[Journal of the Simplified Spelling Society, 1989/1 pp.21-29 later designated J10]
[Chris Upward: see Journals, Newsletters, Leaflets, Media, Pamflet 15, Book, Papers.]

Conflicting Eficiency Criteria in the Cut Speling - 2.

Christopher Upward.

This paper continus th discussion of th dilemas of Cut Speling (CS) introduced in Journal 1988/3 J9 (pp17-23). Th cut spelngs used here ar fairly radicl, and readrs wil find many of th mor problmatic forms discussd in th presnt articl and its predecesr. Readrs ar invited to coment on th forms used.


Speling reform aims primarily to make sound-symbl corespondnce mor predictbl. Redundnt letrs (wich by definition conflict with regulr sound-symbl corespondnce) ar particulrly trublsm featurs of th traditionl orthografy (TO) of english, and CS takls them by systmatic omission. Most omitd letrs fal into one of 3 categris: 1) they may be silent letrs, 2) they may represent post-accentul shwa befor <l, m, n, r>, or 3) they may be dubld consnnts. Cuting redundnt letrs has furthr advantajs: it makes riting mor economicl, is fairly esy to mastr, and rarely chanjes th apearance of words so drasticly that they becom hard to recognize (i.e. CS is hyly compatbl, both bakwrds and forwrds, with TO). But these qualitis of regularity, econmy, simplicity and compatbility somtimes conflict with each othr, and then decisions hav to be taken as to wich shud hav priority. Th foloing questions in particutr wer discussd: 1) wethr CS may produce exessiv brevity, 2) how diftclt CS may be to lern, 3) wich CS forms ar hardst to recognize (problms of forwrds compatbility), and 4) wich TO forms wud be hardst to recognize for readrs ho had lernt to read and rite only in CS (problms of bakwrds compatbility).



5.1. 'Long' and 'short' vowls.

Similr conflicts of orthografic eficiency arise wen it coms to ensuring that th so-cald long vowels, as in raid, read, ride, rode, rude, ar relybly distinguishd from ther short equivlnts, as in pat, pet, pit, pot, put/pull. Th roman alfabet laks any clear way of making this distinction, altho it is centrl to th fonolojy of english. TO is notorius for its inconsistncy on this point, as observd in pairs like proper: toper, hint: pint, ration: nation, river: diver, gone: tone, wich giv no indication to readrs that ther vowls ar difrntly pronounced, nor wud riters gess from ther pronunciation that th vowls ar spelt th same.

5.2. 'Majic' <e> in TO.

One of th devices that TO uses, howevr inconsistntly, to distinguish long and short vowls is 'majic' <e>, as wen th silent final 'majic' <e> in hate indicates that th vowl <a> is pronounced long, wile th absnce of final <e> in hat tels us that th vowl is short. Apart from its inherent ilojicality, ther ar at least two practicl objections to 'majic' <e>. Th first is sycolojicl: 'majic' <e> confuses lernrs by interupting norml left-to-ryt line-scaning; so wen they encountr a word like waste, th letr-sequence first sujests a word begining with th sound was, and only wen th 'majic' <e> is rejistrd (and its significnce undrstood) thre letrs later dos th readr realise that th preceding <a> has a quite difrnt valu (but th pair wasted: lasted shos how unrelybl an indicator that <e> is). Th secnd objection to 'majic' <e> is that it givs rise to numerus inconsistncis and uncertntis wen sufixs ar add; so th 'majic' <e> disapears in waging (but not in ageing), and is optionl in lik(e)able. Howevr, altho an ideal rationl orthografy for english wud certnly not use 'majic' <e>, it is so widespred in TO that it canot be elimnated without drasticly afecting th apearance of th ritn languaj (as hapnd wen New Speling respelt hate as haet).

5.3. Distinguishing 'majic' <e> in CS.

Both to prevent disturbing chanjes to th apearance of words and because 'majic' <e> is strictly speaking not redundnt, CS retains it. Howevr its retention dos cause adults som dificlty wen they first lern to rite CS, as they hav to distinguish th very comn redundnt final <e> (as in are, imagine and numerus othr words), wich is cut in CS (ar, imajn etc), from th tru 'majic' <e>, wich is not cut. Th first step is to distinguish 'majic' <e> in words such as hate (with its long vowl) from th misleading and redundnt final <e> in words like have, wher th preceding vowl is short. Th distinction may be a litl less obvius wen two consnnts folo th vowl, but it aplys in exactly th same way; so chanje, table, title etc must keep th final <e>, but flanj, rabl, litl etc, with ther preceding short vowls, do not. Th distinction is initialy also somtimes overlookd wen th silent <e> ocurs modialy rathr than finaly, tho th rule is th same here too; thus wile th secnd <e> is redundnt in TO seven, it must be kept in even to indicate that th preceding vowl is long (CS sevn: even). Once lernrs hav masted th distinction, th CS forms ar seen to ofr an altogethr mor predictbl represention of th difrnt vowls than dos TO.

Wat this distinction amounts to in terms of CS rules is that Rule 1 (redundnt letrs ar cut) must take precednce over Rule 2 (post-accentul shwa befor <l, m, n, r> is cut). Th secnd <e> in even is not made redundnt by a foloing sylabic <n> (as it is in sevn) and canot be cut by Rule 2 because it is preservd by Rule 1. Th difrnce is also seen in th sylabl structur of th two words: sevn consists of <sev> + sylabic <n>, wile even consists of <eve> + sylabic <n>.

Th foloing CS pairs demnstrate these distinctions:
rabl: label
dinr: diner
apl: maple
rivr: dive
latr: late
hovr: over
sevn: even
supr: super

5.4. Aberant preceding vowls.

Th above patrn also aplys wen preceding <o> has abnorml valu, as in th pair lovr: mover. Altho <o> here has two of th norml valus of <u>, th length distinction stil aplys, exactly as it did between hovr and over.

This efect arises in sevrl comn monosylabls too. Th final <e> in gone, shone is clearly redundnt as these words do not rym with tone; CS Rule 1 therfor produces gon, shon (ryming with on), but leves tone uncut. Now altho done ryms with fun, not with on, and idealy it shud therfor perhaps be spelt dun, this wud rase problms of bakwrds compatbility: a child ho had lernt dun wud find it hard to decifer done, wich apears to rym with tone. CS don therfor merely shos that th <o> is short and that don dos not rym with tone; in fact CS alyns don with its ryms son, ton. Th final <e> is similrly cut in words like com, som, lov, dov, glov wich then no longr misleadingly resembl home, cove etc, altho th visul collection between lov: love etc is preservd for th sake of bakwrds compatbility. CS here compromises between th eficiency criteria of perfect sound-symbl corespondnce and bakwrds compatbility.

