[Journal of the Simplified Spelling Society, J29, 2001 pp19-25]

Perceived English spelling rules for vowel sounds in single syllable words.

by Peter Whitmore.

Peter Whitmore started his career as a chemical engineer. He currently manages a company, started with his wife in 1981, that publishes classified advertising. He has lived in New Zealand, Australia and Canada. Because of his long-standing interest in English spelling, he has been an SSS member for several years and, more recently, a member of the SSS email group.

Abstract.

In order to test what people perceive to be the inherent spelling rules that govern English, 20 participants were each asked aurally to write down 108 non-existent words. The spellings were then analysed. The test was confined to vowel usage in single-syllable words. The results showed strong consensus about the spelling of the short vowels (a, e, i, o and u) and, when they are followed by single consonants, the long vowels (a, e, i, o, and u as in both "fume" and "boot"). There was also consensus about the long a, e, o and u (goo) as word endings, about the long e, and about u (mooch) sounds before consonant clusters. The same was true of the ar and or sounds in all positions, as well as the ou (now) sound as a word ending. Addtionally, the results provide support for some non-traditional spelling patterns. Further work based on this technique might yield useful results in other areas of English spelling. A spreadsheet template for conducting this test is available from the author.

Introduction.

One way to simplify English spelling would be to build on the existing spelling patterns in the language. Ideally, to do this so that the resulting text looks familiar and is easy to read, one would need to adopt patterns that people intuitively feel comfortable with. At the same time, the patterns must fit together to produce a clear, consistent and cohesive spelling system.

The main problem with this approach seems to be deciding which patterns to adopt. For example, the long o sound at the end of a word is commonly spelt as oe (toe), ow (flow) or just o (go). Which, if any, of these is the appropriate form to accept? How does one take into account that there are a limited number of words in very high usage that end in o, such as go, so and no, but many more words that end in ow, such as flow, slow, and glow, which tend to be less common?

In this study we sidestep the difficulties of working from an analysis of the existing written language. Instead, we test for what people intuitively perceive to be the appropriate pattern for spelling certain sounds. The study was limited to vowel sounds in single syllable words.

Method.

Twenty people were each asked to write down a series of 108 non-existent, one-syllable words that were read out to them, using whatever spelling they felt was clear and obvious. Although the objective was to look specifically at vowel sounds, the participants were not told this.

The participants were all female, with good spelling compared to the general population. (They are all employed to take classified advertising for a New Zealand publication.) They generally have a New Zealand style of speech, which is somewhat similar to that in Southern England. In particular, with relevance to this study, they do not tend to pronounce the letter r strongly or at all in words like hard, shirt and pork. They also use the long soft a in words like class, dance and aunt, and pronounce the long u in words like due and neutral with a yu sound.

Three different words were asked for each vowel sound and each position of the vowel in the word examined, except in the case of the short vowel sounds, a, e, i, o and u, for which only one word was asked. An attempt was made not to choose words which, when read out, were obviously reminiscent of existing words.

For presentation to the participants, the words were grouped. Each of the tables below represents one group. For example, the first group is short vowels in the medial position in the word. Within each group the words were given in a mixed order so that words requiring the same vowel sound were not asked consecutively.

The test was first tried on four people who were not part of the main study. This trial led to some minor modifications to the words used, where these were being misinterpreted. Because the results were so consistent, it also lead to a reduction, from three words per vowel sound to one word per vowel sound, for the short vowels a, e, i, o and u.

The vowel sounds included in the study were:
The short vowels a, e, i, o and u
The long vowels, a, e, i, o u (yu, fume) and u (oo, boot)
The sounds ar, air, er and or
The vowels aa (palm), au (caught), ou (shout) and uu (book)

The trial showed that special care was needed in presenting words containing the ar, er, or and air sounds, the aa and au sounds, and the st word ending. For the vowel-plus-r sounds, I explained to the group how US pronunciation would sound the r in words like 'start', and I then pronounced these words with (my best effort at) a US accent. For aa and au words, I explained that in US pronunciation one would not hear an r sound in these words. There was a tendency for people to spell the st ending as sed (as in used) and I had to explain that such words did not contain a d sound.

