The Design of DRE

(version 2.3, current as of February 13, 2005)

The Goals of DRE

DRE (Diacritically Regularized English) is an improved spelling system for English.  It borrows heavily from two other spelling systems: Axel Wijk's Regularized English, and Bob Boden's SRS.  The concept of using diacritics to clarify English pronunciation comes from SRS; much of the respelling methodology comes from Regularized English.  Additionally, DRE draws heavily from Wijk's research into English spelling and phonology presented in his book on Regularized English.

DRE is intended to meet certain goals.  In brief, the goals are as follows:

  1. DRE is based on pronunciation rules which are, by and large, the same as those of current English spelling.  It uses diacritical marks to clarify pronunciation.  When a word can be marked with diacritics to accurately indicate its pronunciation, its spelling is not otherwise changed.
  2. When the pronunciation of a word cannot be indicated by the mere addition of diacritics, DRE calls for the word to be respelled.  Respellings are chosen to be accurate, natural, patterned, and to resemble the traditional spelling as much as possible.  The DRE pronunciation rules have been chosen to minimize the necessity of respelling.
  3. DRE tries to avoid innovative spellings with no precedent in traditional spelling.  Where such innovations are necessary, they are used in spelling common words, so that they will become familiar and recognizable to readers.
  4. When two familiar words have the same pronunciation but different traditional spellings, DRE continues to spell them distinctly.  It does not merge homophones.
  5. DRE has been designed to be readable and to improve on traditional spelling even if all diacritics are omitted.
  6. DRE is intended for use with American English.  But, subject to my limitations, it takes British English into account.  When American and British pronunciation differ, whenever possible, the corresponding spellings differ only in the diacritics used.  I am open to changes in DRE to allow it to better represent British English.
  7. Though not an original goal of DRE, in practice it has the desirable property of tolerating pronunciation variances, which is to say that frequently alternate pronunciations for a word have spellings which differ only in the diacritics.
  8. Though not an original goal of DRE, in practice it elucidates word relationships.  The common portions of related words often differ only in the diacritics, even when pronunciation differs.  This is a desirable property of traditional spelling which DRE preserves.

Certain goals of many alternate English spelling systems are explicitly not goals of DRE.

  1. DRE is not phonemic.  It is neither one spelling per sound nor one sound per spelling.
  2. DRE is not logical.  It is, however, somewhat more logical than traditional spelling.
  3. DRE is not simple.  It is, however, highly patterned, and I believe the patterns to be easy to grasp.
  4. DRE is not economical.  If a DRE respelling happens to be longer than the current spelling, so what?
  5. DRE is not uncompromising.  The goals of DRE do not always lead to clearcut solutions, and much give-and-take has been required to balance them successfully.

The DRE "Dictionary"

I have assembled a list of moderately common words and their DRE spellings.  The spellings are based on American pronunciation, as described by several American dictionaries, with input on British pronunciation from the Longman Pronunciation Dictionary.  I refer to DRE as defined by this dictionary as the "Tentative American Reference for DRE", reduced to the acronym TARDRE.  TARDRE is emphatically not intended to represent my own pronunciation.  Further, the existence of the dictionary should not inhibit anyone from using DRE to represent their own pronunciation whenever that pronunciation is felt to be more correct or representative than my "official" one.

Presently, DRE is nothing more than an amusement of the author.  However, I am pleased with it, and am gradually coming to think it might actually have some practical value or merit.  DRE was designed to accommodate variance in pronunciation, and those with interest in it are free to use it as seems right to them.  If, as seems most unlikely, DRE were to see widespread use, I would have little control over its evolution, and one expects that the TARDRE dictionary would be superseded by one or more professional dictionaries of larger scope.  Americans do love their rules, and I anticipate that there will always be books defining right and wrong ways to spell despite, or perhaps because of, the variety of accents in which American English is spoken.

I do urge users of DRE to start by adhering to the spellings in the DRE dictionary until they are fluent enough with the way the system works to become part of its evolution themselves.  The spellings in this document are all TARDRE spellings, except where alternate design choices or British spellings are shown.

More information on the DRE dictionary may be found here.

Overview of DRE

Given the size of this portion of the document, you may well question whether it is in fact an "overview".  Nevertheless, while the broad outlines of DRE are presented here in some detail, there are some aspects which are omitted or barely touched upon.  The actual definition of DRE is in this document; while it is considerably terser than this presentation, it is the final authority for the rules of DRE.

a.  Vowel sounds and diacritics.

One of the characteristics of English spelling is the association of multiple vowel sounds with the same letter.  DRE, to some extent, organizes the chaos with the use of diacritics.

In DRE, the unadorned form of each vowel represents its traditional short sound, or the schwa: hat, banana, bed, rivet, sin, devil, hot, lemon, bug, album, myth, analyst.  Additionally, unaccented y is used at the end of words for an unstressed long e sound, as in: silly

When a vowel is used with an acute accent, it indicates the regular long sound of the vowel: báby, léthal, tíger, tótal, fúneral, pýthon
The letter e is normally silent if unmarked at the end a word.  As in traditional spelling, silent e is used to indicate the previous vowel is long, or to give a preceding c the sound of s, as in: ráte, sauce.  In strict DRE, the presence of the silent e does not allow the acute accent of the preceding vowel to be omitted.

When a vowel (other than y) is used with a grave accent, it indicates a different free vowel sound associated with the vowel: fàther, sèànce, rádìó, dòg, trùth.  The case of ù requires a bit of explanation.  Many words, such as <duty>, are pronounced by some speakers with the ú sound (with an implied y) and by others with the ù sound (without one).  The ú spelling is used for such words if the pronunciation with y is considered the primary spelling for either British or American English.  Thus, <duty> is spelled dúty, but <lute> is spelled lùte, since the /lju:t/ pronunciation is not the most common for either variety of English.  (See Design Issues below for an extended discussion of the ú/ù dichotomy.)

When certain vowels are used with a circumflex accent, it indicates either a specific unusual short sound, or a schwa.  The circumflex form is often used to indicate a schwa in situations where otherwise the vowel would be assumed to be stressed.  Examples:  âny, furnâce, prêtty, kíndnêss, ôven, purpôse, pûsh.

The vowel ë indicates the regular short e sound or a schwa.  It is used most commonly in vowel combinations.  Examples: belovëd, dúët.  The vowel ü indicates a shortened long u, as in linóléüm, regülar or ampütáte.  The vowel ï is used in words such as aprécïáte; this rather special usage is described later in the discussion of -ci- in part (e) below.