5.5. Othr letrs as long-vowl indicators.

Less obvius than th abov distinctions is th fact that final <b> in comb, climb, tomb, womb also indicates a preceding long vowl - if it wer deleted, comb wud becom com, tomb wud becom tom, and climb wud apear to rym with him. This silent or 'majic' <b> must therfor be kept, altho it is cut wen th preceding vowl is short, as in lam, lim, bom, aplom, crum, dum, plum, sucum, thum.

We hav alredy seen wy label canot be cut: th <e> indicates a preceding long vowl. Howevr in many similr words th unstressd vowl-letr is not <e> as in label: it is <a> in total, <i> in evil, <o> in idol, <u> in ultimatum. Nevrthless th same patrn aplys as with 'majic' <e>, so ensuring th vowl-contrasts in botl: total, devl: evil, symbl: idol, albm: ultimatum. For th riter an unecesry and unpredictbl vowl-letr is removed from bottle, devil, symbol, album, wile th readr can clearly se th distinction between long and short vowls. Othr exampls of this distinction in CS ar: catl: fatal, metl: fetal, litl: vital, grovl: oval, sutl: brutal. In most such words howevr th preceding vowl is short, and th unstressd vowl-letr can be cut: signl, rebl, pebl, nostrl, pistl, consl, dificlt; but removal, yokel, able with ther preceding long vowls ar not cut. Similrly TO pattern becoms patrn, but patron is not cut (in fact th <o> in patron must also be kept to distinguish th -CVC sequence from th -VCC of pattern).

Sycolojicly th user is here becoming acustmd to new, mor relybl fonografotactic patrns. Wheras in TO a nonce word setal myt be pronounced in one of at least thre ways, as settle, seetle or see-tall, in CS th readr nos that setal is not pronounced as settle; indeed settle: metal merj to look like th ryms they ar. Rule 1 cuts final <e> from settle and Rule 3 simplifys th dubld <l>, giving setl; wile Rule 2 cuts th <a> from metal, giving metl.

Th speling-ambiguity of post-accentul shwa is howevr retaind aftr long vowls, since its removal wud require a hyly disruptiv chanje to th speling of th long vowl (removal myt hav to be reritn remuuvl, for instnce). Here again CS is compromising between th two eficiency criteria of compatbility and regulr sound-symbl corespondnce.

5.6. Alternativ cuts for long <e>.

Wile long <a, i, o> ar comnly spelt with 'majic' <e> (rate, rite, role), long <e> rarely is, forms like Chinese, complete, eve being relativly unusul. Insted long <e> is comnly representd by a digraf such as <ee, ea, ie, ei>. Howevr, wen these ocur in a final sylabl befor certn consnnts (e.g. <v, z>), silent <e> is oftn add, wich then servs efectivly as a secnd (and hence redundnt) indicator of th preceding long <e>; so in receive th long <e>-sound is shown first by th digraf <ei>, and secndly by th final <e>.

Wich letr shud then be cut in such words, th final <e> or th digraf? Th set leave, sleeve, receive, believe cud be cut to leav, sleev, receiv, believ. Howevr th variant digrafs constitute a notorius speling-trap in TO, and it wud therfor be mor helpful to reduce al these words to ther comn denomnator, th sylabl eve (itself a TO form). We then obtain th regularity of eve, leve, sleve, receve, beleve. Ocasionl slyt disadvantajs do arise: th plural of leaf wud presumably hav to be leavs (despite th verb leves), and receit, belief cud not be cut to mach receve, beleve. (Similrly th conection between waif: waive wud be hidn if th latr wer merjd with wave; but few notice th conection in TO anyway.) TO howevr contains many such anomlis (e.g. proceed: procedure, speak: speech, comparative: comparison, message: messenger), and th benefits of regulrizing believe, receive etc must surely outwei these disadvantajs. We here face a conflict of morfolojicl versus fonografic regularity, and we ar giving priority to th latr.

Othr words containing this dubly indicated long <e> with foloing sylabic <l> ar less esily regulrized. Evil myt provide a modl for weevil, but no cuting procedur can alyn eagle with legal. Th <eo> in people is hyly anomlus, causing lernrs real dificlty, and CS peple is a clear improvement. But shud its rym steeple then be cut to mach, as steple, and likewise beadle: needle alynd as bedle: nedle? Ther ar som posbl objections to these forms: cuting these centrl, stressd digrafs may be visuly mor disruptiv than cuting a later, unstressd letr; th TO variations <ea, ee> ar comn and distinctiv digrafs, less liabl to confusion than <ei, ie, eo>; and a mor useful kind of regulrisation for these words myt therfor be th standrd patrn of final consnnt plus <l> (wich must be th long-term ideal speling for such endings anyway), giving weevl, eagl, steepl, beadl, needl. With these words we ar having to make sutlr, mor individul discrimnations in deciding th CS form.


6.1. Regulr simplification.

Dubld consnnts ar somtimes said to be useful indicators of a preceding short vowl in TO, but usuly they ar unecesry and/or inconsistnt. CS Rule 3 cals for them to be simplifyd, wich, in conjunction with th othr CS Rules, can mostly be don quite straitforwrdly, indeed th regularity of th riting systm is therby gretly improved.

Dubld consnnts usuly ocur in english in certn wel-defined environmnts, as in ebb, bubble, accommodate, committee. Al of these can be safely simplifyd by Rule 3, giving eb, bubl, acomodate, comitee. So we hav simplification in monosylabls (e.g. eb, od, tif, eg, wel, bur), in words containing short stressd vowls and ending in sylabic <l, m, n, r> (e.g. bubl, rotn, copr), and in words containing asimilated latin prefixs (e.g. arive, imaculat). Th gain in eficiency is one of predictbility, as wel as econmy: no mor is ther uncertnty about wethr or not to dubl th consonnt in such words; inconsistncy as between th cognates abbreviate: abridge, affray: afraid disapear; and ther is rarely dificlty over bakwrds or forwrds compatbility. In adition ther is oftn gretr acordnce with th speling of othr languajs, especialy spanish (e.g. acomodación); compare also CS comitee with singl <m, t> in comité (french, spanish) and Komitee (jermn). Consnnt digrafs functioning as dubld consnnts ar similrly simplifyd, <ck, cq, dj, xc> being cut to <k, q, j, x>, as in lok, aquit, ajust, exept.

