Analysis
r Phonotactics. neutralization, and underspecification
67
unaspirated plosives are unreleased finally, in which case the fmal plosives
languages, we expect the unaspirated ones to have a wider distribution in
are allophones of the voiceless unaspirated phonemes.
any individual language. In this case, the fmal unreleased plosives would be
The ana1yst faced with data like those in (94) will need to dig further to considered allophones of the unaspirated plosives.
determine which of the two word-initial plosives might be more likely to
The predOminating view in many current theOries of phonology is that
have the Imreleased plosives as allophones. For example, if in Taiwanese
the unreleased plosives should not be assigned to either of the word-initial
there were aspirated affricates word initially, but no unaspirated ones, and if
plosives, but considered to be neutral or unspecified (or underspecified) for
simple unaspirated affricates occurred word fmally, we would know that a
the feature [spread], a feature which is contrastive in initial position. In
rule of de-aspiration was operative in the language. In such cases by the
such an analysis, the unreleased plosives in word fmal position have been
tendency for symmetry, we might opt for the analysis of the aspirated called UNDERSPEClFIED SEGMENTS or ARCHIPHONEMES.
plosives as having the unreleased plosive allophones word fmally. It is not uncommon to see archiphonemes represented with upper case
However, Taiwanese actually has both aspirated and unaspirated letters, as an indication of their UNDERSPECIFIED nature. In such a phono-
affricates word initially, and neither occurs word finally. And there is no logical representation, the forms in the second column of (94) would be as
other snuctural evidence to suggest that one analysis is preferable. in (95).
English is another example of this second kind of neutralization. The
phonemes It I and Idl contrast in word-initial and word-fmal positions (to (95) IkaKI 'horn'
versus do. hal versus had) yet between vowels (where the second vowel is IpiTI 'pen '
unstressed) neither occurs. Instead, a flap [e ] (or in some transcriptions a ItaPI 'answer'
flap ped d) occurs as in laller and ladder. Intervocalically, the contrast be- I kuTI 'bone'
~
tween ItI and Idl is neutralized, and only [e] occurs. Therefore, [e] could be I bKI 'nation'
analyzed as an allophone of either It I or Id/. - IpeTl ' others'
Lacking conclusive internal or structural data to resolve the issue in such
~
cases, the analyst may look for extemal evidence; that is, evidence not The notion that underlies this analysis is that these fmal plosives are
found in the data under analysis. In fact such evidence is always the last specified in their phonological representation for all features except
word in validating an analysis, even one which seems as clearcut as those [spread]. It is considered redundant, and therefore urmecessary, to represent
discussed earlier in this chapter. The topic of external evidence is deserving the aspiration since this can be done by a FEATURE FILLING rule, like (96),
of considerable attention, but only an introductory presentation of the no- once for all such forms.
tion can be made here (for a fuller discussion the reader is referred to
Kenstowicz and Kisseberth 1979: 154-175). (96) [- sonorant] - [ - spread] I _#
In looking for external evidence, the analyst is looking for some indica-
tion of what is cognitively REAL to the speaker, for it is that reality which is For the Russian examples in (92), the final consonants can also be ana-
expected to be manifested in the analysis. The analyst might study related lyzed as archiphonemes, that is, as unspecified for voicing. The data for
languaaes to see what patterns can be supported as being charactenstlc of Russian are more complex than this might lead us to believe, however, as
the ....lan~age group as a whole. More to the point, however,. are data indi- we indicate in the following chapter. Suffice it to say now that other types
cating more directly how the speakers of the language mterpret such of evidence are often needed in such cases to elucidate the appropriate anal-
patterns. Observation of speech errors, word games (like pig latin), and .or- ysis. In RUSSian, this additional evidence comes from another aspect of the
thography experiments, e.g., facts detailing the relative ease of teaching grammar called morphophonemics. In other cases, such evidence may come
reading and writing using the alternative analyses, are common approaches from outside the language data themselves, such as native-speaker reactions
here. There are, however, rare cases in which no relevant data can be found. to linguistic data.
!l is in these cases that the theoretical controversy is found. As a rather clear example of underspecification, consider the syllable-fi-
In the past, some analysts have made arbitrary analytical decisions in nal nasal in Asbeninka (Payne 1981). Nasal consonants Im/ and Inl contrast
such cases. Others argue, in the case of Taiwanese, that since aspirated syllable initially and intervocalically in Asbeninka. Preceding obstruents,
voiceless plosives are more rare than unaspirated ones in the world's however, only nasal consonants at the same point of articulation as the
 68

