[Journal of the Simplified Spelling Society, 1987/3 pp21-24 later designated J6.]
[Chris Upward: see Journals, Newsletters, Leaflets, Media, Pamflet 15, Book, Papers.]

Can Cut Speling Cut Mispeling?

Christopher Upward.

Th Cut Speling (CS) used in this articl removes redundnt letrs from t.o. as folos: it cuts letrs irelevnt to pronunciation (det), it uses sylabic <l, m, n, r> in most post-accentual sylabls (metl, atm, prism, detr), it regulizes inflections as <d, s> (hintd, bushs), and it simplifys most dubld consonnts (eg, bigr, agravate). Wher pronunciation reqires it, CS also replaces <gh, ph> by <f> (tuf, filosofy), <g> by <j> <jinjr), and <ig> by <y> (syn, hy). Readrs ar invited to coment on th spelings used.

1. SIGNIFICNCE OF ERR-ANALYSIS.

1.1 Practicalitis of reform.
Th evolution of speling-reform proposals, from th pure fonemicism of New Speling, [1] thru th symbl-sound but not sound-symbl regularity of Wijks Regularized English [2] and th micro-stajes of Lindgrens Spelling Reform - A New Approach [3], to CS (omitting redundnt letrs), shos a groing awareness of th practicl implications of reform. These practicl implications hav two aspects: firstly, we hav to considr th practicalitis of implementation, of teaching new spelings, of geting publishrs to adopt them, of ensuring they ar compatbl with tradition orthografy (t.o.), etc.; and secndly, our incresed undrstanding of th sycolojy of litracy, th problms of ilitracy, th natur of dyslexia, etc., leads us to ask, what improvemnts to th way english is ritn cud most benefit those ho most sufr th burdns of t.o. at presnt? This articl is concernd with th secnd of these aspects.

1.2 Diagnosing th ils of t.o.
It has been recognised for centuris that inconsistncy of sound-symbl corespondnce in ritn english lies at th root of its dificltis. Wat has until recently not been so carefuly considrd is wich particulr inconsistncis users find most trublsm. If we ar now proposing reform by stajes, we need to be clear about th ansr to that qestion, so that we can decide wich featurs of t.o. ar most urjntly in need of treatmnt in a Staje 1 reform. Rathr than simply concluding that, for instnce, long vowls shud hav priority because, in systemic terms, they ar especialy prone to inconsistncy, we shud examn th errs most freqently made by users. Wile lists of individual words (like any, friend) wich cause beginrs special dificity hav long been in existnce [4], it is no less importnt to analyse th dificltis of mor advanced users resling with a ful adult vocablry, to se wat patrns of dificlty can be identifyd. A convenient sorce of relevnt data ar corpora of speling mistakes, of wich thre wil be investigated in this articl, to test th hypothesis that it is redundnt letrs wich ar most comnly mispelt and that CS therfor ofrs th gretst benefit to users as a Staje 1 reform.

1.3 Err corpora.
Alredy an amount of mispeling-data is availbl. Adam Browns paper at th Societys 1987 confrnce [5] examnd th errs made by two categris of foren lernr, Ugandns and Singaporeans. And Roger Mitton of Birkbeck College, University of London, has assembld corpora of mispelings from a wide ranje of nativ-speaking and foren sorces. They ar now held by the Oxford Text Archive [6], and include our Corpus 3 (se Section 4 belo), as wel as one drawn from th same sorce as our Corpus 2 (se Section 3 belo).

1.4 Definition of mispeling
We may define mispelings as forms not found in dictionris, but this definition is not always adeqat. Thus, if dictionris giv rare alternativ or historic forms (comick for comic), dos ther use constitute a mispeling? And if a british riter uses an americn form (gray for grey), is that an err. Then ther is th qestion of word-boundris - is eye lashes a 'corect' speling since it contains th ryt letrs? And dos misuse of apostrofes constitute a mispeling? Somtimes th distinction between a gramaticl and a speling mistake is blurd: is was a mispeling of were in th frase we was ... ? These uncertntis mean that th statistics givn belo ar subject to a marjn of err, tho th brod trends reveald ar not therby invalidated.