A few TO forms contain a dubld <l> or <s> aftr a long vowl: roll, tulle, wholly, drolly, camellia, bass (in music, not th fish), gross. Singl <l> in TO control, mule, holy, Celia provides a modl for CS rol, tule, droly, camelia. Wholly cud likewise merj with holy, but if it is perceved to be pronounced with a lengthnd /l/, it myt exeptionly be modld on solely as holely; clearly it must not merj with holly. Bass, gross wud be left uncut by th CS sub-rule preserving final <ss> (se § 6.11 belo); certnly bas, gros do not acord with any presnt speling patrns.

Problms arise in a few environmnts if al dubld consnnts ar simplifyd regardless, notebly in disylabic words not ending in <l, m, n, r>, especialy in disylabic words ending in <y>, th <-ing> forms of monosylabic verbs and widely with <rr> and <ss>. These cases present a conflict between activ transfer eficiency and fonografic predictbility, and ar discussd in th rest of section 6 and in section 7.

6.2. Medial consnnts.

Medial consonnts in disylabic words not ending in sylabic <l, m, n, r> ar hyly inconsistnt in ther patrns of dubling in TO. Som variations with words ending in <y> ar: abbey: cabby: baby; ferry: very: query: eery; eddy: ready; city: bitty; filly: happily: wily; choppy: copy: ropy: dopey; ston(e)y: money; study: Judy: muddy. Mecanicl aplication of th CS rules wud produce th foloing forms: aby: caby: baby, fery: very: query: (e)ery, edy: redy, city: bity; rily: hapily: wily; chopy: copy: ropy: dop(e)y, ston(e)y: mony; study: Judy: mudy. Th TO forms ar caractrised by two-way confusion of sound-symbl and symbl-sound corespondnce; but if we simplify al th dubld consnnts, we replace th familr two-way confusion by a new one-way confusion: th simplifyd forms tel us how to rite words, but giv us even less gidance than befor as to ther pronunciation.

Such problms ar rarer and perhaps less serius in longr words. So th paralel between nulity: nudity: credulity may be acceptbl, both because of difring word-length and because th valu of <u> in nulity is derived from nul, wheras in nudity it derives from nude. Elsewher a difrnce of stress may be obscured by simplification, as wen dilema is ritn paralel with cinema; but since TO alredy givs paralel spelings to thre difrnt patrns of pronunciation in cinema, enema, oedema (US edema), a fourth variant in dilema may be acceptbl as scarcely worsning th confusion. Howevr a fulr study of polysylabic forms containing dubld consnnts is needd befor any conclusions can be confidntly drawn.

If th medial consnnts in al such words wer consistntly ritn singl, a gret hazrd of TO wud be removed for th riter; but insofar as dubld consnnts indicate a preceding short vowl in TO, ther simplification wud be unhelpful to th readr; th balance of advantaj is therfor unclear. In som cases simplification of dubld consnnts merjs minml pairs with short and long vowls, so introducing total ambiguity (tinny: tiny); but elsewher th ambiguity is only implicit (and then, to varying degrees). Th form query may inhibit us from shortning ferry to mach very, but (apart from names such as Cody, Thody) ther ar no such awkwrd paralels to inhibit us from alyning shoddy, toddy with body.

6.3. Stratejis for CS.

Sevrl posbl ways of dealing with th problm sujest themselvs, non of them entirely satisfactry:

1. CS cud simplify al dubld consnnts regardless of ambiguity, speling holly: holy, tinny: tiny identicly.

2. CS cud simplify al dubld consnnts exept wen hetrofones result, riting droly, foly, joly but not cuting holly, tinny.

3. CS cud simplify dubld consnnts exept wher grafotactic ambiguity arises, riting droly by analojy with holy, but not simplifying folly, jolly, and riting shody by analojy with body because no ordnry words end, in -ody pronounced with long <o>.

4. CS cud leve al medial dubld consnnts in such polysylabic words uncut.

6.4. Th -ing form of verbs.

Th abov problm of dubld consnnts is particulrly serius in th <-ing> form of monosylabic verbs.

In longr verbs simplification brings considrbl advantajs by elimnating th major uncertnty about dubling final consnnts befor sufixs. Thus th variation between committing, visiting, itself hyly conduciv to mispeling, leves benefit(t)ing, formatting unclear or anomlus, and practice varis with som endings (especialy <-l>) between americn and british speling: US traveling, kidnaping, worshiping, british travelling, kidnapping, worshipping. Al these ar regulrized in CS comiting, visiting, benefiting, formating, travling, kidnaping, worshiping. In som cases th stress-patrn is then obscured: <tt, rr> in comitting, occurring tel us th secnd vowl is stressd, wile in visiting, murmuring th first vowl is; howevr TO oftn also fails to indicate stress (considr formatting, benefitting, travelling: compelling, procuring: murmuring), and th econmy and regularity of th CS forms ar surely prefrbl to th presnt confusion.

With monosylabls howevr ther ar over 60 sets (mostly pairs) of verbs like hop: hope hose <-ing> forms wud by th mecanicl aplication of th CS rules becom hetrografs, e.g. hoping for both verbs. Th problm arises because TO confuses th <-ing> forms of pairs of verbs wenevr th speling of ther base-form difrs only by th 'majic' <e>. Thus, since TO uses <hop-> as th base for th <-ing> form of hope (<hop+ing>), it needs anothr way of distinguishing hoping from th <-ing> form of hop - and it dos so by dubling th <p> of hop, thus hopping.

Th CS Rules provide a straitforwrd solution to th uncertnty of consnnt dubling befor othr sufixs (hopd: hoped, hopr: hoper), but befor <-ing> CS faces a most intractbl problm: at worst, to rite hoping as th <-ing> form of both hop and hope wud introduce a major new ambiguity into ritn english. It is tru that TO tolrates actul ambiguity in th case of bathing wich can derive from both to bath or to bathe, and comn words like having, coming apear to rym with shaving, homing, but it can hardly be acceptbl to spred this ambiguity to som 60 othr sets of verbs too.

