Robert Robinson's alphabet and seventeenth-century English phonetics


Material Information

Robert Robinson's alphabet and seventeenth-century English phonetics
Physical Description:
viii, 154 leaves : ill. ; 28 cm.
Harrison, Thomas Carlton, 1943-
Publication Date:


Subjects / Keywords:
English language -- Phonetics -- Early modern, 1500-1700   ( lcsh )
English language -- Phonetics   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis--University of Florida.
Includes bibliographical references (leaves 146-153).
Statement of Responsibility:
by Thomas Carlton Harrison.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000085750
notis - AAK1103
oclc - 05349172
System ID:

Full Text








I should like to thank the members of my committee,

Dr. Jayne C. Harder, Dr. William J. Sullivan, Dr. Kevin

McCarthy, and Dr. William C. Childers, without whose aid

and encouragement this dissertation could not have been

completed. Thanks are due also to my wife and family, whose

patience and support made the task of writing much easier.




LIST OF TABLES .. . ... iv





Notes . . 47


Notes . .. 86


Notes . .. . 106


Notes . .. . 115

V ROBINSON'S ALPHABET ... .. ... .. 116

Notes . . 136

Robinson's Transcription of Barnfield's
Lady Pecunia . .. 138






Table Page

1 Old English Consonant Spellings ..... .13

2 Old English Vowel and Diphthong Spellings 14

3 Long Vowels in 10E and eME .. 17

4 Old French and Early Middle English Vowels 20

5 Middle English Vowel Spellings .. 35



1 The Scheme of Stratificational Phonology .

2 French Nasalization . .

3 The Realization of English Obstruent

Phonemes . . .

4 The Tactics of Syllable Onset Clusters .

5 The Tactics of Syllable Final Clusters .

6 The Tactics of Syllable Onset Clusters and

Syllable Final Clusters Combined .

7 Robinson's Alphabet . .

8 The Second Stanza of Lady Pecunia .



. 102

. 108

. 110

. 111

. 114

. 120

. 125

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Thomas Carlton Harrison

December 1978

Chairman: Jayne C. Harder
Major Department: Linguistics

The first phonetic alphabet for English, published by

Robert Robinson in 1617, is analyzed, along with the forces

which led to its invention. The English spelling system,

generally considered to be the most irregular and inconsistent

of any in the world, was reaching its present state in the

seventeenth century. Robinson gives the irregularity of

the English spelling system as his reason for inventing an

auxiliary alphabet to help foreigners and children learn to

read English.

English spelling appears to be irregular for two

reasons: first, it contains genuine irregularities as the

result of borrowing words from other languages with their

spellings intact, so that English words reflect the spelling

systems of several languages; and, second, insofar as English

spelling is regular, it corresponds not to a phonetic or

phonemic level, but to the more abstract morphophonemic, or


phonological level. The primary sources of English spelling

are Old English, French (from Modern French loan words as

well as from Old French or Anglo-Norman), and Latin. The

irregularity results from the translation of loan words

into English pronunciation with no accompanying change in

the spelling. Many apparent irregularities of English

spelling result from the fact that English spelling corres-

ponds to a deeper level of phonology than the pronunciation:

the use of the same vowel spelling in sane-sanity reflects

the morphophonemic similarity of the words. Generative and

stratificational phonology are similar in giving a consistent

representation of this level, although they differ in most

other respects.

In his alphabet and transcriptions of English, Robinson

suggests the notions of an abstract phoneme for English and

of phonemic voice. Although the phoneme has been rejected

by generative linguistics, it can be redefined without the

procedural bias of American structural linguistics so that

it is acceptable to post-Chomskyan linguistics. The more

abstract phonemic is in fact a necessary part of stratifi-

cational linguistics. Phonemic voice occurs in English in

obstruent clusters. The separation of phonemic voice from

the obstruent clusters is accompanied by neutralization of

the morphons (or morphophonemes) underlying the phonemes,

which results in the presence of archiphoenemes for the

obstruents of English.


Robinson actually uses aspiration to indicate devoicing

in his alphabet and transcriptions. He uses what is in

effect a devoicing phoneme at the same level of abstraction
and with the same accompanying neutralization as occur in

the analysis of English obstruents with phonemic voice.

Robinson's treatment of phonemic devoicing, or aspiration,

has problems involving the sonants involved in English consonant

clusters and consistency of application as devoicing; the

same symbol is used for devoicing and initial [u]. When he

errs, Robinson generally errs on the side of oversimplifica-

tion, using too few symbols rather than too many. His

apparent intention is to create an alphabet and system of

transcription that correspond to an abstract phonemic level

like that of stratificational or Prague school linguistics.

In terms of the relationship of alphabetic spelling

to phonological theory, the presence of a phonemic as well

as a morphophonemic level makes the proponents of the theory

less likely to declare a single orthographic principle than

the proponents of theories which have only one consistent






This is a dissertation in the history of linquistics.

It had its genesis in a combination of interests--history

of the English language, phonological theory, and the

relationship of sound and spelling in English. The early

Modern English period offers ample opportunity for all these

interests. In fact, it seems to combine them at every turn,

for the sixteenth and seventeenth centuries produced the

first spelling reformers and the first phoneticians in

English; phonetics was usually the outcome of proposals for

spelling reform. John Cheke, Thomas Smith, John Hart, and

William Bullokar combine to form a group of spelling

reformers covering the last half of the sixteenth century.

Robert Robinson, whose alphabet (1617) is the first

auxiliary phonetic alphabet for English (not intended to

replace standard orthography), is placed at the end of this

line by Dobson (1968), although Robinson is not a spelling

reformer. Robinson writes with apparent awareness of the

general feeling for spelling reform; his alphabet, he

says, is to make up for the defects of standard English


The impetus for spelling reform in English comes

from the reform in the pronunciation of classical and

Homeric Greek in the universities, usually associated with


the name of Erasmus. Greek had long been taught with the

pronunciation of contemporary Greek, but Erasmus and other

classical scholars, including John Cheke and Thomas Smith

at Cambridge, argued for a pronunciation nearer that indi-

cated by the spelling of Greek. In the conservative aca-

demic pronunciation, for example, the vowel letters I L, I,

and the digraphs oc and oL, were pronounced with the value

of Italian i. The Erasmian pronunciation postulated three

distinct vowels and two diphthongs (Cheke, 1555). The

chancellor of Cambridge, Stephen Gardiner, issued an edict

in 1542 forbidding the use of the new pronunciation by

teachers or students. There followed an exchange of letters

between Cheke and Smith, and Gardiner. Cheke finally left

England for the continent, and took his own letters and

Gardiner's (including the edict of 1542). He published

them all at Basel in 1555, under the title De pronuntiatione

linguae Graecae (Drerup, 1930: 93-95). Cheke and Smith are

exceptionally important because they figure prominently in

the controversy over Greek, and they began the tradition

of spelling reform in English.

Although Robinson nowhere mentions any predecessors

in phonetics, his alphabet itself suggests his knowledge of

Greek, and he apparently matriculated at Cambridge in 1615,

sixty years after Cheke's publication of his letters.

Robinson must then have been aware of the controversy and

must have seen the results of the ultimate success of the

Erasmian pronunciation of Greek (see Chapter V). Robinson

seems to fit at the end of a tradition which was at the

same time the first movement for English spelling reform

and the first English school of phonetics.

The English spelling system has the reputation of

being among the most inconsistent and unphonetic in the

world, and it evidently owes some of its present reputation

to early Modern English. In fact, the English spelling

system had reached its present state by the end of the

early Modern period. The questions present themselves:

how inconsistent is the English spelling system, and how

did it arrive at its present state of inconsistency? In

Chapter I, I shall attempt to answer these questions.

In this chapter, I shall trace the development of the

English spelling system from its beginnings in the Old

English period through the French influence on Middle

English and the Latinizing of the early Renaissance, the

influences that gave English the variety of alternate spell-

ings that it contained in early Modern English.

In Chapter II, I shall deal with the relationship

between alphabetic writing and the two dominant American

theories of phonology--generative and stratificational

phonology--along with their predecessors, American struc-

turalism and Prague school phonology. This chapter will

be a comparison of the two theories as they relate to

English orthography.

In Chapter III, I shall examine the theoretical

status of the phoneme in generative theory, which generally

rejects it, and stratificational theory, which accepts it.

The phoneme as defined by stratificational phonology is the

basis for Chapter IV, and I shall analyze the notion of

phonemic voice in English obstruent clusters.

Robinson's transcriptions indicate that he used the

term aspiration to indicate something like the modern

concept of phonemic voice in consonants and consonant

clusters. In Chapter V, I shall analyze Robinson's alpha-

bet and transcriptions; and in Chapter VI, I shall examine

Robinson's system of phonology, based on the evidence we


Finally, I shall give consideration to the effects

of an essentially descriptive phonology like that of

Robinson or of the stratification linguists on the notion

of the relationship between phonology and spelling.



In this chapter I shall examine the development of

English spelling and try to explain why the orthographic

system of English in the early Modern English period

prompted the first attempts at spelling reform in English.

There were clear causes for these attempts other than the

irregularity of the spelling, for example, the impetus for

reform arising from the controversy over the pronunciation

of classical Greek mentioned in the Introduction.

The early spelling reformers all considered the

ideal spelling system to be phonetic or phonemic. King

(1969: 213) comments that "spelling reformers seem to make

the best autonomous phonemicists." Although the notion

that spelling should reflect etymology was plainly being

put into practice during the fifteenth and sixteenth

centuries in the Latinization of the spelling of words

like nation (Middle English nacion), the notion does not

appear in the writings of orthoepists until the eighteenth

century (Elphinston, 1795). The earliest spelling

reformers in English, Sir John Cheke and Sir Thomas Smith,

however, being classical scholars, must have been aware that

Latinization was one of the contributing factors in the

irregularity of English spelling. This awareness was


undoubtedly one of the factors which motivated Cheke's

desire (Cheke, 1561) to avoid foreign words in writing


Considered as a phonetic representation, English

spelling has been irregular since the Middle English period,

but it has been irregular in two different senses. Middle

English spelling was more or less phonetic in the sense

that the same sound tended to be spelled in the same way

whenever it occurred, although there were often variant

spellings of the same sound, as in the spelling of the

word peace below. A given spelling generally indicated

the same sound wherever it occurred. In Modern English,

spelling has come to represent the same word with the same

letters whenever it occurs (see Jespersen, 1909, 1:3). In

Middle English pes, pees, and pais all represented the

same word, 'peace'; e, ee, and ai all represented long

open [e:]. In Modern English the influence of printing

and etymology have combined to change the system so that

a particular word is spelled the same way every time it

occurs, although a different word with the same sound might

be spelled differently. The vowels of wait and name have

been identical since the seventeenth century, but their

spellings have been frozen, so that they are always kept

separate orthographically.

A number of influences can be found in the makeup of

Middle and Modern English spelling. The foundation of

English spelling was, of course, laid in the Old English

period when the alphabet of the Roman missionaries was

combined with the Irish alphabet to represent Old English

phonology. Old English spelling seems to have been quite

regular, but the Norman Conquest brought with it an enor-

mous number of loan words, their French spellings intact,

which added variant spellings for many sounds of English

(for instance, the ai spelling for [e:] in pais, 'peace,'

above). French influence was responsible for a number of

changes in the spelling system of English, changes which

affected native words as well as loan words, such as the

use of u to represent the vowel [y] and sch, rather than

Old English sc, to represent [s]. Middle English spelling,

which T shall discuss in detail below, is a combination of

the Old English system with the Old French system.

After the complication of the English spelling system

by the influence of French, the sound-spelling relationship

was made more irregular during the fifteenth and sixteenth

centuries by the sound changes that mark the difference

between Middle and Modern English. The Great Vowel Shift

itself was a regular sound change and left consistent,

although different, sound-spelling correspondences. Many

of the changes, however, involved the consonants, such as

the loss of initial velar stops before n as in knight and

gnat, which left a residue in Modern English spelling.

Along with the vowel shift and other sound changes,

the tremendous number of loan words from Latin further com-

plicated the spelling of English. Not only were words

borrowed from Latin with the original spelling intact, but

many words already in the language which were originally

Latin were respelled according to Latin principles. The

b in debt was introduced apparently to reflect the Latin;

the word had been spelled more phonetically as dette in

Middle English.

It has sometimes been suggested that the irregulari-

ties of English spelling are the result of sound changes,

like the Great Vowel Shift, without accompanying changes

in spelling, but this explanation is not strictly true.

It is rather spelling change unrelated to sound change

that has made the system irregular. I shall return

presently to a detailed examination of the factors out-

lined above.

In any study involving the development of English

phonology, as this one is, the monuments of English philo-

logy, Luick's Historische Grammatik (1921), and Jordan's

Handbuch der mittel-englishen Grammatik (1934), are indis-

pensable. Many philological works of smaller scope are

valuable since, in examining the evidence of early English

sound change, they catalog and organize many samples of

spelling. Among these works are Neumann's (1904),

Lekebusch's (1906), and Kjerrstrom's (1946). Among the

histories of English, Wyld's History of colloquial English

(1936) must be singled out for the thoroughness of its

treatment of the fifteenth century, a crucial period for

the development of English spelling as well as English pro-


Few philologists or linguists have actually done more

than suggest the outline of the development of English

spelling. Jespersen (1909) sketches its development, but

he is clearly more interested in the development of English

phonology than of English spelling. William A. Craigie has

been concerned with spelling, however, rather than phonology.

His book on English spelling (1927) is primarily concerned

with the spelling of Modern English, but it suggests that

the origin and history of a word are indicated by its

spelling. Craigie's pamphlet, Some anomalies of English

spelling (1942), is concerned exclusively with the history

of English spelling, but, as its title suggests, it is incom-

plete. Craigie has also adopted an alphabetic organization,

listing each spelling and its sources. While this organiza-

tion makes clear the history of each spelling treated

(vowel, consonant, or digraph), it tends to obscure many of

the general relationships among the phonology, sound change,

and the spelling system.