The trial showed an unexpectedly strong propensity to give spellings involving the gh pairing, which in combination with other letters can produce a wide variety of vowel sounds. Because I did not believe that such spellings would be useful in simplifying English, I suggested to the participants in the full study that they steer away from such spellings unless they clearly thought them to be the best option.

One particularly literate woman pointed out that two of the words actually did exist - thew and dorp. However, no one else tested knew these words, so the effect on the test results would have been minimal.

Results and discussion.

Spelling patterns that received less than 5% support are reported as "other." Replies where the participant had changed the form of the word, so that it did not conform to what was being tested, were discarded from the analysis.

The support for a specific spelling pattern is determined by the number of words written using that pattern, as compared to the total number of valid replies. For example, the result of ee (76%) for long e in the medial position followed by a single consonant means that 76% of the valid words for the vowel sound in that position contained the ee spelling.

In these results an arbitrary consonant is represented by C and a cluster of two consonants by CC.

The short vowels, a e i, o and u.

The test for these vowels was done only for the medial position within a three character word. (Short vowel sounds do not normally occur as word endings, and their spelling is not normally affected when they precede consonant clusters.) The results are shown in Table I.

Table I.

Vowelshort ashort eshort ishort o short u
Model wordmatmetbitmopmud
Test wordpabteppifyomfub
       
Resultspab = 20tep = 19pif = 11 yom = 19fub = 18
 total = 20other = 1piff = 8 other =1fubb = 1
  total = 20other = 1 total = 20other = 1
    total = 20  total = 20


Using the case of short i as an example of how to interpret the Table I, the vowel is pronounced as in the model word 'bit' and the word asked was 'pif'. There were 20 valid replies of which 11 were 'pif', 8 were 'piff' and one was another spelling.

The results show that there is a very clear understanding of how the short vowel sounds should be spelt (though not such a clear understanding about doubling of the trailing consonant). The four answers that did not follow the pattern can probably be put down to mis-hearing of the word required.

The long vowels a, e, i, o, u (yu) and u (oo)

For all the remaining cases, including this one, each participant was asked to spell three different words for each vowel sound, and for each position of the vowel in the word. The results for the vowel in the medial position, followed by a single consonant, are given in Table II.

Table II.

Vowellong along elong ilong o long u (yu)long u (oo)
Modelmatemeetlinenote fumeboot
Test wordsvateleebribetobe pyudemoog
 rabeveenkimerofe vyuneroop
 mavereepyivevoze nyupezoob
ResultsaCe = 50ee = 44iCe = 57 oCe = 36uCe = 36oo = 38
 ai = 4ea = 4other = 2 oC = 6ewC = 8uCe = 15
 other = 5e = 4total = 59 ooCe = 4euC = 6ouCe = 3
 total = 59ei = 3  other = 14other = 7other = 4
  other = 3  total = 60total = 57total = 60
  total = 58     


While the results are not as unequivocal as for the short vowels, there is clearly a very strong understanding of how the long vowel should be formed. For long a, i, o and u (yu) the predominant patterns were aCe (85%), iCe (97%), oCe (60%) and uCe (63%). For long e and u (oo) they were ee (76%) and oo (63%).

In all cases support was much higher than for the next most popular spelling. Note in particular that the fairly common ea spelling for long e gained only 7% support. Surprisingly, for the long u (oo) sound, the uCe spelling gained 25% support, even though this spelling would normally make the yu sound.

The results for the spelling of the long vowels at the end of a word are shown in Table III.

Table III.

Vowellong along elong i long olong u (yu)long u (oo)
Modeldaytreepiego duetoo
Test wordschaydreesnie clogyufroo
 blaysteeclie drothyuvoo
 vayvreezie votyuthoo
Resultsay = 43ee = 48i = 24 o = 39ew = 17u = 30
 ey = 3e = 3y = 16 oe = 7u = 16oo = 20
 a = 3other = 9ie = 10 ough = 6ue = 10ew = 3
 other = 10total = 60igh = 3 ow = 5eu = 4ue = 3
 total = 59 other = 7 other = 3ui = 4other = 4
   total = 60 total = 60other = 6total = 60
      total = 57 


The long a, e and o endings, ay (73%), ee (80%) and o (65%) received strong support. Results for the other vowels were not so clear. For long i, the i ending (40%) had moderate competition from y (27%) and ie (17%). For long u (yu), the ew ending (30%) narrowly edged out u (28%) with ue (18%) further back. And for long u (oo), the u ending (50%) was moderately challenged by the oo ending (33%).