6.5. Cataloging hoping-typ ambiguitis.

If th CS Rules giv two difrntly pronounced words th same speling wen in TO they ar spelt difrntly, total ambiguity results, as wen <pp> in hopping is simplifyd. To establish th size of th problm and compile a catalog of such ambiguitis, a matrix was drawn up for each vowl, with th posbl sylabl-initial consnnts and consnnt-strings forming th verticl axis and th posbl final consnnt patrns forming th horizontl axis. (Th inventry of posbl consnnts was taken from Gimsons tables of foneme sequences. [1]) Th resulting grid shud then ensure that al posbl combinations of fonemes in monosylabis ar acountd for.

Th catalog, giving only th TO form of th short-vowl verb in each pair, now folos. Som pairs only merj wen othr CS rules ar also aplyd, as wen plaiting is cut to plating, <ck> is cut to <k> (tacking: taking), or silent initial <w> disapears (wrapping: raping).

With <a>, 23 pairs arise from plait, plan, bar, bath, back, chaff, tap, tack, quack, lamb, stack, scrap, snack, slack, slat, shack, spar, star, scar, sham, mat, rack, rat, have (if CS cuts th <l> in halve). Ambiguity is avoidd within CS between rag, wag and rage, wage if th latr ar respelt raje, waje, but problms of compatbility arise as th adult lernr wil misread CS waging as TO waging, and conversely th child wil misread TO waging as wagging. Th <-ing> form of both wrap, rap wud apear as raping, and perhaps of ball, bawl as baling.

With <e>, no pairs wer found, because nearly al monosylabic verbs prefer digrafs such as <ea, ee, ei, ie> to 'majic' <e>. Sell: seal therfor remain distinct as sealing: seling, and speling is unambiguus. Th arcaic verb to mete is exeptionl in having 'majic' <e>, but it has no short vowl equivlnt. League is presumed cut to leag, not lege.

With <i>, some 20 pairs arise from pip, pin, pill, till, twin, bid, grip, sit, spit, strip, snip, shin, jib, lick, fill, rid, whip (if paird with wipe), wit (if paird with white), will, whirr (if paird with wire). Fonemicly tip: type belong here, but if <y> is kept to represent th long vowl, no ambiguity arises between tiping: typing; similrly, asuming sign is respelt syn, no ambiguity arises with sin. Ther is one trio (if <wh> is simplifyd to <w>) in win: wine: whine.

With <o>, 11 pairs arise from top, dot, cock, cod, chock, stock, slop, lop, mop, knot, rob, hop. If CS keeps 'majic' <b> in comb, combing: coming remain distinct.

With <u>, no pairs wer found; retention of final <ss> in CS prevents a merjr of fussing: fusing.

6.6. Potential and grafotactic ambiguity.

Th danjer of ambiguity is acute and explicit with th abov pairs such as hopping: hoping, but ther ar many othr cases of latent, or potential, ambiguity. For instnce, of th pair slim: slime, only slim normly ocurs as a verb, and sliming shud therfor self-evidntly derive from to slim. Howevr, english is so productiv of new verbs that a sentnce like (TO) the hippopotamus was sliming itself with mud can perfectly wel arise. This means that wenevr a short-vowl monosylabl forms a pair with a long-vowl monosylabl, ther is at least th potential for ambiguity, as wen with lack: take a short-vowl verb pairs with a long-vowl noun, or with fad:fade, wen th reverse ocurs, or wen with rip: ripe a verb pairs with an ajectiv.

Not merely myt sliming concevebly be derived from a verb to slime, but it paralels long-vowl forms such as timing, miming, so producing wat we myt cal grafotactic ambiguity: th readr is familir with th string <-iming> pronounced with long <i>, and lernrs at least wil therfor tend to constru th same patrn elsewher as representing th same sound. Sliming thus constitutes a dubl ambiguity, potential and grafotactic. Elsewher we may hav only grafotactic ambiguity; so if CS cuts th <b> from th verb to thumb and then rites thuming, ther is neithr actul nor potential ambiguity, as no form thume with 'majic' <e> exists; but th ryming paralel fuming sujests a difrnt fonografic corespondnce, leading to posbl mispronunciation of thuming.

Of corse, a numbr of comn verbs alredy exibit such grafotactic ambiguity in TO, as wen coming, having, giving, loving apear to rym with homing, shaving, driving, roving (or moving); yet skild readrs and riters accept them without demur. Cud we not therfor equaly wel accept mor such disparitis, as between sliming: miming, thuming: fuming? Howevr, perhaps th existing ambiguitis ar acceptd precisely because they ocur in such comn words and ar therfor soon lernt. Furthrmor it is not skild users, but lernrs, wethr nativ-speaking beginrs or foren students unsure of pronunciation, hose intrests we must here considr, since it is they ho most need unambiguus symbl-sound corespondnce. They may soon lern to distinguish coming: homing and th consnnt-dubling patrn of slimming: miming; but sliming: timing can only be distinguishd by lerning individul words - th very dificlty that lies at th hart of th problms of TO.

6.7. Ansrs to th hopping: hoping problm.

In principl at least 4 aproachs to th problm ar concevebl, tho not al ar equaly practicbl.

1. A new grafeme cud be used for th long vowl, hose TO speling with 'majic' <e> lies at th root of th problm. For instnce hope cud be spelt hohp, with th regulr <-ing> form hohping, so enabling th dubl <p> in hopping to be simplifyd without ambiguity within CS. Unfortunatly this solution involvs bakwrds and forwrds ambiguity between TO and CS: th adult wud stil misread hoping in CS as TO hoping, and th child wud mistake it for hopping in TO.

2. A less radicl inovation wud be to insist on morfolojicl regularity in th speling of th <-ing> forms of long-vowl monosylabls. Th <-ing> form of hope wud then be hopeing, wile hopping cud be ritn with one <p>; this patrn ocurs ocasionly in TO alredy, as in ageing (tho not raging), and in th distinction between singing: singeing (contrast hinging). Unfortunatly it also entails th same disadvantajs of bakwrds and forwrds ambiguity as dos th hohping solution.

3. A third solution wud be to accept th ambiguity of hoping as th <-ing> form of both hop and hope - and th ambiguity of th <-ing> forms of al th othr pairs too. This solution has hy activ transfer eficiency, but conflicts with som basic principls of good speling, abov al that a word shud be unambiguusly recognizebl from its ritn form.

4. This leves th retention of th TO forms as th only practicbl solution - but at th expense of activ transfer eficiency because of the numerous exeptions to CS Rule 3.