More recently, Richard Venezky (1965) has examined

the background of Modern English spelling in his Ph.D.

dissertation. Venezky has organized his work alphabetically

by 'spelling units': letters of combinations of letters

(like th or sh) which are used as units. His treatment of

the development of the consonantal spelling units is

generally excellent, but the alphabetical organization, like

that of Craigie's work, seems to hinder a full treatment of

the vowel spellings, and he has little to say about the

occurrence of double consonant letters. Venezky is also

primarily interested in the spelling-to-sound correspondences

of present-day English, and these correspondences differ, if

only slightly, from those of early Modern English. Neverthe-

less, my debt to Venezky's work will become plain as I


Many other sources not concerned directly with ortho-

graphy--or, for that matter, necessarily with English--will

be cited in the course of this chapter. Of these, two are

of exceptional importance. Mary Serjeantson's (1936) study

of the loan words in English is invaluable, and Margaret K.

Pope's From Latin to Modern French (1934) contains one of the

few recent treatments of Anglo-Norman phonology.

I shall turn first to the spelling system of Old

English, which, after the sound changes between Old and

Middle English, formed the basis for Middle English spelling.

I shall examine the phonology and spelling of loan words

from Anglo-Norman and Old French along with the scribal

changes effected by French scribes. These elements are

combined in the spelling of Middle English. From Middle

English spelling I shall turn to the sound changes that

mark the distinction between Middle and Modern English.

These changes, along with the fixation of English spelling

word by word and the influence of classical Latin, formed

Modern English spelling much as we know it today.

The Old English alphabet was a combination of the

alphabet introduced by Roman missionaries and the Irish

alphabet (Jensen, 1969: 531). It had four letters not to

be found in the Modern English alphabet: consonant letters

thorn (7 ), eth ( ), and wynn (p), and the vowel letter

ash (ae). and both represented both [ ] and [ ];

p represented [w]; and [a] represented the same low front

vowel (as in cat) that it represents in modern phonetic

alphabets. The letters k and z did not occur in Old English

except in occasional loan words.

Some of the consonant letters represented sounds

different from those they represent in Modern English. c

and g represented the voiceless and voiced velar stops [k]

and [g], respectively, in conjunction with low front and

back front, as in candel, 'candle,' and gad, 'goad.' The

same letters represented the affricate [c] and the palatal

glide [j], respectively, in conjunction with mid and high
vowels and diphthongs, as in ceorl, 'churl,' and gea, 'yea.'

g also represented the voiced velar fricative [y] when it

occurred intervocalically in conjunction with back vowels,

as in dagas, 'days.' Doubled consonant letters indicated

consonant length, as in pyffan [pyf:an] 'puff,' settan

[set:an] 'set,' and spinnan [spin:an] 'spin.' Short frica-

tives were voiced intervocalically, as in heofon [heavon]

'heaven,' and aeqer [ca:jder] 'either.' Voiced fricatives,

then, occurred in Old English only as phonetic variants of

the voiceless fricatives.


In addition, sc represented [s], as in scip, 'ship,'

and cg represented [j], as in ecg, 'edge.'

The following is a table of the sound-spelling corres-

pondences of Old English adapted from Venezky (1965: 195).

Table 1

Old English Consonant Spellings





















Posited Phonetic Value






[k], [c]

[g], [j]


[s], [z]

[ts], [dz]

[ ] [d]

[f], [v]


[h], [1 ], [x]






The vowel letters of Old English were those of the

Latin alphabet--a, e, o, u--to which ae and y were added.

The digraph oe represented [0] which was formed from i-umlaut

of [o] and was the early unrounded to [e] (Campbell, 1959:

76-77). y represented the high front rounded vowel [y].

In addition to the simple vowels there were four diphthongs

in Old English, spelled ea, eo, io, and ie. The same

spellings represented both long and short vowels and

diphthongs.3 Below is a table of the Old English vowels

and their posited phonetic values:

Table 2

Old English Vowel and Diphthong Spellings


i [i], [i:]

y [y], [y:]
e [e], [e:]

aE [ae] [aa:]

a [a], [a:]

o [o], [o:]

u [u], [u:]


ea [mea], [ae:a]

eo [eo], [e:o]

io [io], [i:o]

ie [ia], [i: ]

As suggested above, the same vowel and diphthong

spellings represented long and short vowels and diphthongs;

the vowels in biddan 'bid,' and bidan 'bide,' as in Latin,
were not distinguished in the original orthography.

The sound changes which separate Old English from

Middle English began in the late Old English period. The

changes important for this study begin with the Old English

lengthening and shortening of vowels and diphthbngs before

certain consonant clusters, and the later monophthongiza-

tion of the Old English diphthongs. Important changes in

early Middle English are vowel lengthening in open syllables

and the loss of contrastive consonant length. Other signi-

ficant changes are the rise of Middle English diphthongs

and the reduction of unstressed vowels in inflectional

endings. This reduction began earlier but was not reflected

directly in the spelling until the Middle English period.

The significance of these changes is that they make the

spelling system and phonological system of Old English

into those systems which existed when Norman French began

its influence on English early in the Middle English period.

Old English short vowels and diphthongs were lengthened

before a liquid or nasal plus a homorganic stop: [ld],

cTld 'child'; [rd], heard 'hard'; [mb], climban 'climb';

[nd], bindan 'bind'; [Ijg], lang 'long.' Short vowels and

diphthongs were also lengthened before [r] plus [1], [n],

or [d] : [rl], eorl 'earl'; [rn], bearn 'child'; [rd],

eorSe 'earth.' A third consonant following the lengthening

so that child 'child,' had a long vowel, while cildru

'children,' did not (Jordan, 1934: 39-40).

The Old English long and short diphthongs were

leveled to long and short vowels in late Old English. ea

was leveled to [e:] in stream 'stream'; eo was leveled to
5 -
[2:]in deor 'dear'; le and its short equivalent occurred

only in West Saxon (Campbell, 1959: 68, 107, 126-128);

their history is not significant for the development of

Modern Standard English. The short diphthongs were

leveled to equivalent short vowels: ea and eo became [e]

and [0]. The io and eo spellings had become confused

with each other early in the Old English period, most

likely because the sounds they represented had been


In the Old English period the reduction of unstressed

vowels had also begun, especially in suffixes. Front

vowels [i], [e], and [a] all came to be written e fairly

early in the Old English period. In late Old English a,

o, and u came to be used interchangeably for unstressed

back vowels. By the eleventh century the distinction

between front and back unstressed vowels was being lost,

so that hlefdigen, minas could be written for hlaefdigan,

mines (Campbell: 153-157).

The long vowels showed changes between late Old

English and early Middle English. The following table of

correspondences is from Jordan (1934: 68-78).

Table 3

Long Vowels in 10E and eME

10E ( eME

[i:] [i:]

[e:] [e:]

[1 :] [#:], [e:]

[a : ] [e:]
[a-:] le:]

[a:] [o:]

[o:] [o:]

[u:] [u:]

[y:] [y:J, [i:]

The high front and mid front rounded vowels [y(:)] and

[((:)] remained in Middle English, especially in the southern

dialects, for some time. [0:] was eventually unrounded to

[e:]. [y:] remained at least in some dialects throughout

the Middle English period; it may in fact have survived into

Modern English (Melchior, 1972; see also note 14 of this


These vowel changes altered the relationship between

long and short vowels. In Old English, as mentioned above,

there were long and short versions of each vowel. After the

changes in early Middle English, this was no longer true.

There were two long mid front vowels, [e:] and [e:], as

compared to one short vowel [e]; there were likewise two long

mid back vowels compared to one short vowel [o]. Old English

[a] came to be written a in Middle English, but there is

some doubt whether it was ever centralized, as the spelling


Two Middle English sound changes which left a permanent

stamp on the spelling tactics of English were vowel lengthen-

ing in open syllables and the loss of contrastive consonant

length in the thirteenth century. These changes left

English with the pervasive--but not entirely consistent--

spelling patterns of double consonant letters as signals of

preceding short vowels, and of 'mute e' as a signal of a

preceding long vowel.

Doubled consonant letters in Old English words like

settan, middel, spillan, siJan 'after,' or missan represented

phonetically-long consonants. Single consonant letters in

words like witan 'know,' and a9 eling 'noble' represented

phonetically-short consonants. Long consonants occurred only

intervocalically and only after short vowels. Although

doubled consonant letters occurred in final position, they do

not seem to have represented final long consonants.

In the thirteenth century short stressed vowels were

lengthened in open syllables (before single short consonants

followed by a vowel) in words like nama 'name.' Long conso-

nants blocked open syllable lengthening, so that words like

settan retained short vowels in Middle English. When
consonant length was lost, also in the thirteenth century,
consonant length was lost, also in the thirteenth century,

the spelling with doubled consonant letters was generally

retained, leaving doubled consonant letters as a signal of

a preceding short vowel.

Later in Middle English, vowels in inflectional endings

were reduced to schwa, and the spelling of those vowels was

changed to e, so that OE nama became ME name (pronounced

[na:ma]). In late Middle English, when final schwa was

dropped, final 'mute e' remained as a signal of a preceding

long vowel.

Another effect of the loss of consonant length was

the rise of voiced fricatives in English. In Old English

contrastive voicing did not occur in the fricatives. Voice-

less [f], [e], and [s] occurred initially and finally; long

fricatives were also voiceless. Intervocalic short frica-

tives were voiced in words like heofon 'heaven,' aeleling

'noble,' and bosm 'bosom.' Long fricatives were always

voiceless. When contrastive length was lost in the medial

consonants, short voiceless fricatives remained in words like

missan and siL7an, and these consonants were in contrast to

the short voiced fricatives in words like heofon (Kurath,


After the Norman Conquest the influence of French loan

words and French scribal tradition had such a pervasive

influence on English spelling that the orthography of

English from late Middle English on can be considered a

combination of French and native English spelling.

The orthographies of both French and English were, of

course, based on Latin orthography. In addition, both lan-

guages were phonetically similar in many respects. The

similarities and dissimilarities between the vowels and

diphthongs are especially significant. The following vowel

charts for Old French and early Middle English are based on

those of Pope (1934: 433, 435).9

Table 4

Old French and Early Middle English Vowels

Old French Vowels

i y u

e o
e o


Diphthongs: aj, ej, oj, uj, yj, we, aw, ow, ew
Triphthongs: ieu, eau

Early Middle English Vowels

Long Short

i y: u: i y u

e: o: e 5 o

e: o: a:



Diphthongs: aj, ej, aw, ow, ew, iw

The differences between the consonant systems of Old

French and Middle English were not in fact significant for

the spelling. French had the alveolar affricates [ts] and

[dz] (spelled with the letters c and z in cent and treze)

which were lacking in Middle English. These sounds changed

to [s] and [z], respectively, on the continent during the

thirteenth century, somewhat earlier in Anglo-Norman (Pope,

1934: 93, 450). In addition, French had a palatal nasal [n]

and lateral [ A ] (in digne and vaillant); the nasal survives

into Modern French, but the lateral changed to a palatal

glide [j] during the Middle French period on the continent

(Pope: 274). In Anglo-Norman [n] and [\] changed to [n] and

[1], respectively, preceded by a palatal glide in words like

desdeign and soleil, apparently in conformity with the

Middle English consonant system (Pope: 450).

Among the vowels, the greatest difference was in the

length distinction of English, which did not occur in French.

French, on the other hand, had many diphthongs and triphthongs

which did not occur in English. The diphthongs of Middle

English--[aj], [ej], [ow], [ew], and [iw] in day, eye, flowen,

newe, and stiward--were falling diphthongs; that is, they had

offglides rather than onglides like the rising diphthong in

French oui. The diphthongs of Old French included all the

diphthongs of Middle English except the [iw] of stiward.10

In addition, French had the falling diphthongs [oj], [uj],

and [yj] in ioie, puint 'point,' and fruit. French also had

rising diphthongs [qi] in cuiver 'quiver' and [je] in chef


Anglo-Norman brought with it a number of triphthongs,

and more arose in the course of its development in England

(Pope: 445-446). Of the triphthongs, only two seem to have

been borrowed into English--[ieu] and [eau], as in lieu and

beauty or quarreau 'quarry.' These triphthongs may have

actually been reduced to [ew] by the time they were borrowed

(Pope: 446).12

Of course, English phonology differed from Anglo-Norman

in other ways than the phonetic segments. Perhaps the most

important difference was in the placement and strength of

word stress. Middle English stress was much stronger than

that of Old French. The result of this stress in Anglo-

Norman was the acceleration of the reduction of unstressed

syllables. Homophonous countertonic vowels, as in graanter,

coalesced with the tonic vowel to yield granter (Pope: 437-439).

English stress was morphological, falling on the stem

syllable of the word, whereas in French stress tended to fall

on the penultimate syllable, regardless of the placement of

the stem. In loan words, therefore, English tended to shift

the stress forward to the stem syllable when it fell on some

other syllable in French. Thus French pite and envie tended

to become English pite and envie (Serjeantson, 1936: 295).

In those cases where the phonetic elements of French

and English corresponded closely, as in the open and close mid

vowels [e], [e], [o] and [o], little or no change had to be

made for loan words from French to be absorbed into English;

prechen 'to preach,' for example, could be taken into English

with little phonetic change.3 Most loan words, however,

underwent some change when they were borrowed into English.

Serjeantson (1936: 295-300) provides a summary of

the changes that took place before the during the transfer

of loan words from Anglo-Norman into Middle English. Unless

otherwise noted below, it is from her appendix on the

phonology of French loan words that I have drawn my informa-

tion on these changes.