The interpretation of many of the above vowels changes when they are followed by the letter r.  An unaccented vowel followed by a single r has one of the two vowel sounds of the word <murder>, depending on stress.  (These two sounds seem pretty much the same to Americans.)  Examples: saròng, permit, dirt, actor, burden.  "yr" is an exception, pronounced as in pyramid.  Some long vowel letters (plus one short) are altered in a regular fashion when followed by r: vèry, mìracle, wôrk, cúre, rùral.  Also, vowels preceded by a double r do not have a diacritic, but, except for u, are pronounced differently from their "single r" sound: marry, terrible, irritáte, horrid, current.  For many speakers, the sounds of "arr" and "âr" are the same.  When a word is commonly pronounced with /{r/ (as in marry), DRE spells it with "arr" when possible.

Two additional vowel symbols, ä and ö, are used in association with the letter r.  The ä appears, for instance, in the word polärity, and the ö appears in the British spelling of förest.  These letters do not occur all that often, and the reasons for their existence are somewhat obscure.  They are discussed further under Design Issues.

The above rules are perhaps more complicated than one would like.  But they have two important characteristics: they allow most English words to escape respelling, and they use the accent marks in a way that is systematic and easy to master.

b.  Vowel digraphs

Like traditional English spelling, DRE augments its vowel repertoire by using two letters to indicate a single sound.  Some digraphs have more than one interpretation, distinguished by the use of diacritics.  This part of the overview will not attempt to describe all the digraphs in detail, but rather explain the principles behind them.

The digraph "oo" is rather special.  In essence, it is treated as a sixth short vowel.  The long form of "oo" is "oó"; in contrast to most vowel digraphs using an accent, the sound is unrelated to the component "ó".  Examples: hook, boót.

All the digraph sounds of DRE can be spelled equally well with a single, possibly accented, vowel.  But without the use of digraphs, it would be necessary to respell a huge number of words, and it would be impossible to distinguish many homophones.

Most digraphs without accents are pronounced in a way you would expect from traditional spelling: hurrah, bait, play, maul, flaw, seat, feed, vein, they, sleuth, crew, brief, coat, toe, toil, boy, house, cow.  As with the single vowels, these digraphs may be shortened in sound when followed by an r: fair, dear, beer, their, pier, soar.  In general, i and y have the same effect as the second element of a digraph, as do u and w.  That is, "ai" and "ay" have the same sound, as do "ou" and "ow".

Most digraphs with accents are pronounced the same as the accented letter, as in: bazaàr, antennaé, samuraí, kaýak, heàrt, matinèe, steín, eýe, kéy, líe, bròad, sóul, lów, còurt, soùp, cúe, trùe, guíde, buý, dýe.  The digraphs éu and éw are used like ú, as in éuréka, féw.  The accent is placed over the e because of the oddity (and unavailability in computer fonts) of an accented w.  (Unfortunately, the requirement to accent the e rather than the w leads to trouble for a number of words containing an unintentional éw digraph, such as préwoshed and réwínd.  This ambiguity seems unvaoidable.)

One might question why diacritics are not used in some of the unaccented digraphs.  áy, ée, èi, ìe and óa, among others, would parallel the digraphs of the previous paragraph nicely.  The reason is to facilitate spelling vowel combinations.  If the word <seen> is written as séen, how then are we supposed to spell préeminent?  Besides, since the use of digraphs with English is unusual, it is probably best to minimize it.  Since we all know how to pronounce ee now, without any accents, why make things complicated?

All the above digraphs occur in stressed syllables.  DRE also has two vowel digraphs used in unstressed syllables, and pronounced as the schwa.  These are âi, as in mountâin, and ôu, as in pòrôus.  The digraph eû, used before r in both stressed and unstressed syllables, as in saboteûr, is also an exception.

Finally, DRE has two innovative vowel digraphs, which are not used in traditional spelling: aa (saamon) and oû (shoûd).  I attempt to justify these constructions in later sections.

c.  Consonants

DRE's handling of consonant spellings is simpler than its handling of vowels because there are no consonant diacritics.  As with vowels, consonant letters may be combined into digraphs with a different pronunciation than their constituents.  This section discusses the single consonants, while the next discusses the digraphs.

DRE, like traditional English spelling, allows consonants to be doubled.  With the exceptions of -cc- and -rr-, doubling a consonant implies no change to the sound of the consonant.  Use of a double consonant implies the preceding vowel is stressed, except where the circumflex diacritic is used.  DRE generally undoubles a consonant in words where this pattern is not followed - see Double Consonants below.

DRE allows silent consonants only in a few specific cases, notably initial kn-, ps- and wr-.  These exceptions are made primarily to facilitate the preservation of different spellings for homophones.

While the use of most consonants in DRE is unsurprising, a few require special commentary.

d.  Consonant digraphs and trigraphs

In addition to the i/j combinations discussed in the previous section, DRE uses the following consonant digraphs and trigraphs:  ch, ck, cq, cqu, dj, jh, kh, ng, ph, qu, sc, sh, tch, th, tz, wh, xc.

Most of these are familiar combinations, used in DRE exactly as in traditional spelling.  A few of them call for additional commentary.

e.  Aggravátion and tòrtjur

One of the more illogical conventions of written English is the change in the pronunciation of c, s and t (and occasionally x) before a (usually) unstressed syllable beginning with i or u (or occasionally e).  In the case of i (or e), the vowel is usually not pronounced; in the case of u, it is.  Examples (traditionally spelled): vicious, ocean, conscious, appreciate, vision, measure, passion, pressure, donation, question, ratio, righteous, torture, noxious, luxury. 

DRE classifies such words into four sorts, and applies a consistent spelling to each.  The first sort is composed of those words in which the consonant is followed by an i which is not pronounced (leaving out words in which ti is pronounced as ch).  In this case, the English spelling of this part of the word is retained, except that s is replaced by z if the sound is voiced:  viciôus, consciôus, vizion, passion, dónátion, noxiôus.  The second sort is composed of words in which the consonant is followed by a u, or when ti is pronounced as ch.  In this case, the letter j is inserted before the u (or replaces the i in ti), again with s changing to z if the sound is voiced, or with the j replacing the second s when s is doubled: mezjur, presjur, questjon, tòrtjur, luxjùry.  The third sort contains words where the following vowel is an e.  These words are respelled as if the vowel were an i: ócian, ríetjôus.  Finally, the fourth sort is composed of words where the vowel is an i pronounced as long e (ì).  In these words, the i is spelled with an umlaut: aprécïáte, rátïó.