6.8. Problms of defining an exeption-rule.

Idealy a simpl exeption-rule is needd, to tel adult lernrs wen to keep dubld consnnts. Howevr it is not imediatly aparent wat rule cud ensure that comma retains <mm>, wile command, commence, commend, comment, commerce, committee, common, communism al alyn with comedy, comet, comic, comity. It is somtimes sujestd that dubld consnnts be kept aftr stressd vowls; comma wud then remain uncut - but so wud comment, commerce, common, communism, tho not ther cognates comercial, comunity. Altogethr this seems a most untidy aproach.

Ther ar also complications wen one trys to formulate a rule for th -ing forms of verbs. If we say that monosylabic verbs with a short vowl shud dubl ther final consnnt wen ading -ing, wat shud be don in cases like TO coming, having, thumbing, spreading, sweating? Do we rite comming, havving, thumming, spredding, swetting, and therby lengthn th spelings? Uncertnty also arises with compound forms; if <tt> is kept in betting, setting, fitting, that reintroduces th hole problm of consnnt dubling in inflectd forms wich Rule 3 was desynd to resolv.

Simplr than this wud be a blaklist of individul words wich wud be ambiguus (i.e. with th same speling as othr, difrntly pronounced words) if ther dubld consnnts wer simplifyd. Section 7.8 belo wil develop such a blaklist.

6.9. Consonnt-dubling between morfemes.

Consnnts that ar dubld or repeatd across morfeme-boundris also create dificlty for users in TO, as for instnce wen 20% of a group of british university students sitting ther final examnations faild to dubl th <n> in TO openness. Such dubling arises particulrly in th foloing environmnts.

CS simplifys these variations by nevr riting <ll>: words ending in <-l> ad <-y>, wile othrs ad <-ly>: badly, solely, cooly, fuly, droly, ably, simply, basicly. Doutful cases ar wholly, wich myt be respelt holy or holely, and th pair latterly: laterally both apear as latrly.

On first aquaintnce with CS, adult lernrs may feel unesy at simplifications such as singl <n> in unecesry, as they may sense a certn lengthning of th /n/ in ther speech wich they wud wish to spel <nn>. Howevr, as in italian, th lengthning is not clearcut or invariable and if one makes an exeption to th consnnt-simplification rule in unecesry, uncertntis arise elsewher: shud th negativ prefix in- similrly be retaind as in TO innumerable, and if so, shud th prepositions prefix in- likewise produce <nn> in innovate etc? But then we begin to reintroduce th hole minefield of consnnt dubling as found in TO. Th rule that bound morfemes do not lead to dubld consnnts is simplst and most economicl for both readrs and riters.

6.10. Dubld <r>.

Usuly <rr> can be simplifyd like othr consnnts. Just as egg, copper, accommodate becom eg, copr, acomodate in CS, so purr, horror, arrive becom pur, horr, arive. Aftr most vowls ther ar paralel forms with singl <r> in TO wich provide a modl: so TO carrot can alyn with carat, caret and french carotte. Howevr disylabic words ending in <-y> again rase dificltis: just as th atemt to alyn choppy with copy runs up against th disturbing paralel of ropy, so if we try to alyn ferry with very, we face th disturbing paralel of query. Likewise simplification of <rr> in carry, sorry, harry runs up against vary, story, fury, altho in americn speech sorry: story may rym. Furthr complications ar that, depending on meaning, tarry may rym with carry or with starry, and furrier may hav th vowl of hurry or of furry. Perhaps th derivativ ajectivs starry, tarry shud be analysed as star+y, tar+y and hence ritn stary, tary (despite vary), wile to tarry remains uncut to paralel carry.

A preceding <u> is particulrly inhibiting, as th valu of <u> in hurry rarely ocurs befor a singl consnnt folod by a vowl in TO (bunion, pumice, punish, study being exeptionl). So perhaps burro shud be distinguishd by <n> from bureau, hurricn from Huron, turret from Turin. This may even lead us to question such CS forms as curaj, surogat, ocurence, altho it is notebl that americn speech comnly givs this <ur> th same valu as in fur.

Restrictions may be less severe with othr vowls befor non-<y> endings: ther ar no such obvius grafotactic objections to cuting arrow etc to aro, baro, faro, maro, naro (unless Pharaoh wer cut to Faro), and carion machs Marion. Aftr <e> th paralel of ferret: merit wud seem to alow feret (cf french furet). Aftr <o>, we myt decide that th distinction between sorry: story is not significnt and therfor rite sory, horid (cf florid), moro; ambiguity wud howevr arise between borrow: borough if both wer ritn boro. In americn speech worry has th vowl of word, and th speling wory myt therfor seem satisfactry.

This analysis is not exaustiv, but a prelimnry conclusion seems to be that <rr> can usuly be simplifyd, but not always aftr <u>, nor aftr short <a, e> befor <y>.

6.11. Dubid<s>.

Simplification of <ss> is also problmatic. Final <ss> always indicates voiceless /s/ in TO, wile final <s> is comnly (e.g. in most inflexions) voiced as /z/. Final <ss> therfor canot be simplifyd if we ar to avoid widespred ambiguity, as between princes: princess (indeed discus: discuss do not even difr as to voicing). If a later reform wer to regulrise th /s, z/ sounds jenrly, final <ss> myt prove a useful transition grafeme for distinguishing pairs such as hens: hence (hens: henss).

TO provides no such clear distinction between voiced and voiceless valus of medial <s, ss>. Thus present is herd with both voiced and voiceless <s>; house, use ar voiced as verbs but not as nouns; and <ss> is voiced in dissolve but voiceless in dissolution. Medial <ss> tends to be voiceless (assess, message, dissent, cossack), with dessert, dissolve, hussar, possess, scissors as wel-nown exeptions, and medial intervocalic <s> tends to be voiced (busy, pleasant, visit, weasel, with basin, mason, sausage as exeptions. Wethr <s> befor or aftr conssnts is voiced varis with th consnnt (tho it dos not depend on wethr that consnnt is itself voiced): compare voiced <s> in crimson, dismal with voiceless <s> in ransom, basket. Th voiced-voiceless distinction in valus of medial <s> dos not apear significnt (it has very lo funtionl load) in english, and in most cases CS rites just <s> for both valus: presnt, desert, disolv, disolution, asess, mesaj, disent, cosak, husar, posess, sisrs, busy, plesnt, visit, weasl, basin, mason, sausaj, crimsn, dismi, ransm, basket. No miniml pairs hav been found wich wud then becom ambiguus, altho ocasionl contrasts ar lost, as wen TO feasible: possible becom CS feasbl: posbl, and lernrs hav even less gidance as to voicing than in TO. Howevr, mispronunciation is unobtrusive and th riter enjoys much improved predictbility and econmy.