The low and mid vowels of Anglo-Norman generally

follow the same rules of lengthening as English (see note 13).

They are long in open syllables under stress or before final

consonants (dame, debate, apelen, robe). They are short in

closed syllables (part, lettre, cofre). Old French [a]

was diphthongized to [aw] before nasals and was borrowed

as [aw] into Middle English, as in chance; Anglo-Norman [aw]

was monophthongized to [a:] in loan words before [mb] or [nd]

(strange, chambre). [e] and [o] are long before final

consonants (bek 'beak,' clQs 'close'). Anglo-Norman [e]

was very tense, sometimes appearing as [i:] in Middle

English--Middle English has both free and frire 'friar.'

[i] is long in open syllables when the stress remains

on the same syllable, as in bible. If the stress is shifted,

the vowel remains short (pite 'pity,' diner 'dinner'; see

note 13). When the stress remains the same, [u] may be short

or long in closed syllables (scurge, scourge); it is usually

long in open syllables (route), before a single final

consonant (bout) or before a nasal plus a consonant (count).

When the accent is shifted, [u] is short (super 'supper').

However, [u] is sometimes short in open syllables when

followed by a consonant plus 1 doublee 'double,' couple


[y], when the accent remained the same, was short in

closed syllables (juggen 'judge'); it was long in open

syllables and before a single final consonant (use and duc

'duke'). The history of this sound is obscure and hotly
contested by scholars.4 It seems to have been borrowed, or

to have changed early in English, to [iw], which fell

together with [ew], perhaps as early as 1300. The resulting

sound develops normally into current English [ju:], as in

few and duke, or loses the onglide, [u:] in chute (Dobson,

1968, II: 711-712).

The nasal vowels of French were denasalized in English.

Since Anglo-Norman retained the Old French distinction

between Ca] and ['], these sounds were borrowed with distinct

vowels and following nasals, as in emperor. [a] followed by

a nasal plus a consonant was diphthongized in Anglo-Norman

to rw]. The spelling aun appears early in the thirteenth

century in words like enchauntement (Pope, 1934: 442). The

vowels in these words were denasalized when they were

borrowed into English, but otherwise the words were unchanged.

[aw] before [mb] and [nd] was monophthongized in Middle

English to [a:] in words like chambre and strange, as

mentioned above.

The other French diphthongs likewise had a complex

history in Anglo-Norman and in their adoption into English.

[aj] fell together with [ej] in the eleventh century, and

[ej] was leveled to [e] in the late twelfth and early

thirteenth centuries. The nasal diphthongs followed the

same course of development as the nonnasal ones, falling

together first as [6j], then being leveled to [e], so that

ain, ein, and en represented the same sound in Anglo-Norman,

and defense, mains, and meins were all rhymes (Pope: 444).

Serjeantson (298) adds that [aj] and [ej] remained as

diphthongs longest before [1], [m], and [n], and when final.

The offglide remained before vowels, so that English

borrowed grain, paien, obeien, but pes 'peace' and rec t

'receipt,' which originally had [aj] and [ej], respectively.

A new [aj] had also developed before the conquest from French

[a] before [lj] or [nj], the glide fronting and consonant

and then disappearing, leaving words like tailor and Spaine.

The continental development of [ej] to [oj] was responsible

for later loans like esploit and royal.

The diphthong [oj] was borrowed directly into Middle

English (noise, ioie). Anglo-Norman [uj], from Old French

[oj], comes into Middle English sometimes as ui, sometimes as

oi. The oi spelling, as Serjeantson points out (298), may

represent the [uj] diphthong, since o was often used in

Middle English to represent short [u] (compare sun-son).

Variants of puint 'point,' and builen 'boil' may be spelling

variants, or they may be from continental French, where [oj]

fell together with [oj], so that the normal spelling would

be point, boilen.

The spelling ui could also represent two other

diphthongs in French. It could represent the falling diph-

thong [yj], which occurred in fruit; this diphthong ordinarily

was leveled to [y], and followed the course of development

of that vowel. After [k] it became a rising diphthong [4i]

and, in English as well as continental French, developed into

[wi], as in quiver and squirrel (from Old French cuiver,


Three reflexes occurred in Anglo-Norman for Old French

[ue]. It could remain a falling diphthong and be leveled

to [u] as in buf (OF buef) 'beef'; it could change to a rising

diphthong [we]; or it could change to [0] (Pope: 443-444).

The last reflex was the one borrowed into Middle English,

falling together with Middle English [s] (Serjeantson: 299).

It was spelled ue, eo, oe, and e, in words like people and

boef 'beef.'

The rising diphthong [je] was leveled to [e] in Anglo-

Norman during the latter part of the twelfth century. It was

borrowed into English as [e] in words like chef 'chief,' and

pece 'piece' (see note 11).

Additional diphthongs were formed from the vowels [a],

[e], and [o] when they preceded Old French [A]. [A]

changed into an offglide, forming the diphthongs [aw], [ew],

and [ow] (Pope: 446). These diphthongs were borrowed into

Middle English in faut, souden, peutre, but before labial

consonants they were monophthongized to long vowels, as in
saf 'safe.'

One triphthong [eau] comes from the same course as the

diphthong [ew] above, that is, from [e] plus Old French [A].

In Middle English loan words it fell together with [ew]

and followed the same course of development in beauty (also

spelled beuty).

A number of consonants differed in Anglo-Norman and

the Central French dialect attaining dominance on the

continent. English borrowed heavily from both dialects,

occasionally even borrowing the same word twice, once from

each dialect. This situation is reflected in the pair

warranty (Anglo-Norman) and guarantee (Central French).

Anglo-Norman initial [w] corresponded to Central

French [g] in Germanic loan words (from Germanic [w]),

giving Middle English were 'war,' waste, warisoun, as well

as gile 'guile,' gerdoun guerdonn,' and garisoun. Anglo-

Norman [k] before [a] corresponded to Central French [C],

as in casten 'chasten,' cacchen, catel, compared to

chasten, chacen, chatel. Anglo-Norman [g] corresponded to

Central French [ts], which was being simplified to [s].

Examples from Anglo-Norman are chisel, cacchen 'catch,'

and winch, as opposed to Central French chace, cite 'city,'

and wince. Finally, Anglo-Norman [s] corresponded to

Central French [s], as in norishe, anguishe, compared to


The consonants e ] and [ ] remained longer in Anglo-

Norman than in continental French, perhaps, as Pope suggests

(431-432), because of the presence of these sounds in English.

[e] and [(] developed from [t] and [d], respectively, which

became continuants intervocalically and finally after a

vowel. [e] occurs in a few early loan words like cariteP

'charity' (later carite), and plentep (later plenty). [S]

also developed from [z] before a voiced consonant and from

[s] before [f], as in mecler and eTfreuer; this fricative

was hardened into [d] in Middle English medle (from medler)

and medlar (from mealier) (Pope: 448-449).

Initial [h] does not occur in early loans from French,

except where influenced by classical Latin spelling, as in

hour. But in loan words originally from Germanic, like

haste, heraud, and hardy, initial [h] was retained and

borrowed into English.

Finally, Old French final [n] became [m] be dissimila-

tion after front consonants [1], [r], [n], [s], giving

venim, ransum, and pelegrim (from French venin, ransum,


The leveling of diphthongs in Anglo-Norman led to a

great deal of confusion in the spelling of Anglo-Norman

itself and in the spelling of loan words in English. Anglo-

Norman [aj] and [ej] were both leveled to [e], for which

English ea (from Old English ea, whose Middle English reflex

was [e:]--Jordan, 1934: 97) and Latin ae could also be used.

In the end, five spellings were possible for [fet]--fait,

feit, fet, feat, and fact. To aggravate this situation,

after the thirteenth century continental French [oj] was

changing by way of [we] to [e], adding oi and oe to the

number of possible spellings for {e]. In addition, ee

came to be used to represent final fe], to distinguish it

from [a], in words like donnee and anee, and this doubling

was used from the mid-thirteenth century on to represent

[e] in stressed syllables, as in pees (for pais); Pre-

consonantal s, which had been dropped in pronunciation,

also came to represent 'lengthened fe],' as in fest for Jfet]

(Pope: 460). Finally, then, Ivet] (from videt and vadit)

could be spelled uet, uait, uoit, uoet, uest, or ueet, as

well as ueat or uaet (Pope: 458).

In the transfer of loan words from French to English

it was the French words which accommodated themselves to

the phonology of English. The palatal consonants of French,

Ik] and In], for instance, were changed in English to the

similar alveolar 11] and In], as in valliant and sign.

Frequently there were nearly exact phonetic equivalents

between French and English sounds, like the parallel between

the French vowels and the Middle English long vowels (Table

4 above). The transfer of loan words from French to English

was accomplished by a process of phonological translation,

which replaced a phonetic segment in a French word (like

French Il] in signe) with the English segment most phoneti-

cally similar (In] in English sign).16 Since the loan words

brought with them new spellings, which they retained after

their anglicization, they added new spellings for English
sounds, such as ai for [e:], as in pais 'peace.'

In addition to new spellings brought into English with

loan words, Anglo-Norman scribes wrought a number of other

changes in spelling which were not related to the phonology

of loan words. A peculiarity of Old French orthography was

an uncertainty whether to represent Old French [o] with u
or o, an uncertainty which affected the Middle English

spelling of short [u] (in sun and son, for instance) but

not long [u:]. The spelling of French [o], vacillating

between o and u, as in tor and tur (Modern tour), was resolved

in the thirteenth century on the continent (Beaulieux, 1927:

175), somewhat later in Anglo-Norman, as ou. The ou spelling

originated from the leveling of the [ow] diphthongs in mout

and escoute to [u] (Pope: 278). The ou spelling was generally

adopted for English [u:], as in house and out.

When [u] was regularly spelled ou, the letter u was

more or less exclusively to represent [y], as it does in

Modern French. This spelling was borrowed along with the [y]

sound in words like use and duc. French [y] then fell together

with Middle English [iw], which developed ultimately into

Modern English [ju:] in use and duke (see above and note 14),

giving u as an alternative spelling to Middle English eu, ew,

and iw, as in few and stiward.

The change of [ts] to [s] in Anglo-Norman was responsible

for the use of c with the value [s]. In early Old French [ts]

could be represented by either ti or ci, but c (before i or e)

later became the usual representation.

The French use of h to suggest modification of the pro-

nunciation of a letter was partly responsible for the number

of digraphs consisting of a letter plus h in English. h was

attached to a letter to indicate that its pronunciation was

'not what would normally be expected under the conditions in

which it stood' (Pope: 177); for instance, ch (which occasion-

ally represented [k], as in Latin Christus) generally repre-

sented [ts], as in Charles or cheval. Old English used th
in the earliest manuscripts to represent [e] and [d],1 and h

was used to represent the velar fricative [x], so that there

was some precedent in Old English for the use of h in wh for
f\ 20
[w], gh for [x], and s(c)h for [s]. The wh spelling for

[w] is, of course, simply reversed from the Old English spells

ing hw, as in hwaet 'what.'

I cannot give more than a general outline of Middle

English spelling, since, although the subject has been fre-

quently discussed, it has never received more than cursory
treatment. I do not have access to the materials for a

thorough study, so I must rely largely on secondary sources.

The following sketch of Middle English spelling comes partly

from my own investigation, but it also depends heavily on

Mosse's (1952) discussion of Middle English spelling, on

Robinson's (1961) summary of spelling and pronunciation in

Chaucer, and especially on Venezky's chapter on the develop-

ment of English orthography (1965: 191-214). Although

Venezky's chapter misses some important generalizations in

the development of English spelling, it is excellent in its

treatment of specific Anglo-Norman scribal changes and the

later changes brought about by the etymologizing of the


French influence was pervasive in the spelling of

Middle English vowels, although native developments, like

doubling of long vowels, also left their stamp on the spell-

ing. After the French ou spelling for [u:] was introduced

(although u was still used to represent short [u]--see

below), y then became an alternate spelling for [i(:)] in

words like bydden 'bid,' or byden 'bide.' y was also used

to represent the palatal glide [j] in words like yelpe


u and o were interchangeable spellings for either [u]

or [o], partly because of the French habit of not distin-

guishing the spellings, and partly because of the handwriting

of the scribes. o was frequently substituted for u,

especially in the vicinity of u, m, and n, to avoid a suc-

cession of minims, or downstrokes, which made the letters of

words like some and son difficult to distinguish when spelled

with u (Venezky, 1965: 202).

Old English ae disappeared early and was replaced by

ea, a, or e in words like appel.

Later introductions from French were ei and ie, repre-

senting [e:] and [e:], respectively, in receive and chief.

ei did not become common until the fourteenth century, and

ie not until the fifteenth century.

The reduced vowel [a] in inflectional endings was

represented by e, as in biden 'bide.' When final schwa

disappeared in late Middle English the 'mute e' spelling

remained as a signal of a preceding long vowel.

A development in spelling native to English was the

doubling of vowel letters to represent long vowels,

especially the mid vowels. ee and oo were used frequently

to represent both the open and closed mid vowels. In

sweete and heeth 'heath,' ee represents [e:] and [e:],

respectively. In good and goot 'goat,' oo represents [o:]

and [o:]. The ea and oa spellings which represent [e:] and

[o:], respectively, occurred in the Middle English period

(ea was., of course, an Old English spelling) but were not

put into general use until the fifteenth and sixteenth


The diphthongs of Middle English arise from a number

of sources. [oj] and [uj] seem to have come into English

from French in words like joie and puint 'point.' It is

difficult to tell to what extent the oi (oy) and ui (uy)

spellings represented different sounds, however, because of

the general confusion between u and o spellings.

Early Middle English [aj] and [ej] merged early,

yielding [aj]. The general spelling was ai (ay), so that

the spelling of sail (OE segl) and way (OE weg) reflected

the same vowel as day (OE daeg). A few words like they

and eight preserve the e spelling.