Note that the same process of sound mutation that led to these spellings also led to words in which du is pronounced as ju, such as <procedure> and <gradual>.  These words are respelled with dj rather than simply j to emphasize the parallelism with the above spellings:  procédjur, gradjùal.

Some of the reasons for these design choices are as follows.  The system above is more precise than traditional spelling about indicating the pronunciation.  (The only ambiguity is really the handling of words like <mention>, which some speakers pronounce with sh, and some with ch.  DRE chooses the sh spelling in such cases, to avoid respelling.)  While the use of the letter j is an innovation, its resemblance to the letter i makes it less jarring than one might anticipate.  The fact that the normal j sound is a voiced ch also helps to make the notation seem natural.

Finally, the ï symbol was introduced to improve parallelism between related words, as with aprécïáte, aprécïátion and apréciativ.  A previous version of DRE wrote the first two words above as aprécjìáte and aprécjìátion which, while consistent, obscured the relationship with apréciativ.  Note that the existence of words like asócìátion and pronuncìátion, where c continues to represent the s sound, precluded the use of ì in the role of ï.

f.  Silent e

Like traditional spelling, DRE makes heavy use of the silent e, to indicate that the preceding vowel is long, to soften a c, or to distinguish a final s sound from a plural marker.  Purists may protest that when long vowels are marked, as they are with DRE, the silent e is unnecessary, and could be dropped.  DRE does not do this, for two reasons.  One is that very many words would be respelled as the result of such a change.  The other is that DRE is intended to remain reasonably readable even when all the diacritics are removed.  After removal of the DRE diacritics, the silent e is vital to distinguish plan from plane, bit from bite and slop from slope.

Traditional spelling does contain, however, a number of words where a silent e is used even though the previous vowel is not long, such as valve, some, college, imagine, accurate, native, culture.  DRE drops such e's, except after c, s, or r (other than in -ure pronounced as -oor).  Thus, the DRE spellings valv, sôm, collej, imajin, accürat, nátiv, cultjur.  When a silent e is used after a short vowel and a single c or s, the vowel is marked with a circumflex unless it is i or u, in which case the e is dropped (and c changed to s).  If more than one consonant separates the short vowel and -ce or -se, the e is retained.  Thus: menâce, penance, malis, prince, purpôse, response, lettus, repulse.  See "Ending -s, -ce, -se and -ss" below for more discussion of this logic.

g.  Double consonants

One of the features of traditional spelling which makes using it correctly so difficult is the inconsistent use of double letters.  Consider pairs like addition/edition, balloon/ballot, cannibal/animal.  There is no order.  Often, a double consonant serves to indicate that the syllable is stressed, but counterexamples abound:  announce, cassette, offensive, satellite, dandruff.

It is way more than DRE can do to repair this inconsistency in English spelling.  It has been suggested that consonants should be doubled exactly when the syllable contains a stressed short vowel sound, but this requires respelling many familiar words.  It is also hard to double digraphs in words like mythical or singular without having the spellings look silly.  And finally, using consonant doubling to show vowel shortness leads to undesirable spelling differences between related words: add/adition, astronnomy/asstronommical, capptiv/captivvity, etc.

DRE does ameliorate the double consonant problem in one way.  If a double consonant is used following a schwa, the double consonant is replaced by a single consonant, thus: anounce, casett, ofensiv, satelíte, dandruf.  There are two exceptions to this.  First, the doubling of consonants in closely related words is kept consistent, so that âddition continues to be spelled with a double d to preserve the relationship with add, and the second m in <commendation> is removed to preserve the relationship with comend.

Despite the fact that DRE limits itself to a timid, partial solution to this problem, this particular change is one of those with a large impact on the overall appearance of transcribed text.

h.  DRE Modes

DRE can be written in one of three modes.  Up to this point, I have been discussing strict DRE, which is to say, DRE with all diacritics used, even for extremely common words or when the pronunciation is completely obvious to the reader familiar with English.  There are two other modes of DRE, which may be more practical.  Stripped DRE is DRE using no diacritics at all.  One of the goals of DRE is for stripped DRE to be readable, no more ambiguous than traditional spelling, and generally superior to it.  Intermediate to strict DRE and stripped DRE is reduced DRE.  This is DRE with many diacritics removed.  Reduced DRE removes diacritics from (1) extremely common words, (2) extremely common suffixes and (3) long vowels with an acute accent preceding a single consonant and a silent e.  In practice, this reduces the number of diacritics substantially.

Here is an example text, the preamble to the U.S. Constitution, in strict, reduced and stripped DRE to illustrate the differences:


Wé, the peaple ov the Úníted Státes, in òrder tu fòrm a mòre perfect únyon, establish justis, insjùre domestic tranquility, províde fòr the common defense, promòte the jeneral welfâre, and secúre the blessings ov liberty tu ourselvs and our postèrity, dù òrdain and establish this Constitútion fòr the Úníted Státes ov Amèrica.


We, the peaple ov the Únited States, in òrder tu fòrm a more perfect únyon, establish justis, insjùre domestic tranquility, provide for the common defense, promote the jeneral welfâre, and secure the blessings ov liberty tu ourselvs and our postèrity, du òrdain and establish this Constitútion for the Únited States ov Amèrica.


We, the peaple ov the United States, in order tu form a more perfect unyon, establish justis, insjure domestic tranquility, provide for the common defense, promote the jeneral welfare, and secure the blessings ov liberty tu ourselvs and our posterity, du ordain and establish this Constitution for the United States ov America.

i.  Respelling Principles

When one transcribes a new word into DRE, one or two steps are required.  First, one must determine if the word can be marked with accents to indicate the correct pronunciation.  This step is simple and mechanical.  If the word cannot be so marked, it must be respelled.  Respelling is an intuitive process, not readily reduced to rules.  This section offers some of the guidelines I've developed as I work with the system to choose the best respelling for a variety of words.

The following are the basic requirements for a respelling:

Beyond that, there are the following guidelines:

One final principle is that, while it is sometimes necessary to think "out of the box", it is also usually best to have a prosaic spelling rather than an innovative one, if either will get the job done.