Ther remain som unresolvd cases. Befor sufixs, for instnce, it is unclear wethr CS shud keep final <-ss> or rite medial <s>: profess givs professd, professing, professr, but th rule for medial <s> givs profesd, profesing, profesr (by analojy with predecesr?). Anothr case concerns th regulr TO distinction between patatlized voiceless intrvocalic <ss> in th -ssion endings (mission), wich contrast with voiced -sion (vision), especialy since <ss> also indicates preceding short <a, e, u>, as in th pairs occasion: passion, lesion: session, confusion: concussion. (Howevr single <s> is voiceless aftr a consnnt, as in repulsion, tension.) No length-distinction arises with <i> (wich is nevr long) or <o> (wich is nevr short), but voicing is shown in vision, erosion as oposed to mission. If we keep <ss> in these cases, disparitis may arise in derivativs (e.g. if misl, misiv hav singl <s> despite mission). We here face a conflict of eficiency between morfemic consistncy and consistnt cuting rules.


7.1. Th natur of th problm.

One of th most caractristic dificltis of TO is its ambiguity, wich takes two main forms: a givn speling may hav mor than one pronunciation (hetrofones, as in tear (=rip) and tear (from weeping), and, far mor comn, difrnt spelings may hav to be lernt for words pronounced identicly (hetrografs, such as meat: meet). Al speling reformrs agree that hetrofones shud be difrntly spelt; but they disagree wethr al, or som, hetrografs shud in principl hav th same speling. Th problm of hetrografs and its implications for speling reform in jenrl was examnd at som length in an erlir articl, [2] and we shal now considr how far CS in parliculr resolvs or agravates such ambiguitis. It is clear that wen ambiguity results directly from redundnt letrs, CS can remove it; but th cuting rules produce a ranje of efects, from those that wud be universly acceptd as desirebl, thru those that hav both advantajs and disadvantajs, to those, at th othr extreme, wich actuly create new ambiguitis.

7.2. Hetrofones unafectd by CS,

Norml CS rules canot difrentiate hetrofones if th discrepncis ar not caused by redundnt letrs, as in:

7.3. Hetrofones difrentiated by CS.

Since som letrs in th foloing hetrofones ar redundnt for one of ther pronunciations, CS difrentiates them:

This distinction of hetrofones in CS reduces th ambiguity of TO, especialy for th readr (tho sycolojists may be dismayd at th loss of som of the 'gardn-path' forms they use for testing readrs' responses). In a difrnt way, riters benefit from th improved sound-symbl corespondnce and from th gretr econmy.

7.4. Variety of hetrograf-merjrs in CS.

Wile th abov distinctions al reduce th ambiguity of ritn english, CS also tends to increse its ambiguity wen hetrografs merj. Hetrografs wich in TO hapn to share th esential letrs required to represent ther pronunciation becom homografs in CS by omission of redundnt letrs. Wile such merjrs remove uncertnty for th riter, ho no longr has to remembr alternativ spelings for a givn pronunciation and meaning (e.g. dependent: dependent merj as dependnt), ambiguity is incresed for th readr, ho no longr imediatly nos from th speling wich sense of a word is intendd. Such ambiguity (polysemy) is alredy widespred in TO (e.g. bank can be eithr a rivr-bank or a mony-bank) and is not in itself necesrly undesirebl. Th main reservation speling reformrs must hav about such merjrs is that cumulativly they increse th overal ambiguity of th riting systm. If taken to extremes (for instnce in a languaj with many mor homofones than english) such ambiguity can rendr fonografic riting dysfunctionl; indeed that is perhaps th main reasn wy chinese and japnese hav not been able to adopt th roman alfabet for jenrl use.

Practicl experience with CS has not so far sujestd that such incresed ambiguity causes readrs any jenrl dificlty. Howevr, in terms of bakwrds and forwrds compatbility, th merjrs introduced by CS ar not al of a kind: som giv rise to gretr ambiguity than others, at least transitionly for adult lernrs. We shal now examn th variety of merjrs that ocur.

7.5. Symetricl cuts:

 peace: piece > pece.

TOpeace  piece
 \ /  
CS pece  

Least problmatic ar symetricl cuts, wen two (or mor) hetrografs lose difrnt redundnt letrs to merj in a third, new form. Thus peace loses redundnt <a> and piece loses redundnt <i>, so producing th merjd form pece. This is imediatly decodebl, it resembls both peace and piece, no confusion arises with existing TO spelings, th context makes th meaning clear, and neithr word is likely to be mistaken for th othr (contrast pece of mind, a pece of my mind). Al we hav here is a new pair of homonyms like bank: bank.

An augmentd version of th hetrograf list in Journal 1987/1 J4 servs as th sorce for th foloing 115 sets of TO hetrografs wich wud be merjd symetricly by th mecanicl aplication of th CS rules. Etymologically related forms ar astriskd *: (It must be remembrd that for a variety of reasns, it myt be desirebl to make exeptions in a numbr of th foloing cases in practice.)