What is particularly interesting in these results is that the participants are strongly putting forward very simple spellings that are not actually common in current English. For long i, their preferred ending i (hi) is far less common in English that the y (fly) and ie (pie) endings. For long o, their strongly preferred o ending (go) markedly outstripped oe (12%) and ow (8%), even though this spelling appears in only a small number of current English words. For long u (yu), the untraditional u ending (28%) gained almost as much support as ew (few, 30%), and substantially more than ue (due, 18%). And for long u (oo) the uncommon u ending (guru) clearly outranked oo (goo, 33%), while ue (blue, 5%) received very little support.

The results for the vowel in the medial position, followed by a consonant cluster, are given in Table IV.

Table IV

Vowellong along elong I long olong u (yu)long u (oo)
Modelfaithleastmind coachneutralmooch
Test wordskaichfeechliich goathpeuchbooch
 vaistreestviistvoast feuthnoost
 maishgeethmiith loachveustcoosh
Resultsai = 13ee = 31I = 14o = 14 u = 15oo = 47
 aCCe =13e = 8iCCe = 12 oCCe =14eu = 11ooCCe =3
 a = 10ea = 5y = 10 oa = 9ew = 8other = 6
 atCC = 5other = 9ie = 7 ow = 6uCCe = 8total = 56
 ay = 4total = 53yCCe = 4 oo = 4other = 12 
 aa = 3 other = 7 oe = 3total = 54 
 other = 6 total = 54 other = 5  
 total = 54   total = 55  


While there is strong consensus about the ee spelling (58%) for long e, and the oo spelling (84%) for long u (oo), consensus regarding the other vowels is much less clear. This is perhaps not surprising given the variation in spelling patterns English uses for such words.

For long a, the spellings ai (faith, 24%) and aCCe (bathe, 24%) received equal support. For long i, the single i spelling (mind, 26%) edged out the iCCe spelling (tithe, 22%). For long o, the o (most, 25%) and the oCCe (clothe, 25%) spellings gained equal support, while oa (roast, 16%) was somewhat behind.

The stressed long u (yu) sound before a consonant cluster is uncommon and may not occur at all in single syllable words. The two most popular spellings chosen were u (fuchsia, 28%) and eu (neutral, 20%).

Even though in current English the use of a modifying e after the consonant cluster is rare for some of these vowels, the approach gained some support in all cases except for long e . This is probably a reflection of how strongly understood the pattern is for words in which a single consonant follows the vowel.

The ar, air, er and or vowel sounds.

The results for these sounds in the medial position are summarised in Table V.

Table V.

Vowelaraireror
Modelpartcairnherdcord
Test wordsparbmairtgernnorb
 larnvairnlerfdord
 zarftairbmertzorp
Resultsar = 54er = 21ur = 23 or = 51
 other = 6air = 5er = 22 o = 3
 total = 60ear = 5ir = 3 other = 3
  other = 16e = 3 total = 57
  total = 47 other = 8 
   total = 59  


The results show strong consensus about the ar spelling (90%) and or spelling (80%) in this position.

The air sound caused some difficulty, perhaps because it is not very common other than as a word ending. The preferred er spelling (45%) is not traditional. In the medial position, neither is the ear spelling (11%), though it received the same support as the air spelling (cairn, 11%).

The three common ways of spelling the er sound all received support, with ur (39%) edging out er (37%) and ir (5%) trailing well behind.

The results for the vowel-plus-r sound as word endings are summarised in Table VI.

Table VI.

Vowelaraireror
Modelfarfairherfor
Test wordsvarsairnervor
 nargairgerzor
 larnairzerblor
Resultsar = 38ear = 17ur = 18 or = 35
 arr = 9are = 9er = 15 ore = 12
 are = 4air = 8ir = 12 orr = 5
 a = 4err = 4urr = 3 o = 3
 other = 4er = 4e = 3 other = 5
 total = 59ere = 4other = 7 total = 60
  eer = 3 total = 58 
  ir = 3   
  other = 8   
  total = 60   


As with these sounds in the medial position, there was strong consensus about ar (far, 64%) and or (for, 58%).