following obstruent occur. A period (.) separates the Asheninka syllables in

Phonemic A1UJiysis
I Phonotacrics, neutralization, and Illlderspeciticatioll

b. [tham mwe] 'boats'

69

(97). [i'sben mwe] 'hoes '

[nd3aD mWe] 'bows '
(97) [ma.to] 'moth' [no. to] ' medicinal plant' [tbul mwe] 'water pots'
[ma.pi] 'rock' [am.pe] 'cotton' [?or mwa] 'roads'
[ma.ni.ro] 'deer' [an.ta.mi] 'jungle'
[a.na] 'black dye' [hJ.kiJ 'peanut' c. [tharn] 'boat'
[i'sb;l} ] 'hoe'
Several native speakers were taught to write syllable-initial nasals m and [nd30lJ] 'bow'
n as in mapi 'rock' and ana 'black dye'. They were also taught to write n [thu!]' 'water pot'
before I for the syllable final nasal as in anlami 'juagle'. Wheo writing the [?or] 'road'
syllable final nasal before p , however, (as in ampe 'cotton') they chose 11
(a/lpe). They later learned to write m before p, as it is in Spanish, and each Here it cao be seen that [m], [n] , [D], [1] , aod [c] occur initially (98a) aod
of them then consistently began to write m also before l (am/ami). We may medially (98b), and their voiceless counterparts occur only utterance fmally
take these facts as external evidence that at some uaderlying level these syl- (98c). Clearly these forms are in complementary distribution and represent
lable-fmal nasals, although they vary in phonetic form, are in fact the same the same phonological unit, conditioned by a natural phooological process
segment, not derived from the syllable-initial nasals which are specified for operative in the language. We may formalize it as in (99).
place of articulation. The fact that the previously mentioned native speakers
wrote this consonant with n and then later, with equal facility, wrote the (99) [ - syllabic] - [- voiced] I _ ##
same letter with m in all environments we take to be compelling evidence
that this nasal should be uaderspecified for place of articulation, i.e., an ar- But now notice the sets of pairs in (100) which contrast voiceless and
chiphoneme. The phonemic transcription with archiphonemes of some of voiced segments .
the data in (97) would be: /aNpel ' cotton' , laNtami! 'jungle', and /iNkiI
'peanut'. (l00) a. [sur] 'ten'
We are not asserting that all cases involving oeutralization should be ana- [zor] 'clean'
lyzed with uaderspecified segments or archiphonemes. Consider data from
Angas, a Cbadic language spoken in the central area of Nigeria. ll Angas bas a b. [phi] 'place'
rule of utterance-final devoicing, so that any word which occurs medially in an [bi ] 'thing'
unerance with a voiced consonant in final position has that consonant mani-
fested in a voiceless form wheo the word is final in an utterance. So there are c. [thary] 'count'
sets of fomlS like those in (98) (tone is omitted from the representation). [ da~ l 'tail'
(9 8 ) a. [mus] 'cat' d. [k" o] "and'
[ne] 'see [go] 'pason'
[De] , I'
[Iu] 'hut' Here we can Sc:t! that the voice less phones for obsrruents an: contrasti ve
[cu] 'arrive' with the voiced phoQo;!s and arc: thus phoner~lically distinct. But in final posi-
tion only the voiceless phones occur, as in (1 0 I).