aisle: isle > ile
altar: alter > altr
axel: axle > axl
ball: bawl > bal
bell: belle > bel
boar: bore > bor
boarder: border > bordr
bole: bowl > bol
calendar: calender > calendr
callous: callus* > calus
carpal: carpel > carpl
censer: censor > censr
chough: chuff > chuf
coarser: courser > corsr
core: corps > cor
cubical: cubicle > cubicl
dollar: dolour > dolr
ewe: yew > ew
filter : philtre > filtr
floe: flow > flo
gamble: gambol > gambl
greave: grieve > greve
hall: haul > hal
hangar: hanger > hangr
hoard: horde: whored > hord
immanent: imminent>imnnt
lea: lee > le
literal: littoral > litrl
mall: maul > mal
manner: manor > manr
marten: martin > martn
meddler: medlar > medlr
moor: more > mor
mustard: mustered > mustrd
palette: pallet* > palet
peace: piece > pece
pedal: peddle > pedl
petrel: petrol > petrl
pole: poll > pol
principal: principle >principl
rack: wrack (*?) > rak
radical: radicle* > radicl
retch: wretch > rech
rigger: rigor: rigour > rigr
role: roll > rol
roux: rue > ru
sailer: sailor* > sailr
soh: sow > so
sloe: slow > slo
soared: sword > sord
stationary: stationery* >stationry
tea: tee > te
there: their > ther
tough: tuff > tuf
weaver: weever > wever
which: witch > wich
whoa: woe > wo
all: awl > al
auger: augur > augr
ay: aye: eye > y
baton: batten* > batn
billed: build > bild
board: bored > bord
bolder: boulder > boldr
buyer: byre > byr
call: caul > cal
cannon: canon* > cann
caster: castor(*?) > castr
choler: collar > colr
coal: cole > col
complement: compliment > complmnt
coward: cowered > cowrd
dependant: dependent* >dependnt
ere: heir > er
felloe: fellow > felo
flea: flee > fle
freeze: frieze > freze
gin: (d)jinn > jin
grill: grille > gril
handsome: hansom > hansm
hoar: whore > hor
hostel: hostile > hostl (US)
knight: night > nyt
lessen: lesson > lesn
limb: limn > lim
lumbar: lumber > lumbr
mangel: mangle > mangl
mantel: mantle* > mantl
medal: meddle > medl
missal: missel: mistle: missile > misl
muscle: mussel* > must
ordinance: ordnance* > ordnnce
pea: pee > pe
pedaller: pedlar (US peddler) > pedlr
pendant: pendent* > pendnt
pidgin: pigeon > pijn
pore: poor: pour > por
rabbit: rarebit > rabit
reck: wreck > rek
rapped: wrapped > rapd
rho: roe: row > ro
right: wright > ryt
rough: ruff > ruf
rye: wry > ry
sea: see > se
slight: sleight > slyt
soar: sore > sor
sole: soul > sol
summary: summery >sumry
tenner: tenor > tenr
throe: throw > thro
watt: what > wat
where: wear > wer
whither: wither > withr

7.6. Asymetrical cuts:

 plaice: piece > place.

CS place

A sutly difrnt kind of merjr ocurs if only one hetrograf contains redundnt letrs, and it then merjs with an existing form; so wen plaice loses its redundnt <i>, it turns into place. We may describe this as an asymetricl merjr of th TO pair plaice: place. Somtimes mor than one word is reduced to th same existing form, as wen borne: bourn ar both cut to born. Even tho, as with pece, th meaning wil normly be clear from th context, th visul impact of an aparently familir form with an unfamilir meaning is transitionly disturbing for th adult lernr of CS, place for plaice being in efect a 'gardn path' form. We here hav a conflict between forwrds compatbility and fonografic regularity, th latr having priority over th formr in asymetricl merjrs.

Th CS rules wud, if aplyd in al cases, produce th foloing 114 asymetricl merjrs (how curius that th numbr shud be so close to that for symetricl merjrs!). Cut forms alredy found as alternativs in at least som variants of TO ar markd .

aide > aid*
balled: bawled > bald
baulk > balk*†
bowled > bold
bread > bred
candied > candid
charred > chard
coarse: course > corse
dessert > desert (verb)
Finn > fin
fore: four >for
gauge > gage †
heard > herd
hoarse > horse
jamb > jam
knew > new
know > no
lapse > laps
leaver > lever
maize > maze
mourn > morn
nett > net †
parr > par
pie > pi
psaiter > salter
raise > rase
seamen > semen
soh : sow > so
Storey > story †
tease > teas
too: two> to
warred > ward
weather: whether > wethr
Whig > wig
whine > wine
wholly > holy
wrap > rap
wring > ring
aunt > ant
banned > band
bee > be
borne: bourn(e) > born(*?)
butt > but
canvass > canvas*
chord > cord*
cruise > cruse
eaves > eves
Finnish > finish
forego > forgo †
gnu > nu
heart > hart
hour > our
knave > nave
knit > nit
laager > lager*†
lead > led
llama > lama
malt > mat †
mourning > morning
oh: owe > o (cf IOU)
pease > peas
plaice > place
programme > program †
read > red
scent > sent
soled > sold
straight > strait †
tolled > told
tore > tor
wee > we
when > wen
while > wile
whit > wit
whooping > hooping
wrest > rest
wrote > rote
bade > bad
barred > bard
bogey> bogy (*?)†
buoy > boy
buy: bye > by †
caste > cast
copse > cops
damn > dam
fiancée > fiancé*
flue > flu
fourth > forth
guild > gild †
heroine > heroin
inn > in
kneed > need
knot > not
lamb > lam
learg > lent
low > lo
mooed > mood
mucous > mucus*
oar: ore > or
penned > pend
plumb > plum
reign > rein
savoury>savory †
sell > set †
steppe > step
stye > sty †
tonne > ton *
waive > wave
welled > weld
whet > wet
whin > win
whole > hole
whorled > world
write > rite
wrung > rung

We note that som roots ar involvd in both a symetricl and an asymetricl merjr if bowl: bole becom bol (by analojy with long <o> in control, old), then CS bowled necesrly becoms bold.

7.7. Imperfect merjrs.

Slytly mor disturbing stil for adult readrs than asymetrical merjrs is th case of CS add. Here a dubl shift has taker. place: an orthografic form abandnd by TO, add (wich is cut to ad), is adoptd by CS for TO added.

TOadded add  
CS add  ad

Ther is ambiguity here not within CS, but between CS and TO, wich creates dificltis of bakwrds and forwrds compatbility, tho these ar probbly outweid by th advantaj of systemic regularity: if final consnnts ar simplifyd in eb, eg, then they shud also be simplifyd in ad; and if verbs form ther past tense by ading just <d> as in faded, needd, then ad shud be no exeption but shud inflect as add.

Ther ar a numbr of sets of near-homofones wich myt or myt not be merjd by th CS rules, it being uncertn wethr ther pronunciation shud be considrd identicl. These merjrs may be symetricl or asymetricl, and they wil typicly seem sensbl to speakrs of som accents, but not to speakrs of othrs. For instnce, th <l> in calm, halve is unversly silent, but many speakrs distinguish th vowls of cam: calm, have: halve and myt therfor find it confusing to hav both words spelt cam, hav. Som speakrs ho pronounce th two vowls of city difrntly may wish to keep candid: candied distinct. Othr speakrs myt not accept hostl, misl both for hostel, missal and for hostile, missile; wile othrs again jib at wethr for weather: whether: wether. Probbly few speakrs wud imediatly think of merjing ere: heir with err as er. It needs to be establishd wethr such merjrs ar beneficial in terms of th global rationality of th riting systm; if so, a persuasiv case myt be made for them, but if not, they myt need to be excluded from a CS reform.