The [ow], [ew], [aw], and [iw] diphthongs were spelled

more or less phonetically in arowen, knew, cause, and stiward.

[iw] also developed from Anglo-Norman [y:], so that u was a

possible spelling, as in vertu. The offglide of [ow] was

frequently not represented in the spelling when it occurred

before [x], as in fo(u)ghte or tho(u)ght.

The following chart, adapted from Robinson (1961: xxxi),

summarizes the spelling of the vowels and diphthongs of

fourteenth-century English.

Table 5

Middle English Vowel Spellings























Spelling'- Short Vowels

i, y





u, o

Spelling Long Vowels

i, y

e, ee (eo)

e, ee (ea, ei, ai)

a, aa

o, 00

0, 00

ou, ow, ogh


u, iu, iw

eu, ew

au, aw


ai, ay, ei, ey

oi, oy


this, thyng


yonge, sonne

can, that

oft, lot

but, yonge


shires, ryden



name, caas

holy, rood (vb.)

good, bote

fowles, droghte


vertu, stiward

knew, lewed

cause, draughte

growen, sowle

sayle, day, they

coy, joie

The above table assumes that Old English [y(:)] had

disappeared in late Middle English. Ordinarily OE [y:]

became ME [i:], as in fir 'fire,' or pride. OE [y] became

[i] as in synne 'sin.' This merger also contributed to

the interchangeable use of i and y in Middle English.

Anglo-Norman and Old French [y], on the other hand, was

generally borrowed as [u], and OF [y:] generally came into

English as [iw]--responsible for the u spelling in the

chart. However, [y(:)] did not drop out of all varieties

of English until much later; there is some evidence that

it survived at least into the late seventeenth century (see

note 14).

A number of changes in the spelling of the consonants

of Middle English were made by French scribes. These

changes did not reflect changes in the phonology of the

language, but only in its appearance. My source, unless

otherwise noted, is Venezky (1965: 201-204).

The remaining letters of the runic alphabet, wynn (P),

thorn (P), and eth (C), all disappeared. Wynn was replaced

by w. Eth was replaced by thorn and th, and thorn continued

in use until the introduction of printing. Since the early

type was made in France, there was no letter to represent

English [9] or [T], and th eventually came to be used

exclusively for both sounds.

Yogh (3) represented a number of related but different

sounds in Middle English. It was replaced by g when it

represented [g], as in gras 'grass.' It was replaced by

gh when it represented [x], and by y when it represented [j],

as in knight and yelp, respectively, g was also used to

represent [ ], as in hegge [heza] 'hedge,' and initial [ ]

in French loans like gentile 'gentle.' Initial [3] was also

represented by i (or j, which was a positional variant of

i--OED) in words like ioie 'joy.'

In keeping with the French habit of using h to indi-

cate the changed value of a letter (above), [c] came to

be represented by ch, as in cherle 'churl,' as well as in

French words like the name Charles. [s"] (OE sc, as in scip

'ship') was variously spelled sch, ssh, or sh, as in schal

'shall,' flessh 'flesh,' or fish, eventually settling down

to the modern sh spelling. The order of the Old English hw

sequence was reversed everywhere to wh, as in what (OE hwat).

c in its double value as [k] and [s] was introduced

through French in words like certayne 'certain.' Since c

was used to represent [s] before e and i, k was used to repre-

sent [k] in those positions, as well as before n and 1, as

in kepe, knaue, and kloke 'cloak' (Mosse,1952; 9).

In late Middle English, ph began to be substituted for

f in learned words of Greek origin, like phleume 'phlegm,'

where it was a transliteration of Greek phi.

The doubled consonant letters left from Old English

long consonants took on their role in Middle English as

signals of preceding short stressed vowels in words like

beggar, cribbe 'crib,' dokke 'dock' (vb.), and bicche 'bitch.'

The doubling of consonants was not consistently carried out

in the spelling, however, so that there are doublets like

cache-cacche 'catch,' and comen-commen 'come.'

The fifteenth century marked the transition from

Middle to Modern English. Sound changes like the Great

Vowel Shift were in progress in the fifteenth century, along

with consonant loss and cluster simplification which left

a residue in the spelling of English. The stress system of

English was also changing at this time. In addition to the

sound changes of the fifteenth century, there were other

changes which affected the development of English spelling,

directly or indirectly.

Although Middle English spelling had been subject to

considerable variation, that variation was consistent in

that a certain small number of spellings were used for each

sound (such as e, ee, and rarely ea, ei, ai for [e:]). Varia-

tions in fifteenth-century spelling became inconsistent, or

less consistent, than that of Middle English, for a number

of reasons.

In early fifteenth-century sources like the London

Chronicles, the spelling is very much like that of Chaucer,

but later in the century, in sources like the records of
Parliament and the Paston Letters,2 more variants began to

occur, which finally formed a spelling system very near that

of Modern English. One reason for the instability in

fifteenth-century spelling was an increase in literacy,

which, by adding a large number of people who could read and

write, caused a breakdown in the scribal tradition (Wyld,

1936: 63).

Another cause of the instability was undoubtedly the

influence of regional dialects. However, the scarcity of

material on Middle English dialects makes it difficult to

assess the influence of the dialects on the spelling.23

We know that the population of London was constantly being

recruited from all over England (Chambers and Daunt, 1931:

237), and that people of high rank (like Margaret Paston of

the Paston Letters, who is mentioned above) usually employed

secretaries to whom they dictated their correspondence

(Kihlbom, 1926: xiii).

The greatest variation in the spelling occurred in

the representation of post-tonic unstressed syllables,

where the reduction of the vowels made all the vowel

spellings meaningless. The vowel letters in the final

syllables of taken, lynyn 'linen,' and happen 'happen,' all

represented a pronunciation which was probably virtually

identical with the pronunciation of those syllables today.

In the inflectional endings-- -ed (past participle),

-eth (third person singular), -es (second person singular

and noun plural)--the most frequent variation was -id, -ith,

and -is (or the equivalent -yd, -yth, -ys), as in dwellyd,

semyth 'seemeth,' or horsis. More rarely u occurred, as in

clepud 'clept.' In -en and -en plus a consonant, o and u

occur, as well as i (y), in y-writon 'written,' gotun 'gotten,'

and gravyn 'graven.'

There was a tendency to represent all unstressed vowels

with i or y, on the one hand, or a, o, or u, on the other.

It may be that the i spellingsrepresent [i] pronunciations,

while the a, o, and u spellings represent [a], as in the two

modern pronunciations of stomach [stAmpk] and [stAmik]

(Wyld, 1936: 258-282).

Some of the variations are due to varying stress on

the words concerned--certin from ME certain, certayne from

ME certain (Wyld: 259).

Another occasional variation which apparently repre-

sented a variant pronunciation was e where [i] was expected.

This alternation in words like wreten 'written,' and drevyn

'driven,' seems to be involved with the lengthening of [i]

to [e:] in Middle English.24

While confusion over the spelling of unstressed vowels

was making itself evident, the spelling of vowels in stressed

syllables was changing to the consistent system of the six-

teenth century. Early fifteenth-century sources like the

London Chronicles, where the scribal traditions were pre-

served (Kjerrstrom, 1946: 15), generally reflect Middle

English spelling traditions. Letters are occasionally

doubled to represent long vowels, especially mid vowels, as

in queen or goode. Final e is sometimes used to indicate

vowel length, as in coke 'cook,' or kepe 'keep.' ou (ow)

was used almost exclusively for [u:] (in howse or toun)

except in words which had French [0] and [u], where o

frequently occurred, as in consellyd 'counselled,' or

montaynys 'mountains.' Consonant spellings display a few

variants like f for [v] in fochesave 'vouchsafe,' and p for

[b] in pupplyscyde 'published' (also v for [f] and the Old

English spelling sc for [s]). Of some significance is the

occasional use of sch to represent [sk], as in schole

'school' (a transliteration of Greek ox),as it does in

today's English.

In general, though, early fifteenth-century sources

did not distinguish open and close mid vowels in the spell-

ing any more than Middle English did. Length in both instances

may be indicated by doubling or final mute e. The ea spelling

for [e:] (as in great) occurs rarely in the London Chronicles

examined by Kjerrstrom (1946: 250), the oa spelling for [o:]

not at all.

The distinction in spelling of the long open and close

mid vowels is one of the spelling features that marks the

difference between Middle and Modern English. The Old English

ea spelling was revived to represent [e:], especially in

native words like great (OE great) and deal (OE dal). The

oa spelling for [o:] was introduced late and applied less

consistently in words like boat and oath. Lekebusch (1906:

36), using official records that cover the latter two-thirds

of the fifteenth century, says that the ea spelling becomes

more and more frequent during the last twenty-seven years

of the century. But he records no occurrence of the oa


In the Paston Letters, however, both ea and oa occur,

though not with total consistency. Neumann (1904) records

one instance of the use of ea to represent [e:]--meave 'move.'

Otherwise, ea occurs in familiar spellings like please and

meane, as well as the unfamiliar sease 'seize' (32). oa

occurs in oath and broad, as well as in stoan 'stone' (57).

We can see that by the early Modern period, English

had acquired a bewildering variety of alternate spellings,

especially the vowel spellings that were a legacy of French

loan words and pronunciations. The ie spelling of chief,

the eo of people, as well as the eau of beauty were relics

of distinct French pronunciations which had long since been

lost in English.

If we take the Middle English vowels and their spell-

ings listed in Table 5 above and consider the changes in

English pronunciation in the early Modern English period,

we can see what opportunity for confusion existed in these


In the first place, the Great Vowel Shift changed the

relationships between long and short vowels. Once ME [i:]

had become a diphthong, it was no longer phonetically the

long (or tense) version of [i], and the similar spellings

of bit and bite were phonetically misleading. [u:], as we

have seen, was spelled with a digraph ou (ow), so its

spelling was not similarly misleading. The spellings of

long mid vowels were never entirely consistent, but the

final coalescence of the Middle English front mid vowels

left a single vowel [i:] with seven historical spellings (e,

ee, eo, ea, ei, ie, and ai), none of which was directly

related phonetically to the sound it represented. The

diphthongs [iw] and [ew] were in process of changing from

falling diphthongs to rising [ju:]. The [aw] and [aj]

diphthongs were becoming monophthongized into [3] and [e:].

In addition, certain consonant sounds like [x] and

initial [k] and [g] before nasals were disappearing, in some

instances with compensatory lengthening, as in knight.

French influences had added c with the value of [s], and

qu with the value of [k] or [kw].

The vowel system of English in particular was in flux

in the fifteenth and sixteenth centuries. Two other factors

entered the picture at this point--a new influx of loan

words from Latin with Latin spellings intact, along with

etymological respellings of older Latin loan words, and a

gradual freezing of the spelling system, largely because of

the influence of printing.

The fifteenth through the seventeenth centuries saw

a great influx of words like accommodate (the OED records

the first appearance in 1525) and exaggerate (1553), whose

double consonants were not motivated in the English spelling

system, where double consonants indicated short vowels in

preceding stressed syllables. At any rate, the normal

'double' form for [f] was dg, as in edge, not gg. Other

examples are occasion (1382), occur (1527), succession (1537),

and recommend (spelled recommend by Chaucer in 1386).

In addition to unmotivated double consonants, Latin

spellings and respellings brought problems with vowel

spellings. Since all unstressed vowels are subject to

vowel reduction in English, unstressed vowels are also sub-

ject to spelling variation. The polysyllabic words borrowed

from Latin offer more opportunity for variant spellings of

schwaa' than most of the shorter native English words. So

Renaissance borrowings give us problem words like definite

(1553) and separate (adj., 1526), as well as the -ent

endings in words like existent (1561) and prevalent (1576).

The changes occurring in the pronunciation of early

Modern English combined with large numbers of borrowed words

coming into the language and an already staggering number of

spellings for the vowels and diphthongs to create a situation

in which the sound-spelling relationships in English must

have bordered on the chaotic, as some of the whimsical

spellings from fifteenth-century documents suggest. In the

midst of this situation, and perhaps partly because of the

need for stability it created, spelling began to be frozen.

The introduction of printing and the spread of literacy from

the fifteenth century on, combined with the undoubtedly

tremendous influence of the 1611 King James translation of

the Bible, encouraged and helped printers in choosing one

from among the alternative spellings for each word or mor-

pheme and sticking to that choice. By the middle of the

seventeenth century, English spelling had assumed the form

it was to keep (Baugh, 1957: 256-257). Students of English

literature cannot help but notice that at times it seems

that Chaucer never spells the same word the same way twice;

Shakespeare still exhibits some variant spellings and spell-

ings that strike us as unusual; but Milton's spelling is

virtually identical with our own.

From a historical point of view, what English spell-

ing preserves is a hodgepodge of relics from defunct pro-

nunciations and linguistic sources. Words like knight pre-

serve in spelling consonants which have not been pronounced

for centuries; the spellings of heath and beet reflect a

distinction (ME long open [e:] and long close [e:]) that

died out after the seventeenth century. The spelling of

accommodate and prevalent reflects classical Latin, rather

than English, pronunciation.

Insofar as it reflects English pronunciation at all,

English spelling reflects Middle English pronunciation.

This in itself does not make the spelling remarkably incon-

sistent, phonetically or phonologically. The changes of

the Great Vowel Shift were regular changes, and left

regular, although changed, correspondences between the vowels

and their spellings. It is somewhat awkward from the point

of view of terminology that 'long i' is a diphthong [aj]

and 'long e' is [ij], while 'short i' is phonetically closer

to 'long e' than to 'long i' and 'short e' is phonetically

kin to 'long a' [ej]. If, as Chomsky and Halle (1968)

maintain, the phonological representation is very conserva-

tive and the spelling reflects a level of phonology near


the phonological representation, then the relationship

between the spelling and pronunciation of English stressed

vowels is normal and regular. But other influences than

regular phonetic change have been at work on the spelling

of English, so that the underlying regularity of the

system, as it is, is often very difficult to see.