These guidelines help make the appearance of DRE natural and self-consistent, even when many words have to be respelled.

Here are some examples of application of these guidelines.

DRE Innovations

Mostly, DRE respells words so that they resemble the regular spellings of similar-sounding words.  Nevertheless, DRE introduces a certain number of novel spellings without any traditional precedent.  Such spellings have not been introduced casually, and, when possible, they have been introduced into the spellings of common words, so that readers will become accustomed to seeing them.

Some of these innovations have already been explained and justified, and are only briefly mentioned.  The remainder are discussed in more detail.

a.  j as a modifier letter

The digraphs sj, tj, xj and zj are used in words where the consonant preceding j is followed by a (usually) unstressed syllable, and the normal sound of the consonant has been softened:  insjùre, nátjur, Cristjan, luxjùry, plezjur.  This change is discussed in detail above.

b.  jh

The digraph jh is used for the French -ge sound, as in garàjh, neglijhèe and jhonrë.  Other options for this sound were zh or gj, but I feel that jh is the best guide to the pronunciation.

c.  kh

The digraph kh is used for the k sound in place of ch in scientific and technical words.  This is justified in more detail above.  Note that, traditionally, kh is often used to represent the ch sound of <loch> or <Bach>.  DRE uses the spelling qh in the few English words where this sound occurs, e.g., loqh.

d.  aa

The digraph aa is used in a number of words for the sound of short a, mostly in cases where other respellings are less than satisfactory.  Some words using this digraph are caash (cache), daam (damn), laam (lamb), plaad, saamon and, in American English, haaf, haav (halve), laaf and moraal.  (In British English, these would be spelled haàf, haàv, laàf and moraàl.)  In its rare uses in traditional spelling, aa represents the broad a sound, becoming aà in DRE (e.g., bazaàr).  To my knowledge, aa representing short a occurs in a single word, <baa>, and there in American English only.  See below for a justification of this unorthodox spelling.

e.  Bare q

One of the minor peculiarities of DRE spelling is its fairly frequent use of the letter q without a u, as in conqer, etiqet, liqor, mosqìtó, plaq, qìshe, únìqe, and so on.  These spellings come from dropping the u from qu when it is not pronounced.  These words could generally be spelled with k instead, but the q establishes a link with traditional spelling.  Besides, I think English should be fun, and I like the air of playfulness that these spellings add.

Defects of DRE

It is worth taking a few paragraphs to admit that DRE is not perfect, and does not make English spelling easy or trouble-free.  This section discusses seven particular criticisms that might be made of DRE.

a.  DRE is non-phonemic

Ideally, a spelling system will be phonemic, which is to say that the pronunciation of a word is completely predictable based on its spelling.  DRE fails to achieve this, for reasons both great and small.  The key reasons that DRE fails to be phonemic are:

See below for further discussion of these points.


b.  DRE is too timid

Quite apart from its tolerance of silent e, and its half-hearted stance on double consonants, DRE fails to even address some of the most annoying problems of English spelling.  For instance, the fact that some English words ending in o have an e in their plural (<heroes>) and some do not (<burros>) is quite absurd, and easy to remedy by simply picking one way or the other and sticking to it.  DRE could do this, but doesn't.  Similarly, why do we have farmer but actor, or dominant but different, or changeable but convertible?  What a waste of good brain cells to remember all this, and it would be so easy to fix!

This is hard to answer.  As the objection notes, it would be easy for DRE to just pick one form or another.  The reason may in fact just be timidity.  I feel like I have my hands full with DRE's changes already, and trying to fix all the irregular inflections and suffixes of English seems a bit much to ask.  However aggravating it is to have to consult a dictionary whenever I want to use an -able/ible word, I also doubt that a large percentage of the spelling errors that mar English writing are of this sort.  See below for further discussion of these difficulties.

c.  DRE is too different from traditional spelling

This is the other side of the coin from the previous complaint - DRE is a failure because it is too unlike traditional spelling.  One way in which DRE is unlike traditional spelling is all those diacritics, but this can be remedied by using stripped DRE, at the cost of much of its precision.

Alternately, the complaint may be that while one is perfectly happy with diacritics, too much respelling is required anyway.  Apart from the diacritics, there are four predominant reasons that words are respelled in DRE:  unnecessary silent e's, unnecesssarily doubled consonants, replacement of g by j, and replacement of s by z.  There is nothing to be done about the silent e's, in my opinion - these spellings need to be fixed.  The unnecessary consonant doublings could be retained, as DRE's efforts in this direction are not nearly enough to solve the problem.  It is interesting to note that the last two reasons could be dealt with, for those not offended by diacritics, by adding two consonant diacritics to DRE, say ĝ, pronounced as j (/dZ/), and ŝ, pronounced as z.  Regrettably, these letters are not present in many computer codes and fonts, making their use somewhat difficult.

d.  DRE is too much like traditional spelling

This is the the argument that, while resemblance to current English spelling is desirable, DRE takes it too far.  The main issue here is the retention by DRE of seldom-occurring or foreign letter combinations.  Why shouldn't the DRE words bròad, shófeûr and matinèe instead be braud, shófer and matinay?  DRE is not doing spellers any favors with spellings such as these.

This is indeed a quite valid argument.  Rather than attempt to refute it, I will just say that, in part, DRE was an experiment to determine how close a better spelling system could remain to current spelling, and the answer is "very close".  It is quite possibly that some of DRE's slavish fidelity to traditional spelling is counterproductive.  If DRE were ever to gather enough interest to evolve, this would be a likely area for change.

e.  DRE is too arbitrary

This one can be summed up as:  There is no way to systematically spell DRE.  Unless you know traditional spelling, there are no rules to guide one to the spelling to be used for each sound.  And even if you do, the respelling system is too arbitrary to allow one to predict accurately how a word will be respelled.  There's really no way around it - this complaint is quite valid.

In addition to the basic arbitrariness of allowing many spellings for some sounds, there are other factors which make DRE even more arbitrary.

All these problems are essentially inherited from traditional spelling, and they are worse there.  We learn traditional spelling through an arduous process that combines pattern recognition with memorization of exceptions.  Because DRE refines traditional spelling, it would have to be mastered by those new to English spelling by the same kind of extended process, learning to recognize patterns and memorizing exceptions.  I believe that in DRE the patterns are more visible and the exceptions fewer, and that therefore the process of mastering DRE spelling, though still non-trivial, should be somewhat easier than for traditional spelling.  And I also believe that those already familiar with English spelling will probably find the patterns of DRE respelling can be mastered in a matter of days or weeks, not years.