2. CORPUS 1: UNIVERSITY STUDENTS.

In ther 1987 final Onrs degree examnation 41 british university students had to translate and abstract texts from jermn into english, each student riting about 450 words. Som 50 mispelings wer noted in th scripts: 15 simplifyd a dubld consnnt, and 2 dubld a singl consnnt; 14 mispelt post-accentual and 2 pre-accentual shwa; 2 wer morfofonemic inflections, 1 confused <-able/-ible>, 1 concernd a silent letr, 1 was an omitd apostrofe - and phenomena was ritn for phenomenon 6 times! 22 of th mispelings wer shortr than t.o., and only 11 longr, a fact that if replicated sujests that shortr spelings com natrly and wud be of asistnce in avoiding err. In al, som 70% of th mistakes wud probbly not hav ocurd if th students had ritn by CS rules: ther wud hav been far less uncertnty about consonnt dubling, post-accentual shwa, silent letrs and morfofonemes. Th foloing table ilustrates th patrns, listing errs in 4 grups:

1. errs that CS wud not hav helpd prevent,
2. errs wich CS rules cud hav preventd (markd x),
3. errs shoing som CS featurs (markd ->), and
4. ful CS forms (markd =).



1.
a) T.o.
affect
it's
phenomenon
preceding
unperturbed
b) Mispeling
effect
its
phenomena(6)
preceeding
unperterbed
c) CS form
afect
it's
fenomnn
preceding
unpeturbd
2. coalition
independent
negligible
preamble
resurgence
separate
coallition x
independent x
negligable x
preambel x
resurrgence x
seperate x
coalition
independnt
neglijbl
preambl
resurjnce
seprat
3. accommodating
embellishments
historiographically
occurrences
accomodating->
embelishments->
historiographicly->
occurences->
acomodating
embelishmnts
historiograficly
ocurences
4. connotations
forefront
implies
openness
conotations =
forfront =
implys =
openess(8) =
conotations
forfront
implys
openess


3. CORPUS 2: DAISY ASHFORD.

A rich sorce of mispelings (despite dout about ther total authenticity, th text having been transcribed for printing) by a hyly litrat yungr riter is Daisy Ashfords story The Young Visiters (sic) [7], wich she rote about 1890 at th aje of 9. It is a hilarius mixtur of naïvety and sofistication, cliché-ridn romanticism and realistic, even cynicl social observation, to wich abundnt mispelings giv add spice. Som persistnt errs wer non-alfabetic: unconventionl word-boundris (21 split words like bull rushes, out stretched, and thankyou always ritn as a singl word) and th omission of apostrofes in cant, dont, wont and for posession. Altogethr 444 alfabetic errs wer noted over 69 brief pajes. Many wer repeatd (rarther 35 times!), and som words wer mispelt in two or thre ways (jellous, jelous, jellus; pleashure, pleshure, plesure), but 218 mispelt word-typs wer noted. Th errs fal into th foloing categris (numbrs indicate freqency, astrisks sho CS forms):

3.1 Pronunciation and accent.
A few spelings conforming mor predictbly to standrd pronunciation wer substituted for t.o. forms, thus bronkitis, convay (2), peaple (9), swollowed (2). survayed (2), but a larger number reflect th authresss RP accent. Thus it is a non-rotic speakr ho inserts <r> in gaierty (10), idear (7), ourght, perlitely (2), rarther (35), socierty (4), thourght (11), thorght (2), varse, and omits it in supprised (6); and only a non-hwayr (i.e. one ho dos not distinguish initial <w, wh>) wud hav ritn wiskers, whicker; and horsly for hoarsely implys no distinction between hoarse, horse.