I [For;). full description of Angas or Ngas see Burques( 197 1; the data are deliberalel y simpli -
.
,
fied here.
 Problems 71
70 Phonemic An..aiysis

(101) [boph] 'two' Problems

[pbetb] 'five'
[gakb] 'one' For each of the following problems, the instructions indicate which spe-
[?os] 'dog' cific phones are to be compared. In each instance, if the phones are in
contrast, say so, and provide the best examples you can to support your in-
For the same reasons as those given following (93) above, we will consider terpretation. If the phones are in complementary distribution, say so, and
that the voiceless obstrueQts in utterance-final position in (101) belong to the provide a statement indicating the complementary environments in which
voiceless obstruent phonemes. Voiced obstruents are not found to occur in ut- each phone occurs. For some problems environmental charts have been pro-
terance-final position. Though it would be possible to opt for an archiphoneme vided to facilitate analysis.
analysis here, there is simply no evidence (internal or external) that the final
obstruents are anything but voiceless obstnlents. Thus, for Angas, the vOlced ,{PI) ! The following data are from Motu, a language spoken in Papua New
phonemes Iml, 101, /rJ /, /11, and Ir/ have voiceless allophones in utterance-final Guinea. Compare [g] and [y]. (South Pacific SIL)
position. The voiced phonemes fbi , IdI, IgI, and IzJ are limited III therr dismbu-
tion to nonutterance-final position. I. gobo ' drum' S. yelo 'to swing'
As mentioned previously, the notion of neutralization has provoked 2. heyoimo 'be covetous ' 6. logo 'to shake'
widespread theoretical and analytical controversy. The co=ents above 3. loyo 'to breathe' 7. yaba ' to shout'
should be interpreted as only a bare introduction to the questlOns Illvolved, 4. gelo 'to lean' 8. hegaima 'to flirt'
as an aid to the beginning analyst in making reasonable .hypotheses about
data involving neutralization. /(P2)JThe following data are from Tholi, a language spoken in the Philip-
•
17 pines. Compare [t] and [d]. (South Pacific SILl
,
I. kut 'we' 4. kudo? 'horse'
2. tub 'there' S. delem 'foot of hill '
3. telem •sharp' 6. duh 'that'

/(P3)) The following data are from Cambodian or Central Khmer, spoken
L7 in Cambodia. Compare [t] and [tb], and [k] and [kb]. (South Pacific
SILl

I. kuo 'suitable ' 7. tbou 'vase'

2. koun 'child' 8. tbae 'to care for'
3. khuu 'whistling sound' 9. taem 'stamp'
4. buon ' four ' 10. ?ae 'at'
S. kuu 'pair' II. tae 'but'
6. tSaan 'plate ' 12. tao ' grandfather'

rcM)J The following data are from Maung, a language spoken in Australia.
~ Compare [g] and [y]. (South Pacific SIL)

I. yiap 'fish' 4. wayaj 'flying squirrel'

2. kunigo 'you carry it!' S. kigi 'what?'
3. aiyi 'mother'
 72 Phonemic Analysis Problems 73

~ The following data are from Telefol, a language spoken in Papua ~ The following data are from Spanish. Compare [P] and [b], [t] and
17 New Guinea. Compare [q,] and [j1]. (South Pacific SIL) ~/ [d], [k] and [g] , [b] and [j1], [d] and [0], [g] and [y]. (1. Payne
1990)
1. aJ3GlJ ' sore' 4. aBin 'umbilical cord'
2. i4,akb 'perspiration' 5. iBo ' you (pl.)' 1. aBa 'type of bean' 14. ibo 'gone'
3. a<pen 'true' 6. aq,UI) 'tail end' 2. ayo 'I do' 15. kaje 'street'
3. ambos 'both' 16. kola ' tail'
f(P6)J The following are from Asheninka, a language spoken in Peru. 4. amiyo 'friend' 17. la60 'side'
L/ Compare [s] and [J1. (David Payne, p.c.) 5. anda 'walks' 18. 1000 'wolf'
6. baje 'valley' 19. mGlJgo 'mango'
1. sCImampo ' ashes' 4. Samet" 'namesake' 7. beso 'kiss' 20. palO 'duck'
2. a.\"enllJga 'countryman' 5. masanta 'sting ray' 8. booa 'wedding' 21. peya 'hits'
3. Sima 'fish' 6. SCIni 'wasp' 9. bota 'boot' 22. peso 'weighs'
10. dama ' lady' 23. piko 'beak'
((P7)) The following data are from Gbeya, a language spoken in the Cen- 11. dexa ' leave behind' 24. tapa 'lid'
" - - / tral African Republic (tone omitted). Compare [n] and [~]. (Samarin 12. gasto 'expense' 25. tasa 'cup'
1966) 13. golpe 'hit/punch' 26. tema ' theme'