To represent the vowels of Middle and early Modern
English, I shall use the traditional symbols of the
philologists, which were developed for this purpose. I
shall use a colon to indicate length in the vowels and,
when needed, in the consonants. The short vowels of Middle
English will be represented by simple letter symbols--
[i], [e], [a], [o], [u]. The long high vowels will be
represented by letter symbols plus colons--[i:], [u:];
also the low vowel [a:]. The mid close and open vowels
are represented with dots and hooks, respectively--[e:],
[e:], [o:], [o:]. In the relatively few cases where'I
cite Old English forms, I shall use traditional orthography
with macrons to indicate long vowels and long diphthongs.

Palatal c and g (as opposed to the velar stops from
which they evolved) are distinct sounds in Old English,
and cannot always be predicted on the basis of the surface
phonology. Some knowledge of the previous history of the
language is required (Campbell, 1959: 21-22).

3Although vowel length was not marked in Old English
manuscripts, I shall mark long vowels and diphthongs with
macrons in the examples I cite, following editorial tradi-
tion in works on Old English, for the sake of clarity--
ceorl 'churl,' ceosan 'choose.'

This summary of the Old English vowels and diph-
thongs is, of course, greatly simplified and is given only
to illustrate the development of English spelling. It also
ignores the recent controversies about the phonemic and
phonetic nature of Old English vowels and diphthongs. For
a summary and evaluation of dissenting opinions, see Kuhn
and Quirk (1953), who ultimately defend the traditional
philological view. I have followed Kuhn and Quirk because,
in this article and those mentioned below, their arguments
seem more persuasive, and their treatment of the data more
convincing. For a full treatment of Stockwell's monophthongal
theory, see Stockwell (1958), in which short ea, eo, and io
(West Saxon ie) are assigned with phonetic values [a], [E],
[i], which are central allophonic variants of /ae/, /e/,
and /i/, respectively. A rebuttal of Kuhn and Quirk (1953)
was published by Stockwell and Barritt (1955), in which
many of the original claims made by Stockwell and Barritt
(1951) were withdrawn. In the same issue of Language was
published a counter-rebuttal by Kuhn and Quirk (1955). Fisiak

(1968: 32-33, 35) follows the interpretation of Stockwell
and Barritt (1951) and Stockwell (1953).
lo and eo were not distinct in West Saxon, the
literary dialect (Campbell, 1959: 125-126).

Peters (1967) has called the phonological significance
of Old English long consonants into question. He bases his
objections on the principle of complementary distribution.
Few if any convincing minimal pairs can be found for conso-
nant length in Old English. However, Peters reaches no
firm conclusions as to the status of consonant length. I
have followed what seems to me the more satisfactory analysis
of Kurath (1956).

There is some difficulty with terminology here.
Kurath (1956), writing from the American structuralist
framework, uses the term phonemicallyy long' consonants.
In the generative framework, which, except for Schane
(1971, 1973), does not recognize phonemic contrast, the
change from long to short consonants would involve restruc-
turing--change in the underlying representation--as would
the rise of voiced fricatives concurrent with the loss of
consonant length.

Contrastive consonant length was lost in some of the
northern dialects, like that of Orm, by 1200. In more con-
servative southern dialects, long consonants remained through-
out the fourteenth century (Kurath, 1956: 443-445).

In terms of phonology, the distinction between
(continental) Old French and the Anglo-Norman dialect is
not made consistently. Generally, the same sound changes
seem to have occurred in Anglo-Norman as in Old French.

1Since ME [ew] coalesced with ME [iw] in the fifteenth
century (Fisiak, 1968: 55), words borrowed from French with
[ew] also changed to [iw] (as in peutre 'pewter'), which
develops normally into the Modern English rising diphthong
[ju:] in pewter or few.
OF [je] was monophthongized early in Anglo-Norman
to [e]. It was apparently borrowed as [e:]in Middle English
and spelled with e in words like chef. The ie spelling in
words like chief, relief, fierce was rare before the fifteenth
century (Serjeantson, 1935: 298).

12Professor William J. Sullivan has suggested an
example which must have contained both triphthongs--Beaulieu
Abbey (pronounced [bjulij]. This word also illustrates the
development of these sounds in stressed ([ju]) and unstressed
([ij]) position.

3Anglo-Norman, unlike English, did not have distinct
long and short vowels. In English, of course, the distinc-
tion between open and close mid vowels occurred only in the
long vowels, not in the short vowels. Vowel lengthening in
loan words in English followed the same rules that applied
to native English words. For instance, vowels were
lengthened in open syllables under stress--native name,
French robe. French words with [i] and [u] followed T-his
rule as long as the stress was not shifted (bTble, house);
if the stress was shifted, [i] and [u] were short, as in
1The course of development of ME [ew], [iw], and [y:]
(from OE y) has been hotly debated by philologists. Dobson
(1968, II: 699-713) summarized the arguments and presents
the evidence, along with his own theory of the development
of the sounds up to 1700. Kokeritz (1959) suggests a
different theory for the development of [y:] into [(j)u:].
Melchior (1972) argues from the evidence of Thomas Smith
and John Wallis that [y:] must have survived into the later
seventeenth century.
15Later spellings of these words with 1, as in fault,
or falcon, is due to the reborrowing of a Latinized form
from later French, or to the Latinized respelling of words
in English, which will be discussed later in this chapter.

16Catford (1965: 56-61) discusses phonological trans-
lation as transfer at the phonemic level. In generative
theory the transfer would take place at the systematic
phonetic level, with phonetic similarity defined in terms
of phonetic features. The borrowed word would become
attached to an Anglicized systematic phonemic representation
subject to English rules, such as vowel lengthening in open

1There was also considerable influence of English on
Anglo-Norman spelling. The reflex of early Old French ie
in siet (from Latin sapit) was [e], which was associated-
with Middle English [e:J, reflex*of Old English eo; the
result was that eo was added to the possible spellings of
[e] in Anglo-Norman, so that set, siet, and seot were
interchangeable spellings for [set].
1This confusion stems from the fact that Old French
[o] developed from Latin [o:] and [u], just as OF [e]
developed from Latin [e:] and [i]. The Latin spellings
tended to be preserved with u and o after the distinction
in pronunciation disappeared.

9English had three representations for [ ] and [ ].
th was used from the earliest times in Old English, but
thorn (P), of runic origin, and eth ( ) were introduced
during the Old English period. eth disappeared during
Middle English, but thorn continued in use until the intro-
duction of printing ,(OED).

20The Old English spelling was sc, as in scip 'ship.'
Other spellings for [5] used by French-educated scribes
were s, ss, sz. By the fourteenth century sch was generally
used initially (as in schal 'shall'), with other variants
(sh and ssh) used medially (OED).

21Middle English spelling has, of course, been
approached from the point of view of graphemic theory by
McIntosh and Samuels. But apart from McIntosh's initial
paper (1956), and articles announcing early results by
McIntosh (1963) and Samuels (1963), no general application
has been made of graphemics to Middle English spelling and
I have had to rely on secondary sources for most of
my data on fifteenth-century English. When fifteenth-
century documents are available at all, they frequently
have been edited as historical, rather than linguistic,
documents; variant spellings are often regularized, there-
fore, and other information of linguistic interest is some-
times omitted. In the latest edition of the Paston
Letters by Norman Davis (1971), pains have been taken to
produce a reliable text, and to provide other valuable infor-
mation, such as whether a letter is autograph, or was taken
down by a secretary. Unfortunately, this edition is not
yet complete. Three sources I have found especially helpful
are Kjerrstrom (1946), The language of the London chronicles;
Neumann (1904), Die Orthographie der Paston Letters von
1422-1461; and Lekebusch (1906), Die londoner Urkundensprache
von 1430-1500. I believe that the thoroughness of these
works makes them reliable sources for the spelling of the
fifteenth century, even when their interpretation of the
phonology has been superseded.
2The existing studies of Middle English dialects are
not very helpful in this respect. The data in the last
complete dialect study (Moore, Meech, and Whitehall, 1935)
are so limited that they have little value in discovering
spelling patterns associated with different dialects. The
work begun by McIntosh (1963) and Samuels (1963) is more
promising, but still incomplete.


24Trnka (1959) points to the peculiar alignment of
long and short vowels in Middle English as an early stage
of the Great Vowel Shift. [u:] and [i:] had no equivalent
short vowels. [i] and [u] were lengthened to [e:] and [o:],
respectively. [e] and[o] had [e:] and [o:] as their long
counterparts. Examples illustrating the lengthening of [i]
and [u] are [we:kes] 'weeks,' from [wikas], and [wo:da]
'wood,' from [wude](Trnka: 441).



Alphabetic writing is the representation of the

phonetic or phonological segments of a language by discrete

graphic symbols. The relationship between the symbols and

the segments that they represent is necessarily arbitrary

in the sense that alphabets consist of 'pieces,' of dis-

crete images, which are put together to represent units

of language, while phonetic and phonological segments are

not discrete pieces but are made up of features. The

assignment of a particular graphic shape to a particular

segment is motivated by tradition and is not apparently

related to the structure of the language. The spelling of

some homophones in English, as in 'crewell gnus' for 'cruel

news,' indicates the potential independence of spelling

from phonology in indicating meaning; the visual form of

the words can indicate their meaning independent of sound.

Although arbitrary, alphabetic writing is related more or

less directly to the phonetic or phonological segments of a


Alphabets represent these segments at some level of

abstraction. They do not represent sound either at the

articulatory or the acoustical level since segments are

theoretical constructs, not detectable at the level of


production or perception of speech (see .Ladefoged, 1967 and

Halle, 1964).

The traditional motivation for the assignment of

particular graphic shapes to the spelling of English words

has been treated in Chapter I. This chapter is concerned

with the relationship between phonological theory and English

spelling. In this chapter and Chapter III, I shall lay

the theoretical basis for the discussion of Robinson's

alphabet in Chapter V. And, finally, I hope to achieve a

limited synthesis of the two dominant American theories of

phonology, generative and stratificational phonology, to

the extent that they can be related to alphabetic writing.

My discussion will begin with the ancestors of the

current theories, the American structuralist and Prague

school theories, both of which have exerted a powerful influ-

ence on the contemporary schools.

Alphabetic writing is generally (and often tacitly)

taken to be the optimum representation of a language. It

is associated with the phonology of the language, rather than

with the morphology, in contrast to the Chinese character

system, which is essentially morphemic.1 It is also

associated with the phonological segments, in contrast to

the syllabic writing system of Japanese2 and the consonantal

system of Arabic and Hebrew. Alphabets represent each seg-

ment of the phonology, consonant and vowel, with a letter

or combination of letters. English spelling is morphemic

to the extent that it uses alternate spellings to indicate

the difference between homophones (as in they're, there,

their, or crewell gnus, mentioned above). It nevertheless

seems to operate on the basic principle of one letter for

each segment.

The difficulties of dealing with alphabetic writing

come in relating a given alphabetic system to a particular

theoretical interpretation of the phonological segments of

the language in question. Although some languages, notably

Spanish and Italian, seem to present no problems to the

speaker in relating the sound to the spelling, others are

notorious for their inconsistency in relating sound to

spelling (writing) and spelling to sound (pronunciation).

French3 is especially noted for its variant spellings of

the same sound, while English has both variant spellings of

the same sound and variant pronunciations of the same


If the spelling of a language corresponds to the

phonological segments of that language, however inconsis-

tently, the degree of correspondence between the segments

and their representation can be determined only if there

is a clear picture of what those segments are, independent

of the spelling. Since theories of phonology vary on this

subject, some preliminary discussion of them and their

approaches may be helpful.

Until the development of generative phonology, lin-

guists generally maintained that the alphabetic principle

was the correspondence of one spelling unit or letter to

each phoneme of the language being spelled. Geib (1963: 197)

suggests that the word 'alphabet' refers to 'writing which

expresses the single sounds of a language.' Although Gelb

does not use the term 'phoneme,' his notion of 'single

sounds' suggests phonemics. Bloomfield (1933: 290) states

unequivocally that 'the principle of phonemic or alphabetic

writing,' arrived at by the Greeks, was 'the principle of

using a single symbol for each phoneme,' and that the match

between symbols and phonemes was imperfect only because

the Greeks did not have enough symbols to represent both

long and short vowels.6

The imperfect match between phonemes and letters in

some modern languages can be explained in a number of ways.

The Greek (or Latin) alphabet was not altered sufficiently

to account for all the phonemes of the newly written

language; for instance, English has never consistently dis-

tinguished between [d] and [e], although it had sufficient

symbols in the early Middle English period to do so--thorn

(b ) and eth (S) could each represent either the voiced or

the voiceless dental fricative. In addition, the conserva-

tism of scribes caused (and still causes) them to write

words down as they had seen them written, not as they sounded.

This tendency, combined with the inevitable phonetic change

in any language, eventually changes the relationship between

sound and spelling, as it has in the case of the 'mute e' in

English words like bite, a letter which was pronounced in

Middle English. And in English especially there has been a

tendency to spell (and re-spell) words etymologically, so that

English debt reflects the b in Latin debitum, although the b

has never been pronounced in English, and the Middle English

spelling dette more accurately reflects the pronunciation.

(The changing relationship between English spelling and pro-

nunciation has been discussed in detail in Chapter I).

For the traditional phonemicist, then, a one-letter-

per-phoneme correspondence is optimal: a phonemic represen-

tation is the most efficient possible representation of the

(traditional phonemic) phonology of the language because

the phoneme inventory of the language is the minimum number

and variety of units which can be accurately and consistently

used to represent the pronunciation. Departures from the

phoneme-letter correspondence can be attributed to historical

accident and the pressures of regular phonetic change without

corresponding change in spelling (Bloomfield, 1933: 291-293).

