[Simplified Spelling Society Newsletter, Spring 1986/1, p14 Later designated Journal 2]
[See other journal articles by Edgar Gregersen.]

Morphological Considerations in the Creation of Rational Orthographies.

Edgar A. Gregersen.

[Professor Gregersen teaches and researches in the Department of Anthropology at Queens College and Graduate Center in the City University of New York. He has been writing on aspects of orthography since 1977. This article is an edited version of the address he gave to the Society's Fourth International Conference in July 1985].
"English orthography, despite its often cited inconsistencies, comes remarkably close to being an optimal orthographic system for English." (The Sound Pattern of English, 1968, p.49)
This assertion by Noam Chomsky (perhaps the most influential linguist of the 20th century) and Morris Halle (another linguistic giant) is probably the most serious intellectual blow the spelling reform movement has yet suffered.

It doesn't matter that many linguists, especially in Europe, are no longer particularly entranced by Chomskyan theory. The reality is that Chomsky's prestige in intellectual circles in the United States alone suffices to give his views tremendous clout. This is especially true because all students on the college or university level are exposed to his ideas if they take courses in linguistics. For example, in a widely used college textbook for elementary linguistics courses, An Introduction to Language by Victoria Fromkin and Robert Rodman (1983, 3rd ed.), the question of spelling reform for English is considered - and largely rejected, using Chomskyan arguments (see pp.156-159). Similarly, in an essay on English orthography in The English Heritage Dictionary of the English Language (much beloved by American English teachers), Wayne O'Neill makes the same argument.

Spelling reformers must take up the challenge of Chomsky's critique and see how they can use it to best advantage.

First however, let us consider the relevant points in Chomsky's position. His major point is that variations in the related forms of a word (or, more technically, the allomorphs of a morpheme) should not be indicated if predictable. For example, stress is very often predictable in English words and when it is, it should not be shown. Compare phOtograph-photOgrapher-photogrAphic or hIstory-histOrical-historIcity with a totally regular pattern of stress on the third syllable from the end. However, the morpheme {histori} consequently has three different allomorphs: /hIstori, histOri, historI/ (capital letters indicate stress; other variations are not shown).

Similarly with vowel reduction. The vowel /o/ in the allomorphs of {histori} often is reduced to shwa when not stressed, thus /hIstəri/ or /hIstri/, and /histərI/ as in historicity. The single form from which all the variations can be derived is of course hypothetical. But Chomsky sets it up as an underlying representation and he thinks that this is what should be the basis for conventional orthographies.

He notes also that in different dialects the underlying representations will be the same even tho the actual forms used (the surface representations) may be quite different.

The implications for spelling reform in English are the following.

(1) Related words that can easily be shown as related should not be broken up. The most egregious example of breaking up a morpheme unwarrantedly is the 1971 proposal adopted by the Simplified Spelling Society (SSS) that your and yours be written yoor and yoors but you should be written U. The older 1948 forms in classic Nue Speling were superior (tho not wonderful): U, uer, uerz. I myself prefer something on the order of yu, yuur, yuurz. The form yu is in fact currently sponsored by the Simplified Spelling Society of Canada and was suggested by the now defunct Simplified Spelling Board (an American group).

(2) Inflections should as much as possible be written without variation. A notable example is the regular English plural, which in traditional orthography is written as <-s>. In Nue Speling (of whatever vintage) it is broken up into <-s> and <-z>, as in kats, dogz. The recent proposal (1985) put forth by the SSS's working party that the plural always be written as <-z> is a big improvement and should be adopted immediately. (Interestingly enough, in the book by Fromkin and Rodman mentioned earlier, the authors assert:
"It is doubtful that anyone would suggest that the plural morpheme should be spelled S in cats and Z in dogs. The sound of the morpheme is determined by rules..."
(P. 158).
The authors have clearly not done their homework since virtually all spelling reformers have hitherto broken up the plural morpheme. As a matter of fact, I know of no exceptions apart from the recent SSS proposal.)

(3) Vowels in unstressed syllables must reflect the stressed vowels in related words, e.g. symbol-symbolic, substance-substantial. This is precisely what is done (for the most part) in Nue Speling (simbol-simbolik, susbtans-substanshal) and should be retained, perhaps even being made more consistent (cf. the classic 1948 Nue Speling forms leprus but leprosy; also kalus, jenerus but kalosity, jenerosity.