3.2 Vowls.
Among vowls, th many posbl spelings of /i:/ caused most errs: beleeve, coffie, compleat (3), egerly (2), here (2, for hear), mystearious (4), peaple (9), quear (5), sceenery, seiezed, siezing, searious (2), supearier, viacle (4). Exept for shwa, no othr vowl was mispelt mor than 5 times, th worst being <ou> for <ow> in broun, doun (2), douny, goun.

3.3 Shwa.
Shwa (especially post-accentual) is freqently mispelt. Actual CS forms with sylabografic <l, m, n, r, bl> ar rarely used, but diffrence, choclates sho som afinity with CS, as dos sylabic <l> in musicle, portles, principle (for principal), tassles. But rithum, spasam sho a decided disinclination to use sylabic consnnts even wher t.o. dos. In jenrl shwa is spelt <u> far mor than in t.o.: blossum, consciunce, geniul (2), meershum, quiutly (2), passiun, radiunce, reveruntly, but elsewher is mostly spelt <e>. Cases with foloing <m> wer hansome (2, for hansom); with <n> bason, certin, cushon (3), Domonic, elegent, vacency, vishen; with <r> ancester (4), decerated, grammer, murmer (7), pillers, razer, resterant, somber (2, th US form), superier (6), theater (3, th US form). Few of these errs wud hav ocurd in CS, wich regulises with blossm, meershm, revrntly, hansm, certn, cushn, Domnic, elegnt, ancestr, decrated, gramr, murmr, pilrs, restran(t?), sombr, superir, theatr. Post-accentual shwa befor othr consnnts was mispelt most ofn with <i>, somtimes with <a>: laudible, salid, monagram, elligant, marrage, dowigers, languige; softist, tam a shanter, nativaty, crickit, privite (6), toilit (2), velvit (9). Of these, CS wud only hav helpd with laudbl. Pre-accentual shwa was mispelt far less freqently, but th foloing errs wer noted: mustache (2, th US form), perlitely (2), purplexed, purspired, supprised (6).

3.4 Singl consnnts.
Among consnnts, most errs ocurd predictbly over th varius spelings of /k/, th two valus of <g>, and th palatlized consnnts. For /k/, th oldr <-ick> ending prevaild over modrn <-ic> (domesticks, elastick) and aristockracy, exacktly, knickerboccers, majestikally also ocurd. Th dual use of <g> for /g, ʤ/ and th dual speling of /ʤ/ with <g, j> caused confusion in vage, gorgous, vengance, majic (2), sandwighs (for sandwiches). By reguirizing /ʤ/ as <j>, CS resolvs this problm, riting vage, gorjus, venjnce, majic. Palatlized <s, t> wer ofn spelt <sh>, wethr voiced or not, as in parshial (2), pleashure (2), profeshion, professhon, tishu, vishen, presumshious, sumpshous (3). Th simplr CS forms plesur, tissu, presumtuus, sumtuus wud hav been mor redily lernt than th t.o. forms, but wud not othrwise hav ofrd systrnatic asistnce.

3.5 Consnnt dubling.
Many errs betrayd uncertnty about consnnt dubling. Dubld consnnts wer simplifyd as in CS in adicted, adress*, anounced*, bagage, cariage (3), caroled, chiruped, disapeared, embarassment, flabergasted, hurah, inteligent, necesarry, pasengers, rumaged, shruged, sugested, wraped; but singl consnnts wer dubld aftr short vowls in alpacka, ballance, bouquett, bussy, coupple (2), elligant, ettiquett, finnished, gallopp (2), habbits, hastilly (4), jellous (3), necesarry, pallace (3), and th sufix <-ful> became <-full> in beautifull (2), faithfull, painfull. Such errs wud be unlikely in CS.

3.6 Redundnt letrs.
Othr errs reflect a natrl tendncy to omit redundnt silent letrs (as in CS): aile, bissness, brekfast*, color* (2, US), desperatly, ettiquett, favorite (US), hankerchiefs, hansome, honymoon* (2), kidnys(*?), lether, lovly* (7), merangs (2), mesure, mustache* (US), nervus*, pius* (2), plesant (3), plesure (3), ruge (5), straitening, swet*, vage*, varius*. But a few words ad an extra redundnt <e>: crystale, parasole, proposals, strolle, truely, weare.