The optimum efficiency of a phonemic representation of

the pronunciation of a language is the direct result of the

methods used by American structuralists in establishing

phoneme inventories (the epitome of which is the discovery

of minimal pairs like bit-bet) and of the clearly stated

definition of Prague school linguistics of the phoneme as

the unit of contrast in phonology. From the point of view

of traditional phonemics, therefore, a phonemic alphabet is

the most efficient way to represent the phonology of a


The question to be raised is whether the pronunciation

is what an alphabetic system should represent. It is clear

that English spelling at least in some cases is morphophonemic

rather than phonemic. The letter c, for example, having no

peculiar phonetic or phonemic interpretation, is ideally

suited to represent the morphophonemic alternation between

[k] and [s] in electric-electricity. Because of the French

influence in opaque-opacity, the c reflects only the [s]

part of this alternation, the [k] member being represented

by the French spelling -que.

The c spelling of electric can be seen as representing

a more abstract level of the phonology than the traditional

phonemic level. It can be seen as representing the morpho-

phoneme which is realized as either [k] or [s], depending

on whether or not it is followed by the morpheme -ity.

In traditional phonemic theory, it is the morphopho-

nemics which accounts for the allomorphic variation--such

as the three regular variants of the English plural, [ z],

[s], and [z]--found in most languages. According to Trager

and Smith (1957: 60), 'a full study of English morphophonemics

and vowel sequences that occur, the relation of certain

stresses to specific segmental phoneme structure, and the

relation of intonation to the stresses and junctures; then

would follow a morpheme list with all allomorphs, and an

indefinitely extendable list of morphemes not showing alter-

nation.' Some linguists use morphophonemes in situations

like that in knife /nayF/, in which the /F/ represents a

morphophoneme which may be realized as the phoeme /v/ when


followed by the plural morpheme, and the phoneme /f/ else-

where. Traditional structural phonology, however, does not

postulate a 'morphophonemic level' or representation. For

Harris (1951: 219) the morphophonemic symbol is a device

for marking 'the more common phonological alternations in

a language.' The morphophoneme, then, is a convenient cover

symbol for regular phonological alternations which,

together with specification of morphemic environments and

realizations, predicts those alternations.

The two current theories which have developed this

phonological level have both discarded the term morphophoneme.

Stratificational grammar uses the term morphon for the seg-

ments at this level, and generative grammar uses the term

phonological representation or, for some linguists,

systematic phoneme. The stratificational treatment of mor-

phophonemics is far more complex than the clause above indi-

cates, and I shall discuss its ramifications later in this

chapter. The point to be made here is that, if we proceed

from grammatical to phonological units, it is at the level of

morphons or phonological representations that morphemes or

lexical items (depending on the theory) are first divided

into segments. In other words, it is the most abstract level

of phonology in either theory.

Generative linguists have suggested that, for English

and French specifically, the orthography is near the under-

lying phonological representation, and is therefore optimal

for the adult native speaker of the language. According to

Ch'omsky and Halle (1968: 49)'the fundamental principle of

orthography is that phonetic variation is not indicated

where it is predictable by a general rule. .. Ortho-

graphy is a system designed for readers who know the

language, who understand sentences and therefore know the

surface structure of sentences. Except for unpredict-

able variants (e.g. man-men, buy-bought), an optimal ortho-

graphy would have one representation for each lexical
entry. Chomsky and Halle make no explicit claim that

English spelling corresponds to the underlying phonological

segments. They seem to doubt, in fact, that the phonemic

representation of alphabetic writing, as opposed to their

feature matrices, is 'psychologically real,' But their

definition of an optimal system of orthography makes the

units of the alphabet correspond to phonological segments

at a level very near the underlying representation.

An important difference should be stated here between

generative theory and the other theories of language under

discussion. It is related to the well-known distinction

between competence and performance made by Chomsky and other

generative linguists. Generative theory is concerned not

with manifest speech, but with what the 'ideal native

speaker' knows about his language. This idealistic view

of language accounts for the difficulty Chomsky and Halle

have in relating alphabetic writing to a particular level of


Chomsky and Halle are working with what arsounts to a

sound-image of the phonological segment in the 'abstract

symbols,' which are in fact 'informal abbreviations for

certain complexes of feature' (1968: 10). The cautious

negative statement of the principle of orthography quoted

above, that the orthography does not indicate regular

phonetic variation, makes no precise commitment as to what

the orthography does indicate. It has been commonly

assumed, however--and Chomsky and Halle's first approxima-

tion of the reading process (49-50) encourages this view--

that units of conventional orthography correspond very

roughly with the symbols of the phonological representation

(the systematic phonemes of earlier treatments of generative

phonology). There has been at least one thoroughgoing

attempt to explore the implications of the comments Chomsky

and Halle make about orthography--that of Klima (1972),

which will be discussed later in this chapter.

Before turning to more specific discussions of the

theories, it may be useful here to comment further on the

notion of an 'optimal' orthography. English spelling has

frequently been castigated for its failure to be optimal

in the sense that it does not preserve anything near a

direct correspondence to the pronunciation. This criticism

has been echoed by traditional phonemicists and by

teachers of reading everywhere, and, from their point of

view, English spelling is indeed doubly inconsistent, pre-

serving neither a sound-to-spelling nor a spelling-to-sound

correspondence (see note 4). This failure of spelling to

correspond to pronunciation clearly does make the

teacher's--and the learner's--task in language study more

difficult; therefore, from the standpoint of one trying to

learn English, the spelling system is not optimal. But

the orthography may be optimal from other points of view

than the teacher's and the learner's, as Chomsky and Halle
suggest. The English spelling system may be optimal (or

nearly so) from the point of view of the adult who knows

the language and the spelling system.

Chomsky and Halle's mentalistic linguistics echoes

Sapir's earlier notion of the phoneme as mental image.

Opposed to this was Bloomfield's apparent belief that the

phoneme could be defined in terms of physical sound. Both

these theories will be discussed below in the general dis-

cussion of phonological theory.

A work of major importance, from which I shall draw

heavily for the discussion of American structural phonology,

is W. Freeman Twaddell's monography,"On Defining the Phoneme"

(1935). Twaddell in fact rejects the theories of American

linguists mentioned above, along with the phoneme theory of

Daniel Jones, in favor of his own definition of the phonemes

as an abstract relationship. Although Twaddell writes in the

structural tradition, he reacts against it in his monograph

on the phoneme, and his reactions bring him close in his

definitions to the phoneme of the Prague school.

Twaddell reacts in a manner characteristic of struc-

tural linguistics against theories which involve mentalistic

or psychological definitions. He echoes Bloomfield (1933:

32-34) in his arguments against the 'mental image' defini-

tion of the phoneme. 'Any correlation of phenomena which

can be established on the basis of mental entities or events

can also, and more economically, be established on the basis

of the phenomena themselves' (Twaddell, 1935: 57). The

scientific method, he says (n. 8) is 'quite simply the

convention that mind does not exist.' Specifically, he

attacks arguments for the psychological reality of the

phoneme on the grounds that they rely on negative evidence--

the failure of subjects to record differently sounds (like

the t in tone and the t in stone) which are phonetically


Turning to the linguists, including Bloomfield, who

claim acoustical reality for the phoneme, Twaddell attacks

their position largely on the grounds of the continuous
nature of the acoustical record of speech. It is con-

siderably more difficult, however, for Twaddell to dispose

of the theories of Bloomfield and Daniel Jones than to dis-

pose of the 'mental image' theories. Bloomfield's phoneme,

as Twaddell sees it, is a feature of the speech sounds,

'characteristic of all the speech sounds in question and

characteristic only of these sounds,' while Jones' phoneme

corresponds to the sum of all the speech sounds in question.

Twaddell rejects Bloomfield's 'minimum same of vocal

feature' definition of the phoneme on the grounds that

current research in acoustic phonetics had failed to demon-

strate the feature which supposedly occurred in all instances

of a particular phoneme (63-64).

Jones' definition of the phoneme as 'a family of sounds

in a given language, which are related in character and are

such that no one of them ever occurs in the same surroundings

as any other in words,' fails to account for phonemic over-

lapping in cases like the vowel of dare and air in American

English, which may be assigned to the phoneme in Mary or that

in marry--in dialects where the vowels in these words are

different (64-65). But Jones' definition was not intended

to be 'the theoretical base for the study of phonetic rela-

tionships within a language' (65), but, instead, is intended

for practical use in phonetic transcription.

In addition, Twaddell rejects Morris Swadesh's treat-

ment in"The Phonemic Principle"(1934) because it leaves open

the possibility of arbitrary procedure. The assignment of

the p in spill to the phoneme /p/ rather than to /b/ is

arbitrary on the basis of phonetic similarity (66-67).

If the phoneme cannot be associated with either mental

or physical reality, the alternative for Twaddell is to

regard the phoneme as fiction. 'Although these two procedures

of definition for the phoneme--regarding it as a physical

reality of some order or as a mental (or psychological)

reality--appear to represent the two possibilities, perhaps

they are only subalternatives of one of two possible

procedures' (67). The alternate procedure is definition

of the phoneme as an abstractional fictitious unit. This

definition involves a complicated chain of thirteen inter-

locking definitions leading first to the micro-phoneme (in

9) and finally to the macro-phoneme, which most nearly

resembles the previous notions of the phoneme.

Twaddell defines ordered classes of forms which have

minimal phonological differences (as in the sequence beet:

bit: bait: bet: bat). The minimum phonological difference he

calls the micro-phoneme. The micro-phonemes /i/, /I/, /e/,

/6/, /ae/, in the above sequence are defined, then, as

relationships, as minimum phonological differences, not as


The class pill: till: kill: bill demonstrates the micro-

phonemes /p/, /t/, /k/, /b/. The class nap: gnat: knack: nab

demonstrates a different set of micro-phonemes--different in

that their articulatory specifications are not the same.

The p in pill and the p in nap differ, for example, in

aspiration. These two classes are similarly ordered, however,

since the 'qualitative articulatory differences among the

corresponding phonetic events are similar and in a one-to-one

relation' (73). The p's in pill and nap share the articula-

tory phonetic features bilabial and voiceless; the t's in

till and gnat share the features alveolar and voiceless;

the k's in kill and knack, the features velar and voiceless;

and the b's in bill and nab, the features bilabial and voiced.

'The sum of all similarly ordered terms (micro-phonemes) of


similar minimum phonological differences among fori:s is

called a macro-phoneme' (73). The two ordered classes above,

then, give us the macro-phonemes /p/, /t/, /k/, /b/.

In his Principles of Phonology (1969), written soon

after Twaddell's monograph, N. S. Trubetzkoy responds

directly to Twaddell's ideas. He rejects Twaddell's defini-

tion of the phoneme because he considers it unnecessarily

complex, its complexity being due, he says, to Twaddell's

apparent eagerness to avoid the danger of the phoneme's

being considered a building block for words and sentences,

rather than a relationship. He gives his own definition:

every language presupposes distinctive (phonological)

opposition. The phoneme is a member of such an opposition

that cannot be analyzed into still smaller distinctive

(phonological) units. There is nothing to be changed in

this quite clear and unequivocal definition. Any change can

only lead to unnecessary complications' (Trubetzkoy, 1969:


Trubetzkoy's phoneme corresponds, as he says (42),

exactly with Twaddell's macro-phoneme. From Twaddell's

point of view, this definition leaves open the possibility

of arbitrary procedure like that of assigning the p of spill

to the same phoneme as the p of pill (or to the same phoneme

as the b of bill, the second alternative). Since it is

impossible, on the basis of acoustical or articulatory

criteria, to assign this sound to either the p or the b

macro-phoneme, Twaddell must assign it to a third phoneme:

'The stops of "spill, spare," etc. are significantly bilabial

and stop, but not significantly voiceless; the stops of

"pill, nap, tapper," are significantly bilabial, stop, and

voiceless' (Twaddell: 74). Trubetzkoy and other Prague

school phonologists avoid the problem of arbitrary assignment

of sounds with the notion of neutralization and the archi-

phoneme--the trademark of Prague school phonology.

In Trubetzkoy's book, Prague school phonology becomes

explicitly the study of phonetic features ('properties' in

the Baltaxe translation) and relationships. Most of the

Principles of Phonology is devoted to defining and classify-

ing the relationship between phonemes, which in turn are seen

as bundles of distinctive properties, Perhaps because he

does not limit his treatment to a particular language--nor,

indeed, does he provide any extensive analysis or examples

from any one language--the relational nature of his theory

becomes especially clear.

It is the concept of neutralization and the archiphoneme

which makes the relational nature of Prague school phonology

especially attractive. Twaddell's solution of a third

phoneme in the problem mentioned above leaves open the possi-

bility of considering the phoneme as a piece from which

larger phonological units are built--the notion which Twaddell

was at such great pains to avoid in his definitions. To

comprehend the concept of neutralization in phonology, one

must have a concept of the phoneme as a relationship, or

'fiction' in Twaddell's term.

It is characteristic of Trubetzkoy's approach that he

takes up the logical classification of distinctive opposi-

tions (Chapter III) before he discusses the phonological

classifications (Chapter IV). His phonetic properties are

those now familiar to students of phonology regardless of

their theoretical background. They include the basic divi-

sion of features into vocalic, consonantal, and prosodic

and the features of place ('localization') for consonants.

Although some of Trubetzkoy's terminology may be unfamiliar,

and his treatment of phonetics is more varied than most

others, the notions and classification of the features are

The logical basis of Prague school theory distinguishes

it from the other theories like those of Sapir and Bloom-

field, as well as that of Twaddell. The notion of neutrali-

zation depends on Trubetzkoy's initial classification of

opposition into bilateral and multilateral opposition.