It may not be heaven, but at least it's a much cooler section of hell.

f.  DRE is too complicated

This argument would claim that, even though DRE is systematic, it is too complicated.  The number of diacritics and digraphs is too large to be easily mastered.  In particular, the umlauted vowels, which are only used occasionally, and which are not defined in a particularly natural or consistent way, are a strong indication that the whole thing is too baroque.

Again, this is a hard argument to refute.  I don't find DRE too complicated to use and remember.  But I must admit, I've been working with it for quite some time now, and I also have a knack for mastering complicated systems of rules.  I have some doubts that the goals can be achieved via anything less complicated, but this could be taken as evidence that they were never the right goals to begin with.

I will point out that Vietnamese orthography uses more diacritics than DRE, and people somehow master it.  I will also admit that the tonal nature of the Vietnamese language presents special problems in use of the Latin alphabet which English is spared, and that there is nothing about the English language which would force a completely rational Latin-based spelling system to contain such a plethora of markings.

g.  DRE uses too many diacritics

This argument is generally made in the context of reduced DRE.  Reduced DRE allows diacritics to be left out in certain cases where the pronunciation is obvious to those familiar with English.  But even in reduced DRE, many obvious diacritics are required: càr, pòrk, fáble, enuncìate, obvìous.  DRE would be more accessible if its use of diacritics were limited to words where they were truly necessary.

I consider this argument to be the converse of argument f above:  DRE is not complicated enough.  I personally have made my peace with diacritics, and no longer look at them as unsightly typographic blemishes.  It is true that more rules would allow more diacritics to be eliminated, but I don't feel the cost in complexity is worth it.  If one holds vowels with diacritics to be first-class alphabetic citizens, then why should we go out of our way to shun them?  The present rules for reduced DRE are both extremely simple and quite effective.  Is it not better to use diacritics fairly often than to have to think really hard about whether each one is needed before writing it?

DRE Design Issues

a.  long u

DRE's handling of the uncertainty of pronunciation of "long u" is problematical.  Many English speakers pronounce <tune> as /tju:n/, others as /tu:n/.  The DRE solution is to prefer the pronunciation with /j/, whenever it is the most common pronunciation in either British or American English.  An alternative was just to allow each speaker to spell it as he says it.  After all, DRE allows Americans to dance, but Englishmen to dànce.

My reasons for designing DRE this way are:

I will end by noting that any English speaker who feels strongly about the preferability of the /j/-less pronunciation is free to write the word tùne in DRE, and no harm will come to him.

b.  The ei digraph

The standard pronunciation of ei/ey could be long a (<vein>) or long e (<seize>).  In his Regularized English, Wijk chooses to treat the long a sound for ei/ey as standard, and one sign this was a good decision is the fact that the spelling of <their> is so reasonable, while <weird> is a spelling problem.  In fact, one suspects that the whole "i before e" spelling issue stems from the fact that both ei and ie are frequently pronounced as the long e.

The trouble with this decision comes from the existence of words like <being>, <deity>, and <reinforce>.  There is little choice but to use spellings like déity for these words, which means that éi cannot be used for the plain long e.  This in turns means that either the unmarked ei must be long e after all, or that the spelling ei for long e must be abandoned altogether.  In the former case, we would have spellings like vèin and thèy.  In the latter case, all the ei words with long e have to be respelled.  One problem here is the existence of many words with very natural spellings in which ey is pronounced as long e: key, monkey, alley, honey.  If we preserve these spellings, we have to break the symmetry between ei and ey, which is preserved for all the other i/y digraphs.  Nevertheless, the final resolution is to do exactly that: long a remains the standard sound of ei, the ei words with long e are respelled using ee (seeze, receet), and ey with the sound of long e keeps the éy spelling.

There is still some feeling that the long e sound for ey is more natural than the long a.  This is one reason for the decision to respell certain common words with ey (such as greyt, weyr, eyt).  These familiar words should increase the reader's comfort level with this pronunciation for ey.

c.  Ending -s, -ce, -se and -ss

The use of s as the plural ending causes a number of problems.  Life would be so much simpler if we just used z instead!

Words ending in s, but not ss, are generally assumed to be plurals, with the s pronounced as /z/ unless preceded by a voiceless consonant.  So how should you spell words that end with /s/?  Traditional spelling sometimes uses an s anyway (alas, atlas, axis, cosmos, campus, etc.).  More often, it uses one of -ce, -se or -ss instead.  All of these workarounds present problems.  -ss tends to imply that the syllable containing it is stressed, while -ce and -se imply that the preceding vowel is long.  Sometimes, these assumptions are false: furnace, purchase, compass, princess, goodness, justice, promise, purpose, lettuce.  These spellings collide with the DRE solutions to the misuse of silent e and of double consonants.

DRE's handling of these words depends on the vowel preceding the /s/.  In the case of -ice, -ise or -uce, an ending s cannot be misunderstood as a plural, since no English nouns end in short i or short u.  Thus, we get the DRE spellings justis, promis, lettus.  We cannot use a naked -s in words with the other vowels, due to the likelihood of confusion with words such as sofas, bottles and phótós.  For these words, we use the circumflex accent over the preceding vowel, to cancel the implication of the silent e or double letter: furnâce, purchâse, compâss, princêss, goodnêss, purpôse.  This convention is also used in words where an unstressed a or o is followed by a single s pronounced as /s/, such as atlâs and thermôs.  Finally, for the few words where an ending s would be inappropriately pronounced as /z/, the spelling is changed to use one of the other endings, as in gass, cáoss and yess.

d.  Non-plurals ending in -s

The English language is blessed with several classes of words that have the form of plurals, but which are not.  Some examples are:

Should DRE spell these words with a z, to match the pronunciation and grammatical role, or with s for familiarity?  My instinct was to spell them with a z: backwardz, herz, rábiez, sêriez.  As you can see, this looks rather odd and jarring.  I eventually changed my mind, and now spell all these words with a final s.