3.7 Morfofonemes.
A tendncy to inflect morfofonemicly was noted, chiefly in th past tense of verbs as in CS: arrayd, cheerd*, disserppeard, continud*, marrid (6), obeyd*, reappeard (2), towzld, traveld, valud*; but also in becomeing, desireus, familys, partys. But in dose (3, for does), satisfide th morfeme was not recognised or representd as such at al.

3.8 Miselaneus.
Homofones wer ocasionly confused: fourth (for forth), main (for mane), right (for rite), role (for roll), there (2 for their), and with eat (4 for ate), hair (for heir), were (for wear) ther was confusion of sound as wel. Only 22 of th 444 errs wer not covrd by th abov categris, and mostly involvd miselaneus misrepresentation of vowls or randm slips.

3.9 Cut Speling implications.
Unlike Corpus 1, th Corpus 2 errs use marjnly (7) mor letrs than t.o. But since 97 extra letrs arose from just two typs of err (70 extra <r>s as in rarther, and 27 rongly dubld consonnts), we can stil say that th errs sho a widespred tendncy to shortn words. But mor importntly than th shortning, we se that many of th err-pattns themselvs mor or less directly mach th patrns of regulrisation introduced by th 3 cuting rules of CS: fonograficly redundnt letrs, post-accentual shwa and dubld consnnts.

4. CORPUS 3: TEENAJERS.

Mor representativ than th preceding corpora (which derive from hyly litrat individuals) is th foloing analysis, based on 163 qestionairs completed in 1981 by a cross-section of fourth-year pupils at a Nottingham comprehensiv scool (avraj age 15, 50% of each sex), as part of a survey of teenaje atitudes [8]. Th respondnts ranjed from very competnt spells to th virtualy ilitrat. Th analysis aimd to establish typs of err, but only aproximat relativ freqency, an exact count being imposbl for such reasns as ilejbility, unclassifybility, overlap, repetition and variation of errs, etc. Th ful listing is availbl from th Oxford Text Archive [6]. We here qote only ilustrativ or significnt exampls.

4.1 Primary categris.
In total 1,377 errs wer classifyd, but th numbr of mispelt words was fewr, a form like manors for manners being classifyd both as a vowl and as a consnnt err. 707 errs involvd vowls, 307 consonnts, 233 word-boundris (e.g. kind ness, alot), and 130 apostrofes.

4.2 Vowls.
Of th 707 vowl-errs, 184 involvd th letr <e> or its asociated sounds, 167 involvd shwa, 124 <i> or <y>, 78 <o>, 69 confused digrafs representing difthongs (e.g. jion for join), 63 involvd <u>, and 22 <a>.

4.2.1 <e>. Th prepondrnce of errs involving <e> is lajjly explaind by th freqency of th notorius silent <e> in english; in fact 109 of th 184 <e> errs involvd silent <e> in one way or anothr. In 44 cases a redundnt final <e> was omitd, thus: aprehensivly*, befor*, lovly*, somthing*, oposit*; a dozn of them used th morfofonemic past tense inflection: gatherd, marrid, murderd, lockd*; such forms wud not hav constituted errs in CS, wich makes a point of omiting such redundnt letrs. In anothr 44 words a redundnt silent <e> was addd (arguements, disrupte, senseably, learne, withe); such errs wud be far less likely in CS, wher silent <e> is nevr redundnt. But even wen a silent <e> is not redundnt, our riters ofn faild to apreciate th fact, omiting it in 20 cases wher it indicates a preceding long vowl (besids, completly, guid, entirly, somtims). Like Daisy Ashford, these riters also ofn mispelt long <e> (/i:/), as in areana, beeing, seariously, treet, leve*, se*, peple*, somtimes using forms that wud probbly hav been avoidd in CS (acheive, bereivement, peices CS acheve, berevemnt, peces). Short <e> was predictbly mispelt in such trap-words as agenst, enyone, and th CS forms frend, helthy wer used 5 and 3 times respectivly. Whether came out twice as weather, twice as wheather; CS wethr for both whether, weather wud obviusly hav helpd prevent those errs.

