[Spelling Progress Bulletin, Spring 1983 pp6-11]
[Robert Baker: see Journal, Bulletin.]

SSS Conference 3: Experiment and Research in Spelling Reform.

"An experimental study of attitudes towards English spelling reform,"

by Dr. Robert G. Baker,

Dep't. of Electronics, The Univ., Southampton, England.

Abstract.

Results are discussed of a number of experiments investigating subjective attitudes toward traditional English spelling. The basic paradigm of the experiments was presented at the International Conference of the S.S.S. in Northampton in 1979. Literate adult native speakers of English were presented with lists of English words and asked to respell them in more rational ways. These "reformed" spellings provide information on those linguistic aspects of orthography which people consider to be important. A selection of such results was presented at Northampton.

There is one critical problem in this research. The instruction to re-spell words is deliberately vague and different reformers may have fundamentally different views on what is meant by a more rational spelling. In order to explore this point further, the naive spelling reformers took part in structured interviews based on the changes they had made in the spelling reform task. Reformers were asked to explain, describe, compare and contrast all the reforms they had made.

A statistical analysis of the interview material allows us to make inferences about these reformers' general attitudes towards spelling reform. The technique provides useful insights into popular opinions about English orthography.

Corpus.

Introduction - the spelling reform task,
This paper continues and elaborates on the analysis and discussion presented in an earlier paper published in SPB (Baker, 1980). Both this and the earlier paper focus on the attitudes of naive spellers (i.e. non-specialists with no reforming axe to grind) to the conventions of traditional English spelling. The principle technique used in the investigations is the "spelling reform task" in which lists of English words are presented to naive "spelling reformers" with the instruction that they should first rate the words according to the "rationality" of their spelling, and then provide alternative "more rational" spellings for low-rated words.

The earlier paper presented some results from one such experiment. 23 undergraduates from Stirling Univ. took part in the experiment. The spelling reform task was carried out on a list of 111 English words specifically selected to represent the range of "non-phonemic" [1] spelling conventions in English. The undergraduates' reforms were scored according to whether or not they preserved or destroyed the conventions in question. (See Table 1) It is clear that some conventions, e.g. preservation of plural "s" in "dogs," are more highly valued than others, e.g. preservation of graphemic final '"-e" in "give."

This type of experiment may have various implications for the Simplified Spelling Soc. The responses of naive reformers may indicate "areas of least resistance" in popular attitudes towards English spelling. If gradual step-by-step reform is to be advocated (e.g. Lindgren, 1979), then a properly conducted social survey along the lines of the "spelling reform task" may provide guidance as to the order in which reforms could be made. The work of Kenneth Ives (1979) appears to be already pointing in this direction. In addition, the occasional use of such surveys throughout a period of actual spelling reform should serve as a useful barometer of change. It seems likely that the popular acceptance of any form of change in the current spelling system could radically alter the biasses shown in Table 1.

Footnote [1] : Following the analysis of Albrow (1972), Venezky (1970) and Chomsky and Halle (1968), the representational principles of English spelling can be seen as polysystemic, morphographemic or systemic phonemic, carrying grammatical and semantic information in parallel with phonological information.

Table 1.

Average rationality ratings and proportions of occasions when higher order regularities are preserved in spelling reform task


Rule Level
Average
Rationality Rating
% occasions rule
is preserved

Graphemic
give not giv
freeze not freez
3.71
4.28
42.6
67.8

Phonotactic
fetch not fech
wash not wosh
4.18
4.01
66.7
52.2

Morphemic
walked not walkt
dogs not dogz
3.22
4.71
60.9
87.9

Syntactic
goose not goos
add not ad
3.78
3.53
29.3
52.2

Semantic
seem and seam differentiated
sp. g retained in sign and signal
3.80
3.53
57.2
46.1

Another approach to gradual spelling reform could be through the introduction of lists of reformed vocabulary items rather than through the modification of conventions. The data from the spelling reform task show quite a lot of agreement between people on certain specific words. For example, "bomb" was respelt as "bom" by 100% of the students, "peach" was respelt as "peech" by 96% of the students. Whichever approach is taken, the collection of social survey data could provide valuable propaganda for the proponents of reform.

Individual approaches to the spelling reform task.


The data collected from the spelling reform task raise many questions. In spite of a degree of consensus over the reform of many words, there was still a great deal of individual variation; at least as much as there is between expert reformers. Some people were willing to reform many more words than others and some introduced interesting innovations to cope with certain items. For example, the use of alphabetic letter names to represent phoneme sequences (e.g. "c" for "sea"), the use of hyphens at morpheme boundaries (e.g. "gradyoo-al") or elsewhere (e.g. "ta-ee" for "tie"). It was therefore decided to attempt to find out more about how people were operating in the spelling reform task. The instruction to produce "more rational" spellings had been deliberately vague, in anticipation of the fact that people would differ in their conceptions of orthographic rationality. In order to gain a deeper understanding of individual approaches to the task, each of the 23 undergraduates were given a structured interview. Each person was asked first of all to try to explain or comment on each of the reforms he or she had made. These comments and explanations were then converted into five point rating scales. In a second session, each person was asked to rate each one of his/her reforms according to each of his/her rating scales. Thus if a person had produced the explanation "I'm spelling this the way I pronounce it" for a particular reform, he or she would be asked to decide to what extent this explanation was true of every reform he/ she had made. If it was definitely true of a particular reform, a rating of 5 would be given. If it was definitely untrue, a rating of 1 would be given.