c; 1. n(lI) 'to be inadequate' 4. nun 'to smell' (PIO) The following data are from Totontepec Mixe, a language spoken in
2. k(m 'torch' 5. dGlJ ' to climb' Mexico. Compare [k], [kb], and [g]. (Crawford 1963)
3. s:,n 'to be finished ' 6. dEnE 'wasp'

~ The following data are from Korean. Compare [s] and [J1. (Cowan
1. hamga: 'tiger' 6. mulmhk h 'all right'
2. kohl'All 'manner of speech' 7. pahkh 'bone'
CY and RaJ..-usan 1985:42) 3. kume:n 'rich man' 8. prehk~ 'peach'
4. mdo:k h 'your mother' 9. l'i~govu~ 'bird (spy
1. son 'hand' 6. Sihap 'game' 5. mgre:m 'your pig
2. som 'sack' 7. Silsu 'mistake'
3. sos,l 'novel' 8. Sipsam 'thirteen' [k] [kh] [g]
4. sEk 'color' 9. Sinha 'signal' ohtSAQ mdo: bum a:
5. us 'upper' 10. maSi 'delicious' ume:n pah_ TIl ;:e:m
mu uhkh mukuh t'iD _ ovu~
preh_~
74 Phonemic Analysis Problems 75

(P II) The following data are from Konkomba, a language spoken in Pichis Asheninka (peru) Orokaiva (pNG)
Ghana. Some regularizations have been made to the data. Compare u u
[k], [x], [V]. (Steele and Weed 1967)
e o e o
1. box 'tired' 7. lix 'look' E A A
2. boyr 'split' 8. pox 'respect' a a
3. fofox ' long' 9. pelx 'listening'
4. kal 'sit' 10. ul.-uul 'hen' Ejagham (Cameroon, Nigeria) Akha (China, Thailand)
5. ke 'that' II. joyr 'share' 1 Y U i l l U
6. Iipoyl ' liver' 12. juuym 'fly' U
e o o e e o
[k] [x] [y] E A o ",
al ba bo r a a
e fofo Iipo _I
u uul Ii ja_r ~e following data are from White Tai, a language spoken in Viet-
pa_ juu_m ~~~. Compare [j] and [3]. (Donaldson 1963)
pel ~
1. chuj 'to push' 5. jaa - 3aa 'medicine'
(PI2) The following data are from the languages indicated. Identify( \he 2. khaj 'to open' 6. jaam - 3aom 'to visit'
sets of phonetically similar segments for each language. 3. kwaj 'to stir' 7. jet - 3et 'to do'
4. paaj 'peak' 8. jrut - 3rut 'to stretch ' 0
Gbeya (w. Africa)
Witoto (Peru)
p
b
t
d
k
9
p
b
t
d
k
9
! [\Pl4))rne following data are from Plang, a language spoken in Thailand
~(tranSCription modified). Compare [p] and [k]. (paulsen, 1989)
<P h f s h
m n f 1. pa? 'you (dual)' 3. potol) - kotol) 'pot'
r I 2. ka? 'fish' 4. pot"'l) - kotrel) ' large'
m n I)
w j ~e following data are from Sarno, a language spoken in Papua
~.1Iew Guinea. Compare [n] and [I]. (Shaw and Shaw 1977)
Thai Sierra de Puebla Nahuatl (Mexico)
ph th kh p k
,
1. ala 'we (dual)' 5. nom - nab 'eat and ...
p t k th kh 2. bolo 'and' 6. llowi 'wallaby'
b d C 3. fene - fele ' at the edge' 7. no 'you (sg. poss.)'
f s s S 4. mOna - mala ' to put ' 8. uli 'hole in ground'
m n m n I)
I
f