The case of the pronouns brings up the question of whether <his> and <whose> should also be spelled with an s, even though the z spellings for these (hiz, hoóze) are not too bad.  For consistency, I am now spelling <whose> as hoós, but because of words like <this> and <tennis>, the "is" ending must be pronounced /is/, and so <his> must be spelled hiz.  There is also the question of jolting contractions such as it'z, he'z, that'z, hoó'z, etc.  In these cases, the z derives from the verbs iz and haz, which I see no reason to respell with an s.  Other irregular verb forms of the same sort, such as <was>, <does>, <says> are spelled similarly: wôz, duz, sez.  The respelling of the vowels lessens the strangeness of the z, and the form "sez" is already commonly used in slangy writing.

e.  could, would, one, done

Two sets of similarly spelled, extremely common words are quite challenging to respell in DRE: could/should/would and done/none/one.  The first set needs to be respelled due to the silent l.  The obvious respellings are cood/shood/wood, but unfortunately, wood clashes in the obvious way.  DRE also allows û for this sound, but cûd, shûd and wûd look like attempts to transcribe a backwoods dialect, and cûd is a homograph too.  My solution here was simply to drop the silent l, and write coûd, shoûd and woûd.  The resulting words are so similar to the traditional ones that there is no doubt what they are.  That the ou digraph already has several other forms (ou, óu, and où, not to mention òur and ôus) is unfortunate, but I don't feel that this usage really introduces any new problems.

The -one words are even harder to deal with.  <one> can easily be respelled as wun, but the other two are not so easy, thanks to the existence of the words don, dun and nun.  It is also desirable to spell <done> with a u vowel, to relate it to the verb (the form duen would be grammatically ideal), and equally desirable to spell <none> and <one> similarly to emphasize their related meanings.  One possibility was dône, nône and wône, making use of the property of the circumflex vowels that they are short even when they would be interpreted as long in the absence of the circumflex.  The biggest problem with this solution is that, in stripped DRE, it is hard to read wone as <one>, rather than as some unknown word pronounced /woUn/.  I attempted to solve this by inventing the digraph ôe: giving dôen, wôen and nôen.  This proved too ugly to endure.  My current solution is to spell them dunn, wunn and nunn.  This is not ideal, but seems to be less grating than any other solution I've tried.  A related issue is the spelling of <once>.  Logically speaking, it should probably be spelled wunnce, but I think wunce is adequate.

f.  ör and är

DRE's goal of simultaneously being accurate, avoiding respelling, elucidating word relationships and supporting both British and American English makes dealing with certain vowel-r combinations very challenging.  One problem stems from the fact that in British English the combination orr is pronounced with the short o sound (/Q/), while in America, it is pronounced with /O:/.  (A very small number of words behave similarly in American English, notably <borrow> (/bA:roU/) and <sorry> (/sA:ri/).)  However, some words containing only "or" are also pronounced in this way in Britain, such as <forest> and <orange>.  This situation caused me to introduce the digraph ör for this sound: förest, öranj.  Its use is limited to British English.  (The American <sorry> could be spelled sörry, but I've decided the dictionary spelling will be sorry, as this pronunciation is also used and accepted.  It didn't seem a good idea to introduce yet another letter for the sake of a few words of this sort.)

A similar problem arises in both British and American English with the /{r/ sound.  Many speakers do not have this combination, generally using /Er/ instead.  The /{r/ is most frequently spelled with the arr trigraph, but it also occurs with a single r, as in <aristocracy>.  <aristocracy> could be respelled as arristocracy, but then the close relationship to aristocrat is lost.   The solution is to introduce the är digraph as another spelling of /{r/, allowing the spellings äristocracy and polärity.  Of course, those people who, like myself, fail to distinguish /{r/ and /Er/ are left wondering what all the fuss is about.  Certainly a case could be made that, for both the /Qr/ and /{r/ sounds, a little less precision and a little more simplicity would be an improvement.

Because ä and ö have the same sound as short a and short o respectively, the analogy to the behavior of the DRE ë is quite close.  This makes the interpretation of är and ör somewhat easier than it would otherwise be.

g.  Other vowels before r

O and a are not the only vowels to be troublesome when followed by r.  When one takes the differences between British and American English into account, they are all troublesome.  By and large, however, they are not very troublesome, because current spelling, which DRE imitates, already serves both varieties of English quite well in this area.  There are a few words whose spelling does not accurately reflect the British pronunciation, which DRE respells.  These respellings are worth explaining, because they seem somewhat mysterious and pointless to Americans.

In addition to the vowel differences, British English differs from American English in being non-rhotic, that is, not pronouncing r in some contexts, as following a vowel and preceding a consonant.  This difference presents no issues for DRE.  The rules for pronunciation of r in traditional spelling can be used unchanged.

Here is a summary of the remaining differences:

As can be seen, these distinctions are quite regular, and all that is required of DRE is to deal with the occasional exception.

One other interesting development in this area should be mentioned.  According to the Longman pronunciation dictionary, for a few very common words ending in -ary or -arily, the American pronunciation of /Eri/ or /Er@li/ is becoming common in British speech.  Examples are the words <necessary> and <voluntarily>.  Ironically, this pronunciation would make the current DRE spelling inappropriate - DRE would have to use necesèry and voluntèrily to indicate the pronunciation accurately.  For the moment, I'm leaving the traditional use of a rather than e intact in these words.  If anyone cares about DRE in ten years, this will bear re-examination if this development proves to be the start of a growing trend.

h.  The aa and aà digraphs

Certain words containing a short a sound in American English require respelling due to silent letters or other defects, but cannot use an obvious respelling representing a homonym.  "cache", "damn", "halve", "lamb" and "morale" are examples of such words.  A previous version of DRE used the digraph ae in such cases: caesh, daem, haev, laem and morael.  The problem with this is that many of these words are pronounced in British English with the broad a sound, for which this spelling is inappropriate.  While the combination àe would work in the abstract, it is a combination which a person familiar with the traditional orthography would find quite surprising for this pronunciation.

When I became aware of the conflict with British English, after much consideration I changed the ae digraph to aa, allowing me the use of aà for the British form, which in turn permits the relationship between haaf and haàf to be as clear as that between bath and bàth.  It is indeed unfortunate that aa in traditional spelling is practically never pronounced as short a, but at least, for an American, the intended pronunciation of caash, haaf and moraal is pretty clear.  I chose to use aà rather than àa for the broad form by analogy with oo and oò - one should imagine the diacritic applying to the digraph as a whole.