(4) Regular vowel alternations in stressed syllables reflecting the Great Vowel Shift of Early Modern English have been called by Chomsky and Halle 'without doubt the pivotal process of Modern English phonology' (The Sound Pattern of English p.187). These suggestions are exemplified in the following often quoted examples (the phonemic values are given in the traditional Daniel Jones transcription for RP):

/ei/
/i: /
/ai/
/ou/
- /æ/
- /e/
- /i/
- /o/
sane
serene
divine
verbose
- sanity
- serenity
- divinity
- verbosity
Comparealso: 
/(j)u:/- /Λ/presume- presumption

Some proposals for the reform of English spelling would obscure these alternations by using the 'international' values of vowels, e.g. sein (for sane), devain (for devine), verbous (for verbose). It is really these proposals that Chomsky's main displeasure is directed against.

However, the traditional Nue Speling conventions in fact preserve the relationships in the most practical way possible (other than by introducing some diacritic), viz, by indicating the so-called 'long' value of vowels by vowel plus <-e>: saen, divien, verboes. The actual design of New Speling conventions thus accommodates the regular vowel alternations quite adequately and is a decided virtue.

The recent proposal by the SSS working party to substitute <y> for <ie> would break up the regular alternations. This proposed use of Y should be abandoned.

No orthography can be totally morphemic if only because scholars disagree on what that would entail. Even tho traditional English orthography often does ignore differences between allomorphs, it is not entirely consistent; cf. five but fifty and fifth (one could write fivty and fivth but no one to my knowledge has suggested doing so). It is therefore clear that if major alternations such as the Vowel Shift are dealt with, other details can be safely ignored. Nue Speling does this admirably.

(5) Stress is for the most part predictable, according to Chomsky and Halle. But the rule they offer is so complicated that it has little relevance for a practical orthography. The 1968 version of their rule is known also to be inadequate. However, we can abstract from it a general rule that will cover most cases and then mark the exceptions, say, with a <´> over the stressed vowel.

Altho most spelling reformers in English have ignored the question of stress - largely, I suppose, because it would introduce a diacritic - nevertheless, I believe irregular stress should be shown. Stress is probably the single most unstable element in English pronunciation and traditional stress patterns have been lost in the last few generations or are changing significantly at present simply because of a failure of the standard orthography to indicate it (cf. the pronunciations of éxquísite, hóspítable, lábóratory, and at least in the United States háráss, éfflúent, etc., where the diacritics indicate alternate stresses). In Spanish or Greek, where stress is consistently marked, no comparable variation occurs.

The rule I suggest for regular stress is an abbreviated version of the Chomky-Halle rule:

Stress occurs as far from the end of a word as possible, up to the third syllable (the ante-penult), unless one of the syllables contains a 'long' vowel; then, the first long vowel from the end is stressed.

Hence, words like history-historical-historicity, general, optimal, element, spelling, alive, linguist, obtain would receive no stress mark because their stress is regular. But phonétics, segméntal, succéss, éxercise would be marked. Monosyllabic words would normally not be marked (except perhaps to make differences between homonyms, e.g. wún 'one' vs. wun 'won').

(6) Altho it would be desirable for an orthography to be devised so that a speaker of the language would find a general one-to-one correlation between his own pronunciation and the spelling, for various reasons this is seldom completely practical. However, a speaker should be able to deduce an appropriate pronunciation for his own dialect from the spelling. In short a spelling may show more differences than are phonemic for any one dialect. It should never show fewer.

The recent proposals by the SSS to reduce phonemic contrasts that are almost universally observed (e.g. between cam and calm, fool and full) or very often kept (whine-wine) should be abandoned. Furthermore, the universally observed distinction between the initial consonants in thy-thigh must be shown as in classic Nue Speling, with <dh> vs. <th>.

In point of fact some other distinctions should be introduced into Nue Speling which have not been observed before.

One of these is the distinction generally maintained thruout Scotland in the traditional form of Standard English spoken there between the vowels of fir-fern-fur. Altho some Scots apparently have adopted the RP merger of all these vowels into one, most speakers seem to have preserved the distinction. Furthermore, if these distinctions are not recognized, Scots nationalists might well reject Nue Speling altogether as a form of English imperialism. So for both linguistic and nationalistic reasons, these distinctions must be recognized. By the way, they are observed in Sir James Pitman's i.t.a. and seem to cause little trouble for students who do not make them. The OED recognizes two distinctions, lumping <ir> and <er> together. In Braeking dhe spel: an apeel to komon sens (an SSS publication from 1942), a two-way distinction is at least tolerated (see p.39).

A much more complicated issue involves the vowel of the final syllable of history, every, etc. Spelling reformers and orthoëpists have almost always regarded this vowel as short and as phonemically the same vowel as in pit. This was true even in the United States where dictionary makers always indicated the pronunciation of such Ys as a short I. That is, until 1961 Webster's Third New International Dictionary was published, which listed as the most common pronunciation one equivalent to the <ee> in see (in Jonesian transcription /i:/, in American tradition /iy/ or IPA /ij/).