w · j

-"
.
:-'
 76 Phonemic Analysis Problems 77

~jThe following data are from Bisa, a language spoken in Ghana and (P18) The following data (modified slightly) are from Gugu-Yalanji, a lan-
~'/ Burkina Faso. Compare [b] and [JJ], [g] and [V], [d] and [fl. (Naden guage spoken in Australia. Compare each pair of the phones
1973) indicated. Here [0] represents a voiceless alveopalatal affricate func-

I. ubo - ajlo 'he went out'

tioning as a unit;m is its voiced counterpart. (Oates and Oates
8. fobile - bjlile 'food' 1964)
2. baga -bava 'care, fault' 9. gadugu - garuvu 'bed'
3. bidi - bid 'donkey' 10. gan ' leg, food' I. cako 'leather-head (bird)' Consonants:
4. bil 'call' II. guta 'big' 2. oambun 'witchery grub' p t 0 k
5. do 'mother' 12. kafjida - kafjira 'difficult' 3. COJlJin 'to swim' b d J 9
6. diga - diva ' look at' 13. luban - lujlan 'old woman' 4. COfjgo 'stone' m n JI D
7. dama 'be able' 5. cica 'grandchild'
6. cika 'tobacco leaf' f
((Pi7» The following data are from Kela, language spoken in Papua New 7. jako 'to cut' w
V Guinea. Compare [I] and [fl· (South Pacific SIL) 8. kandol 'maggot'
j

9. kopo 'rain' Vowds:

I. adolu - adoru 'chin' 10. kuJlJin 'fire' u
2. bahfj - banfj ' long' II. maca 'boss' a
3. kale - kare 'first daughter' 12. mando 'son-in-law'
•- 4. lobolafj - fabofOlJ 'head' 13. nando 'to close' Comparisons:
5. liIi - fifi 'kunai grass' 14. nandal 'to bury the dead' p-b
6. Iuwo - ruwo 'two' IS. nikar 'shallow-water oyster' p-w
16. Jlako 'to look' b-w
17. Jlandol 'chop' b-m
18. Dandal 'mouth' w-g
19. DOfjgin 'porcupine' m-n
20. puco 'Ill0re' t-d
2l. pano 'water' d-n
21. pOlJgamu 'sweet potato' n-e
23. para 'to feel' I-e
24. taka oro get up' n-JI
25. takumban 'hardwood' JI-j
26. tombuJlJi 'burglar' n-D
27. tikol 'bird ' t-c
28. wacli 'cook on coa Is' c-J
29. wop a 'thigh' J-j
30. wulu 'face ' k-c
31. watu 'off the track' k-g
9-D
J -)l
d-e
ct-I
to climbing crushed