I have chosen to respell more words than strictly necessary with aa, in order to increase the familiarity of the notation.  Words like caaf, jiraaf, laaf, laam, plaad and saamon could have had other spellings, but using aa in this way increases its visibility, so that the reader becomes more used to seeing it.  For British readers, there are only two very common words (daam and laam) that have the unmarked spelling, but I hope the relationship of these spellings to haàf and laàf will still be apparent.

Improving DRE

I have considered a number of possible improvements to DRE.  I have not adopted any of them, and probably will not, but it could be argued that they would be beneficial.  The first two sets of changes discussed below would make DRE spelling slightly more logical and precise, but I'm not convinced this is particularly important.  The third set attempts to rationalize several inconsistent English spelling patterns, which would be of significant value if successful.  In all these cases, one unfortunate effect of the proposed changes would be to cause DRE to deviate further from the current appearance of English, thereby reducing its familiarity and acceptability to the average reader.  Finally, the last change alters stripped DRE to solve an unpleasant problem inherited from TS, at the cost of making it diverge from the other forms of DRE.

a.  Making DRE more phonemic

As discussed above, DRE is not a phonemic writing system.  In terms of the consonants, there are three primary ambiguities:

  1. The th digraph represents both the voiceless /T/ and voiced /D/ th sounds.  While usually only one of the two sounds is possible for a given spelling, there are a few words where the meaning must be considered in determining the pronunciation.  The pair either/ether is the classic example.  DRE could easily be modified to distinguish these sounds, probably using the classic dh digraph to respresent the voiced sound, as in dhis, widhin, ôdher and soódhe. Many of the most common English words contain the /D/ sound, so this change has a radical impact on the overall appearance of DRE.  A change with slightly less impact is to keep th for the voiced sound, and use tx for the voiceless sound, as in txick, txrill, atxléte and sympatxy.  I like this because it helps preserve the similarity of related words like batx/báthe, bretx/breathe and clòtx/clóthe.  Despite this advantage, most readers of English would no doubt find this convention difficult to get used to.

  2. The letter x is used both for /ks/ and /gz/.  This flaw is straightforward to fix, using a double x for the /gz/ sound.  This combination does not occur in standard written English (or at least did not before the Exxon company came to be), but it is easy to read and not unnatural looking: exxist, auxxilyary.  The main reason not to make this change is that it appears to be completely unnecessary.

  3. The ng digraph is used for three different sounds: /N/ (hung), /Ng/ (hungry) and /ng/ (ungrateful).  While the distinction between /Ng/ and /ng/ is mostly academic, and varies widely between speakers, there is no doubt that the /N/ and /Ng/ sounds are of a rather different character, and that spelling finger and singer so similarly is not ideal.  I've considered a number of modifications to DRE to repair this fault, but most of them are too outlandish-looking to consider for long.  I consider the best solution to be spelling the soft ng sound with a tilde (ñg), as in tañg, siñger, readiñg, leñgth and shañghaí.  (In reduced DRE, the diacritic would be omitted at the end of a word, since the hard ng never occurs in final position.)

As for the vowels, there is one ambiguity in DRE that dwarfs all others: the inability to represent in spelling the difference between a short vowel and a schwa.  The words chapel and lapel use the same letters with different interpretations, as do pozitiv and pozition.  Any change to this would entail a huge change in the appearance of the language, and seems out of the question.  If one was absolutely determined to do such a thing, I would recommend a diacritic under the vowel to indicate a schwa, as in pozịtiv and pọzitiọn.  This is unobtrusive enough to avoid interefering with normal reading speed, or giving a false impression of stress.  Of course, these characters are rarely found on non-Vietnamese keyboards, and are difficult to generate with any but the most Unicode-aware software.  Thus, even if one was inclined to think that making this distinction improves the orthography (which I do not), practical considerations intervene.

b.  Making DRE more logical

Several of the conventions of current English spelling preserved by DRE are particularly illogical.  Proponents of logical spelling might prefer that these regular but illogical features be remedied.  I myself consider the benefits of these changes to be very small.

  1. The word endings -le and -re are particularly illogical - the e is itself not pronounced, but adds a vowel sound before (or a syllabic character to) the preceding consonant.  It's almost as if the spelling system were itself dyslexic, as reversing the order of the two letters generally results in an acceptable spelling (except when the letter c is involved, as in <particle> or <acre>).  The best correction for this, one to bring joy to the hearts of those who think that reformed spelling's task is mostly that of shrinkage, is to just leave the e off: pàrticl, búgl, ácr.  One might be tempted to go further, and omit the final vowel from words like <camel> and <thunder>, but this is entirely contrary to the DRE philosophy of not respelling words whose spellings are already acceptable.

  2. The past tense ending often adds a gratuitous silent e, as in moaned, wished and looked.  Sometimes, this is accompanied by a doubling of the final consonant, simply to avoid ambiguity, as in pinned or stripped.  The plural ending does not work like this; we write looks and pins, not lookes and pinnes.  Logically speaking, the extra e and the doubling should go.  (Unfortunately, the doubling is still needed before -ing, in order to distinguish pinning and pining, though this could be fixed by writing pineing.)  It would be possible to go even further with this by, as is often recommended, spelling the past tense with a t where it is so pronounced: wisht, laaft.  This is not a change I could ever consent to, as I consider the value of using the same letter for all regular past tenses to be far greater than the value of accurately representing the variance in the way this inflection is pronounced.  I note that many proponents of wisht do not also suggest the equally accurate spellings wishez or ribz.

  3. Traditional spelling frequently adds an -e before the s of the plural of a word ending in a vowel, as in chilies and heroes.  For words ending in -o, of which there are very many, this is particularly aggravating, as it is usually difficult to predict whether an e is required or even accepted.  DRE could easily mandate that no e is to be added.  For -i words, the required diacritic on the long vowel prevents ambiguity when the e is omitted.  For o, life is more complicated, as the diacritic would be omitted from an ending o in reduced DRE, but even so, the rules of DRE prohibit the formation of words which end in an unmarked -os.  Dropping the e's from these plurals is probably a change which ought to be made, but I have resisted it out of a general desire to avoid modifying the spellings of grammatical features of the English language.

  4. Another feature of the English plural which is rather illogical is the replacement of ending -y with -ie- in the plural.  Many reformed spelling systems repair this, so that you would have variëtys and remedys, but this is uncalled for in DRE.  The mutation of -y to -ie- is completely regular and predictable, and the ie digraph has the correct pronunciation in DRE, so retaining the y in the plural serves only to make the orthographies diverge without offering a significant benefit.  Similarly, using spellings like béutyful, happynêss and fúryôus to preserve the y of the root word seems misguided, however logical.