Since that time American dictionaries recognize the <ee> pronunciation as the preferred one.

The linguistic reality is in fact reflected in Webster's Third, and it is shocking that it took so long for dictionaries to come to terms with it. It is also ironic that when the British Simplified Spelling Society and the American Simplified Spelling Association decided in 1956 to compromise and reach unanimity in their decisions, the Simplified Spelling Association insisted on showing the <-y> ending and derived forms always with a short <-i>.

This dialect difference between RP and perhaps most dialects of English outside of North America, and dialects of the General American type poses more of a difficulty than is usually understood. In General American the words candid and candied are different. (A pronunciation of <-y> and related forms with /i/ also occurs in America but tends to be very regional and has little prestige; it is used by some Rock singers in imitation of Black English dialects.) Following the general rule implicitly suggested earlier that when dialects differ, the maximally differentiated dialect should serve as the basis for the written norm, we might assume that the American pronunciations should be taken as the norm: we would thus get kandid 'candid', kándeed 'candied', and kandee, 'candy'.

But the short /i/ speakers make another difference: whereas the <ee> speakers never have a contrast between unstressed <ee> and <i> finally, the short /i/ speakers have. In RP, for example, where unstressed <ee> is unusual and /i/ tends to be ousting it (e.g. words from Italian with a written plural in <-i> are now pronounced either with <ee> or /i/, with /i/ forms perhaps now more common: libretti, banditti), nevertheless an <ee>-/i/ distinction is maintained in final unstressed syllables. A pronunciation with <ee> is the norm in words derived from Latin written with <-ae>: alumnae, amoebae, antennae, formulae, larvae, nebulae. The singular stele ends with /i/, the plural stelae with /i:/. Many words recently borrowed from French and some other foreign languages also have final <-ee>: causerie, chassis, debris, gaucherie, jalousie, lingerie, précis, spahi (these words are sometimes pronounced quite differently in the United States: debris is usually stressed on the last syllable, and lingerie usually - to the consternation of purists - rimes with day). Note that whereas in American English mammy and mammae 'mammary glands' are pronounced identically (unless a restored Latin pronunciation is attempted so that the <-ae> is pronounced like the <ai> in aisle), in the rest of the English speaking world they are generally differentiated as mami vs mámee. The usual American pronunciation of the song made famous by Al Jolson containing the words 'my little mammy' would perhaps be misconstrued by an Englishman as at once a lament by a preoperative transsexual and a plea for more silicone.

In this instance an out-and-out compromise must be made. The problem is not final <-y> since even if written with <-i> speakers of General American could simply interpret it as a variant of <-ee>. The problem occurs when the sound is not final, usually before an inflectional ending, where <i> would not do.

One solution has been suggested by Martin Joos: write what in General American is <ee> as <i> in unstressed syllables, and /i/ as <e>. This means kanded for 'candid' and kandid for 'candied'. The major drawback is that the ending <-ing> would have to be written as <-eng>: singeng, etc.

A better solution has been proposed by Edward Rondthaler who recommends that words with General American unstressed <ee> be written with Y both finally and elsewhere: using other Nue Speling conventions we would get kandy 'candy', kandyd 'candied', kandyz 'candies'. This solution in fact simply extends present written usage: altho normally Y is changed to I before inflectional endings, it is not before the possessive <'s> nor with proper names, e.g., Mary's, the two Germanys. (Note also that the plural of fly 'carriage' can be flys; the spelling flies refers to insects.) All we have to do is drop the change-the-Y-to-I rule. The major drawback with this is that it gives two values to Y, one of which is not in accordance with international conventions.

Let me here suggest a third solution, which may be the most far out but which has several advantages from a linguistic point of view.

The main problem with writing <-i>, as I've said before, is before an inflectional ending. If we could overcome that problem we would have the best solution. In the first place, we would want to have the vowel count as short (no matter how it's actually pronounced) so that the stress rule given earlier would not treat history, etc., as irregular (writing <i> would indicate that stress is indeed regular). Secondly, <i> would be in keeping with international conventions, whereas <y> would not, and in particular spellings like <-ny> (as in money, honey, pony) would suggest a pronunciation as in canyon or the Ñ in mañana.