Which people reform for which reasons?


A total of 47 different rating scales was elicited. The maximum any one individual produced was 12 and the minimum was 3. All the scales are presented in Appendix A since spelling reformers may wish to take cognizance of them all.

The most frequently produced explanations and comments were:
1. I'm removing unnecessary letters (17 people).
2. Some British people may pronounce this word differently from the way I've reformed it (15 people).
3. I don't like my reform (14 people).
4. I've spelt this the way I pronounce it (12 people).
It is worth noting that two of these comments reflect an uneasiness about the task. The other two demonstrate a fairly straightforward "phonemic" approach to the task.

In order to find out what sort of people produce which explanations, the 23 undergraduates were classified according to 4 separate objective measures. These are shown in Table 2. The mnemonic labels refer to aspects of reforming performance. The label Conservative is straightforward. Some people reformed far fewer words than others. The maximum number of reforms was 100 out of a possible 111 and the minimum was 17.

Table 2.

Performance of Undergraduate students.

Subject
Removing silent letters
Substituting for ambig. let.
Word better left unchangd
Reform is offensive
Aiming for economy of sym.
*Conservatives
3.0
3.0
3.2
3.0
2.7
Levellers
2.8
3.3
2.2
4.0
2.5
Linguists
2.8
3.0
3.2
2.0
2.7
Phoneticians
2.4
2.5
2.2
4.0
3.5

*Conservatives = few reforms, Levellers = reforms in direction of more frequent sound-spelling, Linguists = reforms to preserve linguistic rules, Phoneticians = consistent one sound=one symb.

Each person was placed in one of 4 status groups (A-D) within the Conservative classification depending on the number of words reformed. Thus Group A Conservatives produced relatively fewer reforms than Group B Conservatives, and so on.

The same procedure was used with the other three classifications. The Leveller classification refers to a measure of spelling reform "regularity." A frequency list of English sound to spelling correspondences (Hanna et al, 1966) was used to establish whether any particular reform represented a more frequent sound/ spelling correspondence than its original spelling. Thus the "ee" spelling in "peech" represents a more frequent sound/spelling than the original "ea". In this instance the spelling "peech" would be assigned a regularity score of 3. If on the other hand a word such as "greed" had been reformed to "gread", a regularity score of 1 would have been given. Correspondences of equal frequency would be assigned a score of 2. The scores were summed and averaged for each reformer to produce an index of the extent to which each reformer made changes in the direction of more frequent sound/symbol correspondences. Group A Levellers had higher average regularity scores than Group B Levellers, and so on.

Group A Linguists were those people whose reforms preserved or enhanced non-phonemic spelling conventions (as in Table 1) more often than Group B Linguists' reforms.

The classification Phonetician is essentially a measure of phoneme/grapheme consistency. A set of ambiguous sound/spelling correspondences was selected from the original word list, such as /k/=c or k, /s/ = s or c, /tsh/ = ch or tch. "Good" phoneticians were those people who consistently and unambiguously selected one or another of the alternatives throughout their reforms.

Comparing these groupings with the explanations and comments produced by the reformers, we find that some explanations are exclusive to certain types of people. Thus a Group A Conservative was the only person who produced the rather prescriptive and reactionary comment "This reform would be unnecessary if people spoke more clearly." Examples of reforms which were highly rated on this scale are "Ingland" and "wimin."

On the other hand, a Group A Phonetician was the only person who made the comment, "It is difficult to represent some of the sounds in this word without introducing new letters," showing a high degree of phonetic awareness. An example of such a word is "revision," where the comment presumably refers to the phoneme /ʒ/ which has no unambiguous representation in English spelling.

A Group A Linguist is the only person to make the comment "I have made this word more difficult to identify." This comment applies particularly to words where the reform produces a homograph, e.g. "maid" reformed to "made," "sign" reformed to "sine", "air" reformed to "ayr." This comment therefore demonstrates an awareness of semantic orthographic conventions, and perhaps a reluctance to break them down (the type of analysis discussed below confirms that this is the case).