Support was split among the three common spellings of the air ending, ear (bear, 28%), are (care, 15%) and air (fair, 13%). Support for the er ending was split in a similar manner to the case when this sound appears within the word, with ur (fur, 31%), er (her, 26%) and ir (sir, 21%).

The aa, au, ou and uu sounds.
The results for these sounds in the medial position are summarised in Table VII. The aa, au, and ou sounds were not tested before a consonant cluster because these sounds are typically made by a two-letter group that preserves its pronunciation, whether followed by a single consonant or a cluster.

Table VII.

VowelaaauouuuCuuCC
Modelpalmcaughtshoutbook push
Test wordspaabdaupfoubchuub muuth
 vaatpauggouptuup fuunt
 saanzaulbougzuud zuuch
Resultsar = 20or = 21ow = 14 oo = 23u = 19
 aa = 19aw = 9ou = 13 u = 14oo = 14
 a = 7all = 7au = 7ou = 6 o = 10
 aCe = 6ou = 3ouCe = 5 au = 4oot = 4
 other = 7au = 3al = 4 o = 4other = 11
 total = 59oo = 3el = 4 uCe = 3total = 58
  other = 14other = 6 other = 6 
  total = 60total = 53 total = 60 


Despite telling the participants that none of the test words in this group contained an r sound, for the aa sound the ar spelling (34%) narrowly edged out the aa spelling (32%). The same situation was repeated for the au sound. If you remove the all spelling (which makes sense in the test word zaul but not in the others), then the or spelling (40%) gained more support than the aw spelling (pawn, 17%), with the au spelling (daub, 5%) well behind.

Support in the ou sound was almost equally divided between ow (town, 26%) and ou (shout, 25%).

Because the uu sound can be spelt in English with just the letter u, it was tested prior to a consonant and prior to a consonant cluster. The leading choice prior to a single consonant was oo (book, 38%), followed by u (put, 14%). Prior to a consonant cluster the order of preference was reversed with u (push, 33%) and oo (whoosh, 24%).

Of the sounds in this group, only aa, au and ou appear as word endings. The results for these vowel sounds are summarised in Table IX.

Table IX.

Vowelaaauou
Modelbaapawnow
Test wordschaagaufou
 glaavaujou
 snaazauzou
Resultsa = 25or = 18ow = 26
 aa = 12aw = 16ou = 4
 ar = 11ore = 10other = 18
 other = 10o = 4total = 48
 total = 58other = 12 
  total = 60 


Again with the aa and au endings we have the difficultly of participants using spellings containing the letter r when there is no r sound. For the aa sound, the preferred choice was a (ma, 43%), followed by aa (baa, 21%) and ar (19%). For the au sound the or ending (30%) predominated, followed by aw (paw, 27%) and another r ending, ore (17%).

For the ou sound, there was a reasonable consensus for the ow ending (now, 54%), with no other spelling gaining significant support.

Conclusion.

While we may view the spelling of our language as rather chaotic, the study revealed a perception that it actually has some very strong rules, if we define a rule as a pattern that receives over 50% support.

The strongest and clearest of the rules relate to the spelling of the short vowel sounds, a, e i, o and u. However, there are also very strong perceived rules relating to all the long vowel sounds followed by single consonants, the long a, e, o and u (oo) word endings, the long e and and u (oo) sounds before consonant clusters, the ar and or sounds in all positions, and the ou sound as a word ending.

The strength of these results indicates that simplifying English by building on the underlying rules already present in the language is a real possibility. The results also indicate that, based on the perceptions of the participants in this test, there are areas where it is not clear what rule should be adopted.

In the broader context, the results also raise questions about how one ascertains what pattern should be adopted as a rule for the purposes of simplifying spelling when there are several alternatives. For example, after examining the existing vocabulary, one would presumably opt for y (try) or possibly ie (pie) as the appropriate rule for the long i word-ending, at least for single syllable words. However, the participants in this test opted primarily for i (hi), which is simpler.

These results need to be treated with some caution. The test was limited to 20 people in one area of the English-speaking world, and the response received depends to some extent on the test word given. Nevertheless, I believe they provide a useful guide, and that the same technique might yield useful results if extended to other areas of English spelling.

Back to the top.