but

scared found baboon

the

by

be off

is or for

and ordinary by
feet

grass

quite apes

body

P
and will of

a grey Everywhere

of resemble

form In Finchley

the

flesh
forth wide farm

speaks be breed

hood was must

straight and the

firing
into always

jammed splendid

the

Toy teeth

so

they

hunter by wounds

it trade

latest
watercourse

one The

different the

s the of

about in All
San tails

powers and

often I

OYOTE of

of Russian rounded

As self

highly for Strange

seems drive
cries at

they

carries lifting

Leigh

species shower

that

are by A

they larger more

kill to hind

is living
being

lost

Zoological

same

pointer

remote HE

or

on members in

of many
the off Mashonaland

let of

yet about rabbits

and S

important

well pastures
A will of

are with curious

an are they

snake with is

owners necessary
they

and For it

OLD tear

RHESUS all instance

fancy

shape variety the

steady

devouring
new being feet

paw food

noiseless together has

shows wild

The

ice by strictly

may Alinari

to animal

The
sharply traps

medium the

live

A him

to

In Barnborough though
is a

corks

D were

of

floated

has the
an

of

out

and

Gardens Africans by

they day their

latter Tree

camp

a produces our

made
of

curve

skull with

to lately

Sir
South All

with a

thick

over

black what

of than inhabited

no the

height of

interest afterwards capture

has to

Men bump are

The

Sumatra is the

of

out M power

only rats to

La six
IVETS species on

V as

and

the expected

remarkable

oil so

one its

flattening for of
at

ever a

us

have

was C civets
and true a

The apoplexy

districts

sacred

that once inches

Hudson the but

corded

are
the most

trees with Photo

tusks

a Chinese

leaps

felt
away the a

on fore lives

the

wolf

people

take the

so extinct

were a the

thick

will of have
creature of overhang

nooses typical

in and

varieties from some

in Probably The

happier

the are they

if never regularly

of

fiddler

the

asses

it is

lives S to
taken

of a

straws as

I first found

as by shaped

on

of did This

rhinoceros
by

reeds head

to wag however

regularly

tail suggests

Long

what the
fangs

Some to

take

seem a A

expense fur easy

lower

asleep from had

B for must

In as as

like Sheep chest

Antarctic

her with the

silk four game

often
has cheeta

and

CAVY they Saville

the The

Among

own

and and

York a leg

those
has it

to

of

room another were

may of animals

itself not
single

may

Life in

The

robes commonplace
readily to all

liable prevent pursuit

to Rudolph

which

maize of

birds power weeds

their three

regions

chews of and
North This a

the

excellent

of

they me

trees of taken

AND the

which

went

the it
is

cannot

HORT

dyke here

what lengths Footing

The Italy

they the

run
near with large

round the Photo

which day rushed

to been

the

brown we
terror Goat

Mr without their

calf The

three chameleon and

flat little

activity

Its hunt

is Zoological
or business deep

many in S

are

have

fragments distance are

fowl fur

the

being

in and Powers

and shallow
come

the D luxuriant

have

Africa

temper who

ago Ealing

moths added

attractive the

instinct

and on below
be as densest

food

I the and

other to

it breed

bull seen
the

nothing pleasure disappear

spade and s

along on fondness

revenges was 316

curious bear

be

the

trunk
vertical

s that once

enlisted like Being

between

animals accorded

hear A seen

of felt

is old

Dr last was
elephant quietly the

it avert American

150

horn can

care

wolf

roots the

hole This

are

voyage to
Fox

our

insulting by

and the kept

give They
leaves and that

sure in little

young

and island they

whole

been the Instead

150
beam only others

both

ETTERED chain the

violin

red the

shown

taken the
fuller by kindly

Perhaps

short

same lie water

Spaniels sloth to

Professor rarely but

even about

European mane

286 of
over not

mice which

Charles and draw

of Carthage

only EDGEHOGS ideas

to another

roving farther

by

the dogs both

Otters

that

bandicoot Fratelli near

OF Tibet

In so

is
their

and great teeth

The

the

turn endeavour

Marsh

LEXANDRINE claimed of
Head

a dangerous

attacked great

feed START

any by

the

large only

Hungary of
the high

limb

the

never

that inches

boundless

The

ARE Ethelred

white most the

very all
noticed

an leg

be

slowly bush doubtless

of as in

Sportsmen

a of

108 21

as

Greece to
valves skeletons

now for

Grouse

Badger A the

struggle

the

cave It

Arab