  5. The use of i as a modifier letter in the endings -cious, -tion, etc. is quite illogical - there is no remnant of any vowel-like qualities to the i here.  DRE has already introduces the digraphs sj, tj and zj in certain words: presjur, questjon, úzjùal.  Why not make the j convention universal, so we would also write eficjent, impresjon, conscjôus, indignátjon, vizjon?  The only difficulty here is the fact that DRE defines tj to be pronounced as /tS/, not /S/, but the rule could be modified to require this pronunciation only before u or after s, t and (optionally) n, and to respell the exceptional word ríetjôus as ríettjôus.  An even more radical proposal would be to eliminate the redundancy of cj, sj and tj altogether, and just use sj for /S/ in all cases, and tj only for /tS/: efisjent, lusjôus, indignásjon, pàrsjal.  Because of the large number of common words using these endings, either form of this change would have a great impact on the overall appearance of DRE.  A less drastic proposal that might be worth doing is to correct those words where the spelling of a suffixed form is gratuitously different from the root, e.g., inflùencial, abolision, transmition, benefitial.

c.  Regularizing suffixes

As discussed above, a serious spelling difficulty that DRE does not address is the irregularity of certain suffixes, especially -able/ible (and the related -ability/ibility), -ance/ence (as well as the related -ancy/ency and -ant/ent) and -er/or.  A phonemic spelling system solves these problems easily by establishing a common spelling for each (e.g., -ibl, -ens, -ur) and sticking to it.  For a system like DRE, whose goal is to change as little as possible, solutions are not so easy to come by.  There are a number of approaches one can take, and they all have their drawbacks.

I have been pondering these issues for quite some time, and have reached no definite conclusions.  If, at the present time, I were to try to enhance DRE to better handle these suffixes, I would try the following (and possibly try something different tomorrow):

  1. The ending -able would be used when the remainder of the word is itself an English word, and -ible would be used otherwise.  -eble would be used rather than -able after a soft c.  Any words ending in -üble (such as solüble and valüble) would continue to do so.  Thus: breykable, díjestable, audible, flammible, repláceble, redúceble.

  2. The case of -ance/ant/ence/ent is the most difficult.   First, I would restrict the applicability of the rules which follow to words where both the -nt form and one of the -nce or -ncy forms exist.  This would leave out words like science and clearance.  Words where a soft c or ci/ti precedes the suffix would retain the current spelling:  magnificent, indécency, pátient, proficient.  All other such words would be respelled with a u vowel: elegunce, differunce, relíunt, prominunt.  This last avoids the effect of seeming to be simply a misspelling, at the cost of having to respell twice as many words.  It also forces the respelling of some extremely familiar words, such as absunt and distunt.  One may question whether this rule is really an improvement; we now have three endings instead of two, and it is still not obvious in some cases which is correct.

  3. The ending -er would be used in all cases where the remainder of the word is itself an English word: viziter, acter, counseler, dictáter, burgler.  In cases like resister/resistor, the word with the more literal meaning would be given the -er ending.  Words like coroner, ancestor and sponsor, where the suffix is less literal, would not be respelled.

It should be obvious that these rules are considerably less than perfect, and it is not at all clear that even significantly more complex rules would do much better.  It also is rather unclear whether these changes would actually contribute significantly to making the language easier to spell.  It is clear, however, that they would greatly decrease the familiarity of DRE text, which is one of DRE's strong points.  For these reasons, I am unable to bring myself to make these changes, despite the seriousness of the problems they address.

d.  Improving stripped DRE

Stripped DRE is DRE with all the diacritics removed.  It is similar to Wijk's Regularized English in somewhat improving English spelling without using diacritics.  It works because, even without diacritics, the changes in DRE from TS were designed to indicate, or at least not mislead about, correct pronunciation.  In one situation, DRE "fixes" TS simply by adding a diacritic, but without the diacritic, the current spelling remains misleading.  This is the situation with words like <epitome> and <cafe>, where a final e is not only not "magic", it is not silent either.  Because DRE goes to a lot of trouble to remove misleading silent e's (as in the spellings fòrtjunat and jenúin), it seems wrong to fail to correct e's which look silent, but are not.  The problem with making such changes is that it causes stripped DRE and the other forms of DRE, which can use diacritics to make this distinction, to diverge.

There are several approaches to fixing this in stripped DRE, differing mainly in the details:

  1. My personal preference for a correction here is to replace a final long e with an -i (epitomi) and a final e or ee pronounced as long a with ey (cafey).  Though -ay would show the pronunciation better than -ey does, it fails to work in words in which the e is preceded by c, like <divorcee> and <fiancee>.  An ending -y could be used in place of -i, and produces more familiar looking words, but this also creates potential confusion in spellings like catastrophy and finaly.

  2. Another possibility is simply to continue to use the DRE diacritics on e at the end of a word: epitomé, cafè.  The biggest issue here is that TS already uses a diacritic in some words of the sort, such as <resumé>, but it is the wrong diacritic for DRE.

  3. A hybird solution could use the -i solution for words like <epitome>, and use é in words like café and divorcée.  I regard changing the direction of the accent as a greater divergence from the other modes of DRE than simply replacing the final e with ey.

Another difficulty with stripped DRE occurs in a number of words where i as pronounced as a long e instead of the long i one would expect, such as <machine>, <police> and <antique>.  Using the letter ì for the vowel takes care of this in strict and reduced DRE, but is not possible in stripped DRE.  Perhaps special stripped forms such as mashien, poliece and antieq should be used.  I note that that <suite> is respelled as swéte even in strict DRE, as it is completely unrecognizable as swìte.  There are of course many other words whose form in stripped DRE leaves much to be desired, but it is probably best to stop here rather than continuing with less prevalent difficulties.

I have not established any of the changes above because, at this time, I regard stripped DRE as being of relatively little interest, and consider reduced DRE to be the most promising form.  If, somehow, there were to be a massive groundswell of enthusiasm for stripped DRE, I suspect I would change the definition as indicated, using the first listed approach to the problem of the unsilent final e.  Until that day, it seems best to keep the definition of stripped DRE simple, even at the cost of considerable imperfection.

To comment on this page, e-mail Alan at

Go to home page
Go to spelling system roster