We could accomplish our goal if we had a device that could indicate that the inflectional endings somehow didn't count as a full-fledged part of the word and were quasi-autonomous. In part the apostrophe used in possessives does this quite well: lady's, etc., which could quite adequately be written as laedi'z. Unfortunately we cannot use the apostrophe to indicate the plural as well (altho Sinclair Eustace was surely on an interesting track when he suggested that it be used in a whole range of phenomena he called junction). If we had yet another 'boundary marker' we would do quite nicely. The 1956 version of Nue Speling used a raised dot or full stop (period) to indicate the separateness of vowel sequences (for the earlier hyphen): medi.eeval, poe.em, kwie.et. I suggest that this be used with the plural and other inflectional endings. Thus, the plural of lady would be laedi.z. [In Webster's Dictionary, compiled by John Gage, 1981 (Baltimore, MD: Ottenbeimer), the final <-y> is always indicated in the pronunciation respelling by the compromise /i·/.]

If we generalize this not only to plurals but to regular past inflections in verbs, we get an unexpected and highly desirable reward. One of the distinctive characteristics of English spoken in Scotland is that traditionally long vowels (in the technical phonetic sense) are shortened when a consonant follows. This means that the vowel in week is as short as the one in wick and the two have to be represented phonemically by separate letters, and not as being in the relationship found in other dialects, e.g. RP /i:/-/i/ in Daniel Jones's conventions.

However, before the past inflection <-d> the traditionally long vowels are long. Thus,

/rod/
/brud/
/rud/
/grid/
'road' but
'brood'
'rude, rood'
'greed'
/ro:d/
/bru:d/
/ru:d/
/ə'gri:d/
'rowed'
'brewed'
'rued'
'agreed'


Similarly, the vowel in side is /ə+i/ but that in sigh is /ai/, and the past tense form keeps the /ai/:

/səid/'side'/said/ 'sighed'.

In short in the standard English of Scotland (as described by William Grant in The Pronunciation of English in Scotland, 1970, College Park, Maryland: McGrath), vowels sometimes differ before inflectional endings and create contrasts not otherwise found. If we extended the use of the stop <.> to the past tense marker, we could deal with these contrasts perfectly.

In at least some dialects of American and Australian English, vowels are also sometimes lengthened before the past. So side and signed contrast but in a way different from the Scots forms; the realizations are /said/ and /sa:id/, respectively. In other dialects, there is a short and long form for /oi/, as in /ə'void/ 'avoid', /to:id/ 'toyed'. A true minimal pair in my own speech is /toid/ (pseudo-Brooklynese pronunciation of 'third', which is more accurately /təid/) and /to:id/ 'toyed'.

Similar phenomena have not been mentioned for RP altho Daniel Jones does talk about certain other morpheme boundary lengthenings. For example, he notes that the /ai/ in bi-plane is longer than in pipeline. Furthermore, he indicates that the word highly has a long /ai/ when used to mean 'in a high manner' (i.e. 'she praised him highly'), but a short one in the meaning of 'very' or 'very much '('he was highly praised'), and also in the name Haile (as in Haile Selassie). (In the fourteenth edition of Jones's English Pronouncing Dictionary (1977) revised by the late A. C. Gimson, indications of such length are dropped altogether because they are 'generally evident from the orthography and from the meaningful segmentation (morpheme boundaries) of the word' (p.xiv) - an unfortunate decision based on a circular argument.)

Let us assume that such distinctions do not occur in RP and a great many other dialects, before inflectional endings. Nevertheless, if we choose to mark such boundaries, this would be no great burden for speakers of any dialect since part of the linguistic competence of all English speakers is to recognize such boundaries for grammatical reasons if not because of pronunciation.

Furthermore, to introduce such a distinction would add considerable clarity to the reformed system since forms traditionally written separately would be distinguished: daze-days, seize-seas/sees, size-sighs, nose-knows, use-ewes; wade-weighed, tide-tied, road-rowed, etc. (in modified Nue Speling: daez-dae.z, seez-see.z, siez-sie.z, noez-noe.z, uez-ue.z, waed-wae.d, tied-tie.d, roed- roe.d, respectively).

As a matter of fact this convention of <-.s> and <-.d> suggests even further that a single regular ending could be adopted for all inflections involved. We find something comparable in traditional orthography where the possessive is normally written <'s> even tho there are 3 variants (all of them shown, unfortunately, in Nue Speling: kat's, dogz, hors'ez). The traditional way is better and modified Nue Speling forms should be kat'z, dog'z, hors'z. But with <.> we can also drop the <-e> in plurals, third-person present tense forms, and regular past inflections. Thus:

plural:
present:
past:
kat.z
hop.z
hop.d
dog.z
rob.z
rob.d
hors.z
wish.z
wish.d

hunt.z
hunt.d

All in all, classic Nue Speling (the 1948 variety) tends to meet the general criteria Chomsky has proposed for a desirable orthography for English. With a few modifications, Nue Speling can counter most of the remaining objections Chomskyites have raised against spelling reform.

Altho English alone has been considered here, it should be clear that similar considerations could be used in the creation of rational orthographies for any and all languages in the world.

Back to the top.