In bilateral opposition the sum of the features that the

two phonemes have in common is common to those two phonemes

alone; 'the basis for comparison of a multilateral opposi-

tion is not limited to the two respective opposition members'

(Trubetzkoy: 68). It follows, of course, that only members

of a bilateral opposition--like English p and b--can be


The treatments of phonology by American linguists

ignored the logical basis of phonological theory perhaps

because they were largely concerned with discovery procedures.

Swadesh's sections in The phonemic principle (1934) on

method, orthography, normalization, and phonetics are clearly

written for the field linguist faced with the task of analyzing

an unfamiliar language. Indeed, the phonemic principle itself

was originally a part of a paper on Chitimacha phonemes,

as Swadesh indicates at the beginning of the paper (32).

Whatever the cause, emphasis on discovery procedures has

come to be associated with American structural linguistics;

an emphasis on logical structure and analysis has come to be

associated with Prague school linguistics.

Twentieth-century phonological theories before trans-

formational-generative phonology differed from each other in

many ways (some of them indicated abovel2), but they had a

common goal--identification and description of the phonemes

of languages. There was also a general tendency to treat

phonology autonomously, without considering the possible

connections between phonology and grammar or semantics--the

legacy no doubt of the neogrammarians' specifications of

linguistic rigor.

Generative grammar introduced a number of changes that

were startling in the context of the American structuralism

with which it was originally placed in direct competition.

Instead of the behavioristic 'no mind' approach of American

structuralism suggested by Twaddell (above), generative grammar

proposed to describe explicitly the ideal native speaker's

knowledge of his language. This frankly mentalistic approach

is derived at least partly from the mentalistic notions of

Sapir, which were rejected by Twaddell, a debt that Chomsky

(1964) acknowledges.

A feature of generative grammar closely related to

its mentalistic approach is its distinction of competence

(what the ideal speaker-learner knows about his language)

and performance (which is affected by such non-linguistic
factors as fatigue and excitement). The business of

grammar is to describe the competence of the native speaker.

In accounting for the speaker's intuition about his language

(Chomsky, 1964: 63), the grammar must account for all possible

forms. In phonology, this means that generative grammar

claims to fill in phonological gaps; therefore, forms like

/nis/ and /blik/ will be predicted, although they do not

actually occur (they are meaningless) in English (Halle, 1962:


Generative phonology differs in several specific ways

from American structuralist phonology. Halle (1959) provides

an argument against the autonomous phoneme, and Chomsky (1964)

amplifies his argument. (The status of the phoneme in

generative and stratificational phonology will be examined

in Chapter III.)

Although they relied on native-speaker intuition (in

responses of 'the same meaning' or 'different meanings') in

their establishment of phonemes, the American structuralists

rejected mentalistic notions (and 'meaning' in general), and

minimized the importance of the phonetic level. They tended

to see the (autonomous) phonemic level as the structurally

significant level, and to view the phonetic level as non-

linguistic,15 although in practice they relied heavily on

feature analysis and 'phonetic universals' (Chomsky, 1964:

92-93). Generative grammar, in contrast, seeks to develop

a universal phonetic theory which will be incorporated into

the phonology at the systematic phonetic level (Jakobson, Fant,

Gunnar,and Halle, 1963; Halle, 1964; and Chomsky and Halle,

1968: 293-329).

At the top of the phonological component is the

phonological representation, provided by the operation of

the readjustment rules (Chomsky and Halle, 1968: 371-372)

as the input to the phonological component. The phonologi-

cal representation is rendered in phonological segments,

which, in their turn, are matrices of distinctive features

which are subject to successive, ordered rules which

finally result in the systematic phonetic level, correspond-

ing to the pronunciation.

A word, phrase or sentence may have written represen-

tations corresponding to its sequence of feature matrices

after the application of each rule. This is called the

derivation of the word. It begins at the most abstract level

of the phonology, the phonological representation. The

application of each successive rule brings the word closer

to the surface level, and each representation is progressively

less abstract. Thus, the derivation of ambiguity (Chomsky and

Halle; 195):






The segment undergoing the change in this sequence is /u/,

which is the cover symbol for the following feature matrix


/u/ vocalic +

consonantal -

high +

back +




round +


Chomsky and Halle's features at this point--the

phonological representation--are not 'phonetic' in the sense

that they have a pronunciation. Features are defined in

articulatory terms only at the lowest--systematic phonetic--

level. Above that level they are abstract categories. Each

place in the phonological matrix represents a category and

its component (165). The category [+tense] has as its

complement [-tense].

In the derivation above, the segment /u/ undergoes a

tensing and unrounding rule, making it +rtensde, and yielding

the second representation [aembig+T+ity ]. A rule of diph-

thongization adds the segment [w], in features:

[w] vocalic


high +

back +




round +


This yields the third representation [aembig+iw+ity.]. A [y]-

glide is added to this form in features:

[y] vocalic -


high +







This yields the form [aembig+ytw+ity to which a rounding

rule applies, changing [T] to [u].

vocalic +


high +

back +




round +

tense +

This finally yields the surface form [abmbig+yuw+ity]. From

the underlying phonological /u/, the rules supply [yuw]. In


/u/ ---- [yuw]

+ vocalic -+-


+ high +++

+ back -++

low ---

anterior ---

coronal --

+ round -++


Since each phonological phrase undergoes the applica-

tion of rules as a unit for rules involving stress in

English (the 'transformational cycle' in Chomsky and Halle),

and the word undergoes the remaining rules, each phonological

phrase or word has its unique derivational history. Since

the language does not move monolithically through the rules,

it makes no sense to attempt to describe a level at which

a representation can be made for the language as a whole

between the phonological and systematic phonetic represen-


The above derivations illustrate the componential

nature of generative phonology. It lacks a segmental

aspect--phonotactics or alternation patterns. The segments

are supplied at the top of the phonology by the lexicon and

the readjustment rules. Admissible segment sequences are

specified by the phonological rules.

The alphabetic representation of English corresponds

roughly to the phonological representation for Chomsky and

Halle. In the example above, /aembig+u+ity/ is nearly

identical with the spelling of the word it represents,

ambiguity; the only segment which does not correspond with

the spelling is /ae/, for which the letter a is always the

English spelling. The phonological representation of

tabular (Chomsky and Halle: 197) is /taebl+aer/, reflecting

the stem table. The spelling corresponds more closely to

the form of the word after the application of a rule that

inserts [u] and a laxing rule which yields the form

[taebul+aer ].

Chomsky and Halle do not state a direct equivalence

between the spelling and the phonological representation,

however, as I have stated. They take the more cautious path

of defining the principle of orthography in negative terns,

suggesting that it should not reflect phonetic variation

where that variation is predicted by a general rule (49).

Halle develops the principle, stating it as, 'ortho-

graphies must contain no symbols that reflect the operation

of phonological rules' (Halle, 1969: 19). The difference

between phonetic [s] and phonetic [z] in consign and design,

respectively, is predicted by a phonetic rule (intervocalic

voicing), and therefore is not represented in the orthography.

Double consonants--not pronounced in English--may also have

a function in the orthograph as they do in the phonology.

The ss sequence in dissent or dissemble blocks the voicing

rule, mentioned above, yielding unvoiced [ss], which is then

reduced by another rule to [s]. The orthographic represen-

tation ss, then, reflects the phonological representation

before the application of the rules, according to the

principle (21).

It is interesting to note that this principle of ortho-

graphy also finds a legitimate use for the letter c. While

c has no distinct sound of its own, always taking the sound

of k or s, this very fact makes it the ideal symbol for the

[k]-[s] alternation in electric-electricity, medical-medicine,

and vocal-vociferous.

Klima (1972) has further pursued the generative ortho-

graphic principle. He suggests four additional principles

in the creation of an optimal orthography: (1) minimal

arbitrariness in representation, (2) minimal redundancy,

(3) sufficient ambiguity or expressiveness, and (4) standardi-

zation (61).

Alphabetic writing is necessarily minimally arbitrary,

as compared with the ideographic of pure syllabic systems.

Alphabets represent the essential segments of the language,

rather than the larger and more numerous elements of syllables

and words. Although an element of arbitrariness is necessary

in representing phonological segments with graphic symbols (a

change from an essentially auditory--albeit abstract--image

to a visual image), alphabetic writing reduced arbitrariness

to a minimum by representing the least numerous and most

frequently repeated elements of language, the phonological

segments. This reduces the number of graphic symbols to a

minimum, making the load on the memory of the language user

as small as possible.

Of course, it is possible for the orthography to indi-

cate distinctive features--p and b are clearly related in

the printed version of English orthography (and in the

orthographies of all other languages using the Latin alpha-

bet). A system that is too thorough in representing the

features, however, runs the risk of violating the first

principle of orthography and representing automatic features.

This unnecessarily phonetic representation also leads

to violation of the principle of economy or minimal redun-

dancy. It would make little sense to have three representa-

tions for the three phonetic k sounds that occur in English

(Klima, 1972: 64): medial in cop, back, labialized in coop;

and palatal in keep.17 The principle of economy in effect is

the same as Chomsky and Halle's orthographic principle.

The principle of expressiveness allows the distinction

of homophones in spelling, as in there and their in English,

and the principle of standardization requires the same word

to be spelled the same way wherever it occurs.

Klima experiments with a number of possible interpreta-

tions of Chomsky and Halle's orthographic principle. He

demonstrates that the relationship between the spelling of

English and any level of the phonology is anything but simple

and straightforward. All the principles he suggests fall

short of adequacy in spelling English. The underlying repre-

sentation of fashion, for example, is /fac+ion/ (70). Tabbed,

tapped, and patted would be spelled tabd, tapt, and pated

by a convention that would permit somewhat less abstract

representations (71). A convention which allows the ortho-

graphy to reflect everything except what is predicted by

regular phonological rules--excluding only those phonetic

effects ascribable to surrounding sound segments and internal

or external word boundary (79)--yields the correct spelling

of oblivion, but the incorrect spellings of rebelyon (rebellion)

and crusyal (crucial).

Klima points out (72) that real orthographies often

reflect the representation of the word rather than its sound

form when the two diverge. For example, the spelling of the

preterite morpheme -ed, cited above, reflects the unity of

the morpheme, rather than the predictable [t], [d], or [Od]

phonetic forms. Thus, spelling in some instances may be only

indirectly phonological.

At any rate, generative phonologists have demonstrated

that spelling in a language like English (and presumably

other languages as well) may be consistent at a level of

language more abstract than pronunciation, although the

optimal nature of the English spelling system has yet to be

conclusively demonstrated--in fact, English spelling reflects

different levels at different times, as the examples cited

here indicate. Phonologists have suggested that in fact

orthography should not reflect the pronunciation, since the

language may be more economically represented at more

abstract levels.

While traditional structural linguists insisted on

viewing writing as the representation of the pronunciation

or the phonology, rather than talking about the pronunciation

of the spelling, the generative interpretation of orthography

has made this distinction as parallel manifestations of an

abstract level of language, so that it makes as much sense

to talk about the pronunciation of the spelling as it does

to talk about the spelling's representing the pronunciation.

King, in his chapter on scribal practice (1969: 203-213),

discusses the competing notions that scribes tend to repre-

sent autonomous phonemes on the one hand (if the analyst is

a traditional phonemicist), or deeper phonological segments

on the other (if the analyst is a generative phonologist).

Scribalpractice, however, is so inconsistent--as in fifteenth-

century English manuscripts--that it is difficult to explain

it consistently on any basis.

It should be observed here that the scribal practice

described in King (1969) is a problem of historical linguis-

tics, while the optimal orthography of Chomsky and Halle and

of Klima (as well as of traditional phonemic theorists) is

a problem of theoretical linguistics. Optimal orthography

is a question of competence in writing, to use generative

terminology, whereas scribal practice is a matter of perform-

ance. The two areas should illuminate each other, but they

cannot be expected to be identical. Historical linguistics

must account for occurring data, while theoretical linguistics

attempts to find the abstract system behind the data, which

is then used to organize and describe the data.

Stratificational theory is an alternative to generative

theory at all levels of analysis, although it has perhaps

been most consistently developed in phonology. The success

that stratificational grammar has had as a basis for machine

translation (see, for example, Lamb, 1964) recommends it

for attention if it had no other appeal. But the theory has

more to recommend it than this basis, and it has many
ancestors other than the computer.

One feature of stratificational grammar that sets it

apart from generative grammar is that it rejects the strict

distinction made by Chomsky and others between competence--

what the speaker knows--and performance--what the speaker

actually produces. This feature is central to Algeo's

(1970: 266-268) classification of stratification grammar as

'process-oriented' as opposed to 'system-oriented' generative

Another important distinction is that stratificational

grammar does not have 'rules' in the same sense that

generative grammar does. The rules of generative grammar

are unidirectional, moving (in phonology) from the more

abstract phonological representation to the more concrete

systematic phonetic level (the pronunciation, in effect).

The lines and nodes of stratificational grammar define rela-

tionships and are bidirectional. In the part of the grammar

relevant here, the morphemes of the morphemic sign pattern

are related to morphons (morphophonemes of earlier theories),

which are related to phonemes by way of the morphonic alter-

nation pattern. Phonemes in turn are related to phonons

(phonetic features) at the bottom of the phonotactics. This

is the 'realizational' portion of the phonology. In addition,

each stratum has a tactic pattern, which is the combinatory

element of that level, specifying the order of the elements,

so that the stratificational scheme may be presented in the

form of Figure 1.


Figure 1

The Scheme of Stratificational Phonology

Stratificational linguists have not declared an ortho-

graphic principle.20 But the nearest equivalent to the

phonological representation of generative grammar is the

morphonic level. This level, the downward component of the

morphemic sign pattern, consists of what were formerly

called morphophonemes. The morphonic alternation patterns

have traditionally been called morphophonemic.