4.2.2 Shwa. Of th 167 shwa-errs only 24 ocurd pre-accentualy (e.g. intulectual, imagration, affence) wher CS ofrs no help. Of th 143 postaccentual cases, CS cud hav helpd in al but 27 (e.g. not with ernbarris, fortunitly, reletivs, confedence); actual CS forms (lisn*, famly*) wer rare, but th regularity of CS wud hav grety reduced th likelihood of err in probabal, resmoserbl, senscebul (CS probbl, responsbl, sensbl); dismell, hassel, littel, modle, travill, vandle, vanderlism, vandolism, famaly, famley, famliy, persnaly (CS disml, hassl, litl, modl, travl, vandl, vandlism, famly, persnly); hansem, hansum, enames, enimes, enimies (CS hansm, enmis); certian, serton, couson, lisen, samariton, pleasent, pregnent, redundent, differant, present, perminently (CS certn, cusn, lisn, samaritn, plesnt, pregnnt, redundnt, difrnt, presnt, permntly); (bachler, figer, neighboor, tempar, militery, misarable (CS bachlr ,figr, neibr, tempr, militry, misrbl).

4.2.3 Othr vowls. Th comn confusion of <-y, - ey> endings (mony, angrey, badley) wud be avoidd in CS, wich spels them al <-y>. Also comnly confused wer <i, y> especialy in inflections (enemys, orgi's, studiing, emtyed, erlyer, trys*); here CS ofrs only partial morfofonemic disambiguation (CS deny, denys, denyd, pity, pitis, pitid). Perhaps therfor CS shud extend its limitd regulisation of <y> for /ai/ (CS myt, desyn, hy), so that <i> is only used for th fonemes in piti, studiing. CS bild, busness wud howevr hav helpd riters avoid biuld, buisness. CS wud hav helpd less comnly with <o> or <u> errs, tho som CS forms did ocur: no* (11, for know), corse*, cruse*, nutral*, and th foloing forms sho a CS tendncy at least: sholders, contry, dose, norish, truble, beutiful (CS sholdrs, cuntry, dos, nurish, trubl, butiful. Among th difthong-digrafs, th 15 mispelings of their/there point strongly in th direction of th CS merjr as ther, a form wich ocurd twice. CS wud not hav helpd with th 22 mispelings involving <a>, exept that CS gards wud hardly hav produced th mispeling gaurds.

4.3 Consnnts.
Silent consnnts caused 37 errs. <gh> was predictbly th bigst trap (17, such as cought, eigther, figthing, laugth, fite, lafe, neibour, tuff); in CS these errs wud hardly arise. Sibilants (24 errs with <c, s, or z>, as consentrate, exsept, resently, sertain, conciderate, percon, criticizm, wize) and <h> (42 errs; omitd as in disonist, macanic, or insertd as in hair, ham, hearn, whould, whorthwhile, whent) wer also trublsm; CS helps rarely with sibilants (exept, persn), but by geting rid of silent <h> and reducing initial <wh> to <w>, CS resolvs much of th <h> problm. Howevr, as in Corpus 1 & 2, consnnt dubling caused most dificlty, with 80 rong dublings (e.g. habbit, accademically, beggining, abillity, allmost, fammily), and 84 rong simplifications (e.g. rubish*, inteligent, comunity*, anoying*, hapiness*, wory*). By simplifying nearly al dubld consnnts, CS makes virtualy a clean sweep with <b, g, l, m, n, p, t> and wud help with <c, f, r, s>. Howevr our riters somtimes rongly simplifyd <cc> (=/ks/ as in acident, sucess, and they confused of/off 5 times (CS keeps <ff> in off). Lastly, palatlised <s, t> caused mispelings like fassion, accationally, proffeson, recetion, manutriction, posistion, relasions - CS rarely helps here (exept perhaps fashn).