Vews difr as to wethr where: wear: were shud al be merjd as wer. Th presnt authr exeptionly now keeps <h> in wher to distinguish where: were, but he merjs wear: were. Othrs howevr argu that if <wh> is jenrly cut to <w>, and if there: their merj as ther (as bear, pear etc ar cut to ber, per), then where: wear: were shud al becom wer. Certnly they ar oftn not clearly distinguishd in speech.

Two comn words, could and done, contain blatantly redundnt letrs, but ther removal merjs them with TO cud, don, wich in most accents ar difrntly pronounced. Here th rarity of TO cud, don shud permit an unproblmatic merjr in th intrests of econmy and wider consistncy: cud then alyns with shud, wud, put, pudding, and don alyns with son, ton.

Th main argumnt in favor of th abov merjrs is that altho adults may find som of them disturbing on first reading, children wud find them mor straitforwrd than th variant forms of TO.

7.8. Unacceptbl ambiguity: a blaklist.

We may accept most or many of th merjrs discussd abov, but othrs go yet furthr in th direction of ambiguity, and wil probbly hav to be excluded. They ar typicly cases wher one word is cut to th same form as anothr hose pronunciation is quite difrnt. Ther is howevr no clear dividing-line between th acceptbl and th unacceptbl, and each posbl merjr has to be jujd on its own merits.

So we hav a case like pall: pawl wich cud be cut like call: caul to th form pal - but pal is a word in its own ryt with difrnt pronunciation; perhaps pall: pawl can be so merjd because they ar rathr uncomn words, or conversely perhaps ther very rarity means they do not justify being thus regulrized, and shud wait for a later reform wen they myt be ritn paul.

Then ther is th group chilled, willed, binned, finned, wich by norml CS procedurs wud merj with child, wild, bind, find, so introducing mor hetrofones like TO the wind: to wind. To prevent this we may exeptionly decide to keep th dubld consnnts, giving chilld, willd, binnd, finnd. Th ultimat solution wud be to distinguish th long vowls, perhaps as in chyld, wyld, bynd, fynd, wynd, but without a transitionl period ther wud stil be problms of bakwrds and forwrds compatbility.

Similrly, dubld consnnts ar needd to distinguish th pairs bellow: below, boggy: bogy, bonny: bony, comma: coma, coral: corral, dully: duly, furry: fury, gammy: gamy, hallow: halo, holly: holy: wholly, knobbly: nobly, navvy: navy, pennies: penis, ragged: raged, spinney: spiny, tinny: tiny, vellum: velum.

Awkwrd individul cases include alley: ally, annual: annul, colour: collar: choler, latterly: laterally wich wud be merjd by th norml aplication of CS rules. If merjrs ar not acceptbl in these cases, individul solutions may be needd, such as aley: aly, anul: anull, color: colr, lairly: latraly.

Th inflectd forms of monosylabic words ending in long /o:/ spelt <ow> ar very trublsm, especialy sho. Ther is no dificlty with shoing, shos (cf going, gos), but showed, shown wud be ambiguus if cut to forms like shod, shoed, shon, and showd, shown ar perhaps th only posbilitis. Similrly CS probbly has to rite bowd, twwd, mown, rowd, sowd, sown from th verbs bo, mo, ro, so etc.

Also serius ar th heterografs toe: tow and doe: doh: dough, wich surely canot be ritn to, do wile those TO forms remain unchanged; th presnt authr wud cut dough to doh, but leve th othr words uncut. Th comparativ forms of th ajectiv lo rase sirnilr problms, lor, lost being totaly misleading; here loer, loest ar prefrbl to lowr, lowst (lowr being an altemativ speling for lour).

Potentialy quite danjerus wud be a merjr of th numerals two, four with to, for (despite th existing cut form forty): misreading of a numeral cud hav serius, even catastrofic, practicl consequences. CS cud howevr recomend that numerals, especially 1, 2, 4, always be ritn numericly, nevr alfabeticly.

7.9. Short cuts thru th complexitis?

Varius aproachs to th abov problms ar posbl. At one extreme, th CS rules may be aplyd regardless of th consequences; that aproach givs us hy activ transfer eficiency at th expense of fonografic predictbility. At th othr extreme, we may seek simpl rules to exclude problm cases (e.g. nevr simplify medial dubld consnnts between vowls in disylabic words), but in th process som useful regulrisations wud be missd. A rathr difrnt aproach wud be to concentrate on th most frequently ocuring base-words (perhaps those listd in Collins COBUILD, [3] or in th LOB/Brown corpora [4]), and delibratly ignor rarer or arcaic words. We may then say that for th purposes of an initial speling reform, an exaustiv lexical analysis is unecesry, and many of th abov merjrs can be left in th relms of th hypotheticl, in th belief that it is of no imediat imporince wethr pawl, ere etc. ar cut to pal, er or not. For practicl reasns th presnt CS Working Party wil in th short term be pursuing this latr aproach.


Th detaild analysis and lists in this articl sujest a paradox: on th one hand, th 3 CS rules ar far simplr to lern than th 40+ rules that wud be needd for a complete fonografic reform of ritn english; but on th othr hand, wen one atemts an extensiv survey of english vocablry, a larj numbr of individul problm-cases emerj. CS is esy to read without instruction, and once th 3 rules ar mastrd, th adult lernr can rite evryday english in CS fluently, indeed mor esily than in TO; yet to produce a substantial dictionry of CS forms requires careful considration of many problmatic minutiae.

Th purpose of this artict (and its predecesr) has not been to provide conclusiv ansrs to th questions rased, but to explain som of th factrs that hav to be taken into acount in seeking th best ansrs. Readrs ar invited to reflect on th isus, and to send in coments, reactions, furthr problm cases, and sujestd solutions.

A foloing articl wil examn th aplication and efects of CS Rule 2, along with a posbl simplification of th use of capitl letrs and apostrofes in english.


[1] A C Gimson An Introduction to the Pronunciation of English, London: Edward Arnold, 3rd edition with corrections 1983, pp. 240-52.

[2] Christopher Upward 'Heterographs in English' in Journal of the Simplified Spelling Society, Vol. 1, 1987/1, J4 pp18-25.

[3] Collins COBUILD English Language Dictionary, London & Glasgow: Collins, 1987.

[4] Knut Hofland/Stig Johansson Word Frequencies in British and American English, Bergen: The Norwegian Computing Centre for the Humanities (Longman), 1982.

Back to the top.