It is at this level--morphonic for stratificational

grammar and phonological representation for generative

grammar--that the two theories show perhaps their greatest

similarity. Stratificationalists have done relatively little

work with the English language, but, when work is done with

English, the morphonic level will no doubt be very similar

to the phonological representation of Chomsky and Halle.

The morphonic level, like its ancestor, the morphopho-
nemic level,21 contains morphemic elements before they are

distinguished into their variant phonemic forms--the plural

morpheme is represented on the morphonic level probably as
/z/,22 the preterite as /d/. The morphophonemic variation

in inflectional suffixes in English is easy to demonstrate.

What may not be immediately obvious is that the morphonic

level will reflect the same conservatism that the generative

phonological representation does. For example, sane and

sanity would have different vowels (/ej/ and /ae/, respectively)

on the phonemic level, but the same vowel at the morphonic


The disagreements between generative and stratifica-

tional grammar involve almost all levels and aspects of

language--the overall structure of the theory; the existence

of morphemic, lexemic, and sememic levels; the existence of

the 'autonomous' phoneme; and the nature of phonetic features--

except this one. Insofar as the phonological representation

of generative theory can account for spelling, then, strati-

ficational theory can also account for it.

Perhaps it should be borne in mind that the process of

creating an orthography in a state of linguistic naivete--a

state in which most orthographies were in fact created--makes

the resulting system very much subject to confusion or preju-

dice on the part of the scribe, not to mention the adequacy

(or inadequacy) of the available symbols. Thus, with the

French influence in Middle English, English scribes learned

to distinguish f and v in the spelling, but French had no

[C] or [8] sounds, and so the distinction, which became

phonemic in Middle English, found no representation in the

spelling. Finally, the print imported from France by William

Caxton contained no letter to represent either the voiced or.

the voiceless dental fricative--which, of course, did not

exist in French. So, after some experimentation with y

(ye for the), [d] and [0] both came to be represented by the

digraph th (see Chapter I).

Historical accidents lead to inconsistencies of the

kind just mentioned, in which one voiceless-voiced pair

(f and v) is distinguished in the spelling, but another

([-] and [e]) is not. Another kind of difficulty is that

English spelling is often unaccountably phonetic. The

voicing alternation in knife-knives is reflected in the

spelling, although in the similar house-houses and bath-
bathe, it is not. If we accept the underlying form of

the preterite and plural, with Shibatani (1972), as /d/ and

/z/, respectively, the -ed spelling in all regular verbs

and the -es spelling after sibilants in nouns are unneces-

sarily phonetic in the generative view.24

Both these kinds of inconsistency have their roots in

the history of the language and the origins and development

of the spelling system. There are also relics in the spell-

ing system.of sounds that have been lost. The initial k

of knife and knight, for example, were pronounced in Middle

English, and the spellings of beet and beat reflect a distinc-

tion in pronunciation that died out in the seventeenth


A different and important kind of non-phonological

phenomenon in English spelling is a number of patterns of

some generality which do not directly reflect the phonology,

but are part of the spelling system as distinct from the

phonology. These include the gh spelling in knight which,

while it is a relic like k, serves to indicate a preceding

long (or tense) vowel.* Similarly, single consonants followed

by 'mute e' indicate a preceding long vowel, and double

consonant letters, also for historical reasons discussed in

Chapter I, indicate preceding short vowels in stressed

syllables--mate has tense [ej] and matter has lax [a ].

Finally, there is a certain amount of spelling

irregularity which may be accounted for by the historical

source of the word, but not by the phonetics of English

phonology at any stage. The double c of accommodate

accurately reflects the Latin pronunciation of double

consonants, but its sixteenth-century adoption into

English (OED) came long after double consonants had ceased

to be pronounced in the thirteenth century. Furthermore,

following an unstressed vowel as it does, it can indicate

nothing about the quality of the vowel (normally reduced to


There is a large number of elements, patterns, and

relics of various kinds in English spelling which would un-

doubtedly be counted as irregular in any theory of phonology.

Nevertheless, much of what has been condemned about English

spelling--such as the letter c with no single phonetic

value, and the letter a used to represent both the [as] of

telegraph and the [a] of telegraphy--can in fact be seen to

be perfectly regular at the morphonic or phonological level.


In Chinese, the morpheme and word are nearly identi-
cal, the major form of morpheme combination being that of
compounding. There Are only a few restricted exceptions
like the men plural ending applicable only to personal
nouns and pronouns like wo 'I' (women 'we'), and haidz
'child' (haidzmen 'children').

See Samuel E. Martin (1972) for a detailed discussion
of the Japanese writing system.

It may be of interest to note that, in a structural-
ist or Prague school phoneme inventory, standard French and
standard English have about the same number of phonemes--
forty, more or less, depending on the authority cited--with
an especially large number of vowels. Spanish and Italian,
on the other hand, have relatively few phonemes, especially

In French a single pronunciation may be represented
by more than one spelling; for example, o: may be spelled
au or eau; a may be spelled en or an, but a given spelling
generally has only one pronunciation. In English, as beet,
beat, weather, illustrate, the irregularity goes both ways--
more than one spelling is possible for a given sound, and
more than one pronunciation is possible for a given spelling.

The distinction is often made between letter, on the
one hand, and grapheme or spelling unit, on the other.
Letters are members of our familiar twenty-six unit alpha-
bet, the abc's. Grapheme and spelling unit are both used
as technical terms to indicate the graphic representation
of language; they are used to avoid the confusion that may
be brought on by the fact that combinations of letters may
represent single segments, as in English th and sh for [8]-
[d] or [s], respectively. Grapheme, as used by Kurath (1964),
for example, seems to indicate the representative of the
phoneme, while spelling units, used by Venezky (1965, 1970)
indicates a spelling system which is more independent of the
phoneme system, having its own rules and patterns. This
distinction reflects a disagreement among linguists as to
the status of alphabetic writing either as a direct repre-
sentation of the phonology or as a system parallel to but
independent of the phonology. I shall discuss this question
later. I believe that, for the purposes of this discussion,
the terms letter and alphabet gain in comprehensibility and
familiarity what they may lose in precision. They are useful
because they preserve the ambiguity in the meanings that
grapheme and spelling unit separate--they can be used to
indicate either phoneme-representative systems or more
independent systems. I have found no reason at this point

to be in total agreement with either interpretation; there-
fore, I shall use the terms letter and alphabet as unambigu-
ously as possible to indicate a segmental spelling system
which may be either phonological or independent.

The relative recency of this view of the relationship
between ancient Greek spelling and pronunciation is attested
by the bitterness of the quarrels in the sixteenth century
between those who wished to change the academic pronuncia-
tion of ancient Greek to make it fit the spelling, and those
who wished to preserve it in the contemporary Greek pronun-
ciation (see the Introduction).

English spelling here represents the alternation at
the phonemic, not the morphophonemic, level.
The last statement in this quotation echoes the fre-
quent observation that English tends to spell the same
morpheme in the same way whenever it occurs, regardless of
changes in pronunciation.
Kiparsky (1975: 277) remarks that language may be
optimal in one of three ways--from the standpoint of
encoding, from that of decoding, and from that of learning.
It cannot be optimal in all three ways at once, and the
resulting tendency toward the optimum in all three areas
causes the constant change that we know to be a fact of

1Bloomfield's definitions of the phoneme are suffi-
ciently ambiguous for Fudge (1970: 80) to identify them
with the 'functional' view, which includes Trubetzkoy and,
probably, Twaddell.

1The great variety of languages cited in Trubetzkoy's
book leads to a far greater degree of refinement in the
system of features than is generally necessary for the treat-
ment of a single language, or a small group of languages.

12There are two European theories of phonology I have
not covered in this chapter because they do not seem to have
an important bearing on the subject--the relationship of
alphabetic writing to phonology. They are the prosodic
analysis of J. R. Firth and F. R. Palmer, and the glosse-
matics of Louis Hjelmslev and his followers. A brief discus-
sion of both theories can be found in Fudge (1970), and the
origins of prosodic analysis have been treated extensively
by Langendoen (1968).

13The competence-performance distinction is similar
to, but not identical with, the langue-parole distinction
of Saussure and the habit-behavior distinction of Hockett,
as Algeo (1970: n. 6) points out.

14Although genbrative phonology claims to fill in
phonological gaps, stratificational linguistics claims that
this is not possible without a phonological base (tactics)
to define the possible combinations of segments (William J.
Sullivan, personal communication).

1Prague school linguistics developed phonetic theory
more fully than American linguistics. This phonetic theory
has been employed more recently in the phonemic theories of
Mulder (1968) and Martinet (1960, 1962).

1Sanford Schane has found the phoneme alive and well,
not between the top and bottom of the derivational chain,
but at the systematic phonetic level. He observes (1971:
520) that generative grammars generate an 'explicitly broad
phonetic representation, which, implicitly, is a representa-
tion of surface contrasts.' The output of generative
phonology is therefore often almost identical with tradi-
tional phonemic representation.

1The fact that cop, coop, and keep have two different
representations for /k/ is due to a rule of spelling that is
only partially phonetic, that c has the value of /s/ before
i and e. The reason for the development of this rule lies
in the French influence on English spelling, discussed in
Chapter I.

18Lamb acknowledges his debt especially to Hjelmslev
in his Epilegomena to a theory of language (1966a).

19Algeo (1970) outlines this and other ways in which
stratificational grammar differs from generative grammar--
the number of parallel generative mechanisms (strata)--six
for English--and the nature of the 'rules' connecting the

2Stratificational linguistics is not likely to declare
a principle of orthography like that of generative linguistics.
The question of the level of the phonology to which spelling
corresponds is considered language-specific. Czech spelling,
for example, is phonemic (William J. Sullivan, personal
It may be of interest to note that the term
morphophonemic was considered in earlier versions of
generative phonology, along with systematic phonemic, for
the level Chomsky and Halle (1968) call the phonological


22Arguing in the generative framework, Shibatani (1972)
concludes that /z/ is the underlying representation for the
plural morpheme, and /d/ for the preterite.

23Of course, standard English spelling cannot make
the voiced-voiceless distinction of bath-bathe. It has
the letters s and z to make the distinction in house-houses,
but English spelling practice follows French in generally
using s to represent both [s] and [z] unless [z] is initial.
2These spellings reflect Middle English forms in
which the vowel--in the ending of walked, for example--was



The concept of the phoneme has been largely rejected

as a theoretical reality by generative linguists since the

arguments against it presented by Halle (1959) and by

Chomsky (1964). However, as indicated in the previous chap-

ter, some theoretical linguists have continued to find the

phoneme to be a viable and important part of their theories.

Lamb (1966b) has demonstrated that the phoneme is defensible;

his arguments will be summarized below.

Algeo (1970) has pointed out that stratificational

and generative linguistics are not arguing at cross purposes.

He suggests that the 'system-oriented' generative grammar

and the 'process-oriented' stratificational grammar in fact

describe different, and perhaps compatible, aspects of

language. I shall try to show in this chapter that the

phoneme is a necessary consequence of a basic requirement

of stratificational linguistics, while the presuppositions

and form of generative theory make it difficult to have a

clear formulation of a phonemic level in that theory.

The disagreement among theories over the viability

of the phoneme has two bases: the question of the proper

object of linguistic enquiry and the question of the form

that the theory takes. The object of generative enquiry is


competence, or the rules that the native speaker knows which

enable him to speak and understand his native language.

Theories which espouse the phonema often take as their

object of enquiry something like the Hjelmslevian notion of

texts, or manifest lanquace, and take as a linguistic proce-

dure the analysis of texts. Lamb (1966a) took texts as the

object of enquiry, with a view to describing the system

underlying the production and comprehension of texts. More

recently, he has suggested that the object also includes the

information system that the native speaker knows which enables

him to speak and understand his language. This notion is

similar to the generative competence, but Lamb explicitly

rejects the generative dichotomy between competence and


The phoneme itself was, of course, conceived as a

theoretical entity long before the distinction between compet-

ence and performance was made. The phoneme was well estab-

lished in linguistic theory before the second world war.

American linguistics of this period was behavioristic in its

orientation. The.concentration of American anthropological

linguistics from Franz Boaz through the second world war on

the recording of languages foreign to the linguist encouraged

a view that linguistic behavior was the proper object of

linguistic endeavor. This concentration also encouraged

the general procedural orientation of American structural

linguistics, which Chomskyan linguists have found objection-


The procedural orientation of the American structural-

ists made the isolation of the distinctive sounds of a

language--the establishment of the phonemes--the first

necessary step in the analysis and recording of the language.

It was this procedural orientation which led Bloomfield to

state (1933: 20) that 'the only useful generalizations about

language are inductive generalizations.' Attacks against

phoneme theory have usually actually been attacks against

the procedural bias of the theories which advocated the


Halle makes one of the early arguments against the

phoneme as a theoretical entity. He provides (1959: 19)

'six formal conditions which phonological descriptions must

satisfy.' Conditions (3), (3a) and (3a-l) are central in

his argument against the phoneme (pages 21-23):

Condition (3): A phonological description
must provide a method of inferring (deriving)
from every phonological representation the utter-
ance symbolized, without recourse to information
not contained in the phonological representation.

Condition (3a): A phonological description
must include instructions for inferring (deriving)
the proper phonological representation of any
speech event, without recourse to information not
contained in the physical signal.

Condition (3a-l): Only utterances which are
different are to be represented by different
sequences of symbols. The number of symbols
employed in all representations must be compatible
with this objective.

Traditionally linguistic descriptions have contained
both representations satisfying Condition (3) alone,
and representations satisfying Conditions (3) and
(3a). The former are usually called 'morphopho-
nemic' to distinguish them from the latter, which
are called 'phonemic.'