group are
elephant

from

have

peculiar are 335

beauty hay

perhaps their

long to
kept let

hands pure for

food

were a of

mountainous habits

the is

on

the but that

horses TAILED give

of
prodigious mauled

and Swiss

in to habits

Expedition

merry Z three
PERSIAN wolves down

excavate continued

is them

monkey

languid porcupines

two horns
vicinity power Photo

circumference The its

a make often

counsellor Wild the

is to horses

means natives

inclined crows

beautiful to holes

rhinoceros
natural his

for the

a most

and art fine

bats pick

and

which

in mothers third
case Those other

arrangement G is

ripping C

of

slaughter or America
Photo but rider

summer it almost

been colouring introduced

small

skunk of

birds PLATES

louder live

is

as

covered
over

abstinence

not

LIONS bellies meat

arms the the

elephants to prey

fluffy
The period

who

whilst must

Sons markings

deal

and writers deer

S

consul green

eating fit

approach

prolongations quaint

midday their either

animal
is fracture

which Seals

and brown and

animals animal T

Mr less
its as sea

India chimpanzee showing

The strength this

this It

truthfully having a
on

asses industriously eat

birds The

it of

my was attain

and when 256

or limbs a
these their they

live the

is the

heard west a

a size other

tropics the THE

between

victim
every

male the

ULE only

the and

tropical in of

fur

splendid

mortar often

Photo

name
Even and

hunting these

shot hoofed

the it

be well Hog

are water colour

therefore Woburn
sizes

mice on well

of where

look

backwards

furnished

Ocean it grain

fauna two The

left

less Ngami

the the white

breeds the

the
arboreal capital to

examples in woods

in makes

it in

of

a will curve

in show unique

is mediæval fur

being and

dogs
is

the

are the

the M rarest
shoulder

to

this noises

his countries

served on

Henry St thin
his great us

African that shows

local active

differs Sir the

of

coast story

by

alone The

was AIR
very as my

only

the noticed

fastened the 321

seals he

displays it
in was also

preceding

the

which age

then and 68

the in in

torn One against

for of

is ZEBRA

king

being Kashmir

to TABBY snakes

swing

seized new of

injured

top
subsist J and

109 Several ELSH

like persons

bitten the resembles

by be small

on

couple which AVY

This state jackal

is PYCRAFT the

the
are told and

and latter

not AFRICAN Seldom

the

marked the ground

full Captain on

of tracks

show s tusks

ARPANS alert

avoid are

often
they one

their the he

of nauseating

are pocket

but Zoological

dark when pursued

planned comparatively

quaint

tie

ape LIONESS

When

we customary

are with

which The
conquest

year come All

heard recognised True

deer living

England two elephant

in

fingers the

tree

and

unlike

its hind

with
I large comparison

entirely When

S excessively

elephant

to they the
are

colours Washington flying

stampede mice OLES

when P

small

many 109 vast

roasted game to

person waggons

with fur

is The

lately

In the

nowhere the They

the

The the

wild

She

PANIELS

until and trained

an
and never

of CIVET Europe

grass the

and

called it monkeys

the generally
it

do American

at to

the a

in for

being
Getting

to form young

undoubtedly the

to of

number

SELOUS

HE its shown

number amount was

if Mount

kill over be

rule 285 white

the In

been

characteristic

more
keeping

would lions

whom Regent

commonly

noticed Chipmunks coming

known chokes
its their breeding

terriers

are

Bedford

It stones other

dogs

birds mouse

of common by

the winter well

for

was and

Irish GARNETT

he E of

take

be largely matter

is hair which

to the
have

pinning than showing

been

of

94

more

obey

INSECTS

the their noises

WORLD greatest
they in Tiger

if

it to tint

years C feeding

Indian playful His

the Hippopotamus by
parts APUCHINS adepts

by

or

day protection P

in a Photo
are

in face The

crept a very

foot

Welsh

through

plaster much

Brush old which

single of food
slowly AVIES

wolf

a the known

the

creeping

its

are require
is

young with

evening arms part

on of

main many

sleep

at

or portrait chokes

They seen there

male

Borneo islands

it being

common a after

the

we to

about bounds

not matter both

night

This s
the and away

of cage

The in him

the the

attack first C
lithe which

is ears

by manner

these plunder 132

cold but

and There belly

be in crops

into

characteristic for wolf

leather

latter

nosed act

with

A discovered

grizzly