4.4 Apostrofes.
Othr comn errs involvd apostrofes and word boundris. CS can do nothing for th latr, but is trying out omission of apostrofe for posession and in negativ contractions like wouldn't (th err would'nt ocurd 6 times in our corpus). If th apostrofe cesed to be used to indicate posession, then th widespred confusion with between <-'s>, <-s'>, plural <s> and presnt tense <s> wud be overcom: in our corpus 50 plurals and 14 presnt tense verb endings wer ritn <'s>, and th tru posessiv apostrofe was omitd 17 times. If <'s> wer only used for contractions of is, as in he's, it's, this dificlty wud larjly disapear.

5. CONCLUSIONS.

5.1 Wat CS acheves.
Our thre corpora hav here provided substantial evidnce that if english speling wer reformd on CS lines, a huje numbr of speling mistakes wud be avoidd. Al thre of th cuting rules target a specific major problm area in t.o. Pronunciation givs no clu as to th use of silent letrs, nor to th choice of letr for post-accentual shwa, nor for consnnt-dubling - and it is precisely in these areas that CS ofrs th gretst relief. Othr notorius speling-anomlis of t.o., such as th vowl letrs used in comn words like any, woman, son, busy or to represent long vowls, pale into insignificnce by comparisn, at least when jujd by th criterion of err-freqency as found in th thre corpora examnd in this articl.

5.2 Furthr qestions.
This latr point dos howevr rase an importnt qestion wich must be directd at CS, and wich is also reflectd in th most serius omission from this study: an analysis of errs made by primary-scool lernrs, for hom th comnst, most elementry words in th languaj ar th first stumbling blok. Certnly CS helps with many such words (e.g. CS frend, hed), but othrs it leves unchanjed. For th beginr, therfor, i.t.a. or AMERICAN may ofr an even mor efectiv systm. Here is an importnt matr to be pursud. Furthr reserch or reporting-bak myt also be intresting on th foloing points:

1. yung lernrs: th Oxford Text Archive has som corpora of mispelings made by primary-scool childrn, and som readrs may hav access to or be able to colect furthr material.

2. childrns inventd spelings: studis by Anne Robinson (Manchestr Polytecnic) cud be relevnt.

3. speling mistakes made in othr european languajs: dos John Downings Comparative Reading contain information on speling?

4. anglo-welsh experience: dos lerning th mor regulr welsh orthografy help childrn in bilingual scools mastr t.o.?

Ther is stil much work to be don.

REFERENCES.

[1] Walter Ripman & William Archer New Spelling, London: Sir Isaac Pitman & Sons, 1948, 6th edition revised by Daniel Jones & Harold Orton.

[2] Axel Wijk Regularized English, Stockholm: Almqvist & Wiksell, 1959.

[3] Harry Lindgren Spelling Reform - A New Approach, Sydney: Alpha Books, 1969.

[4] e.g. th Queensland Department of Education's 80 Words of Special Difficulty (Demons) listed in Garry F Jimmieson 'Spelling Reform 1 - And Nothing Else!' in Simplified Spelling Society Newsletter, J2 Spring 1986, pp.22-23.

[5] Adam Brown 'English Spelling and Non-Native Speakers', to apear in Journal of the Simplified Spelling Soc., 1988, [published as 'A Singaporean Corpus of Misspellings' in J9 1988/3 pp6-10].

[6] Text Archive of the Oxford University Computing Service, 13 Banbury Road, Oxford OX2 6NN.

[7] Daisy Ashford The Young Visiters, London: Chatto & Windus, 1919.

[8] An analysis of th qestionairs was publishd as part of a much mor extensiv socio-sycolojicl survey of yung peples atitudes by Cyril Simmons & Winnie Wade as I like to say what I think, London: Kogan Page, 1984.


Back to the top.