However, many of the explanations are shared to a certain extent by people in all the subgroups. Table 2 illustrates some of the relative biases which people in any particular classification produce the comment or explanation in question. The scores in the table are calculated by assigning a rating to each person who gives the explanation, depending on which status group he/she belongs to within each classification. So, for example, a Group A Conservative producing any particular explanation will contribute a rating of 4 to the Conservative column; a Group D Conservative producing the same explanation contributes a rating of 1. These ratings are added up and divided by the total number of people producing the explanation. This gives an average bias with which particular explanations are produced by particular types of people. So, if all the people who produce a particular explanation are Group A Conservatives, it will receive a Conservative bias score of 4; if they are all Group D Conservatives, it will receive a score of 1. If any explanation was produced by all subjects, a bias score of 2.57 would result.

The explanations given in Table 2 are merely illustrative of this type of analysis. It should also be borne in mind that all the explanations discussed were given spontaneously. If we take "removing silent letters" in Table 2, we see that it is a fairly Conservative-biassed explanation. Those people who only make a few reforms are particularly concerned about "silent letters." Phoneticians, on the other hand, remove silent letters as a matter of course. It is a sine qua non of operating "phonetically" and so there is no need to mention it in their explanations. Phoneticians are relatively more likely to talk in general terms about "Economy of symbols."

"Substituting ambiguous letters" is a particularly common explanation amongst Levellers, and once again a sine qua non for Phoneticians.

The production of the comment "Better left unchanged" is equally biassed towards Conservatives and Linguists (disguised Conservatives?).

As for "Aesthetics," Conservatives and Linguists are probably less likely in the first place to produce reforms which they deem to be aesthetically offensive. However, only one person (in each group), (Grade A Leveller and Phonetician, Grade R Conservative and Grade C Linguist) actually produced this comment (See Appendix A). This type of analysis generally suffers as a result of the limited number of explanations shared by the subjects.

What do the comments and explanations mean?


A more powerful way of examining people's explanations of their own reforming behaviour is to plot the relationships between each individual's different explanations and comments. This can be done by taking each person's matrix of ratings (reformed words by explanations) and putting it through a principal components analysis on a computer. Principal components analysis is a statistical tool for describing the structure of matrices (see Slater, 1976-7).

A few examples will be given to illustrate the technique. In an ideal world, a perfect spelling reformer would produce perfect spelling reforms. He/she would be completely consistent in producing reforms which were phonetically sound, aesthetically satisfying, economic on symbols, and so on. In such a case, the computer would be unable to carry out a principal components analysis. In practice most people are less than perfect.

Figure 1.
Grade A Conservative
Better
uncha






Having
breaking

Aiming for economy
of symbols






Making more
regular






Figure 3.
Grade A Linguist







Rationality of
original
Removing super-
fluous letters



 










 

left
nged

Reform obscures
word origins

Rationality
of original
difficulty
old habits


Improving order
of letters


Making less confusing
for learners









My reform could
be pronounced wrong
Don't like
this change
Spelt as I
pronounce
Some people may
pronounce this
differently from
my reform

Non-phonological
considerations
apply in original





Substituting for
ambiguous letters

I think I had
problems with
this word as
a child
Figure 2.
Grade A Leveller












Spelt as
sounds
Easier to
learn to
read
Change
aesthet
offen


Substi
ambigu




Figure 4
Phonetician
Making easier
for foreigners
to learn
Reform looks
better than
original







Spelt as I pronounce



Difficult to repre-
sent some sounds


My reform could
be pronounced wrong



Don't like
this change


Rationality
of original








Removing superflous
letters
is
ically
sive
Easier to
learn to spell
tuting for
ous letters
Aiming for
economy of
symbols









Spelt as
it sounds

Rationality of
original


Some people may
pronounce this
differently from
my reform



Don't like
this change



Figure 1 shows an example. This figure represents the explanations and comments of a Grade A Conservative, making only 17 changes in the original word list. The way in which the labels for the comments and explanations are spread out in a 2-dimensional space represents their relationships to one another. For example, "Obscuring a word's origins" is quite strongly associated with "better left unchanged." Precisely what this person means by "obscuring a word's origins" can be ascertained by looking back at the way in which this comment is applied to the reforms themselves. It is applied in particular when a derivational relationship between words in the list has been broken down, e.g. kwest = kweschun; fakt = factual.

High "rationality ratings" of the original word are associated with "having difficulty breaking old habits" in the reform. It can be concluded that this is a fairly enlightened Conservative who is aware of the major role played by habit in the acceptance of traditional spelling. Words such as "bom", "gon," and "goos" are highly rated on these scales.

"Making words more regular" is directly opposed to "better left unchanged", so it is clear that this person sees regularization as the principal aim of reform. On the other axis, "economy of symbols" is opposed to "improving the order of letters." The opposition applies particularly to the reforms "ogger" (ogre) and "aker" (acre) where in the first instance at least, the subject has to increase the number of letters in order to avoid a possible pronunciation. In both words the conventional full syllable representation of the final phoneme(s) as "er" may also be construed as a diseconomy. A parallel motive is found in the Nue Spelling's treatment of syllabic "l" in "peepl".

Figure 2 shows a Grade A Leveller. Interestingly "aesthetic offensiveness" is negatively correlated [2] with "not liking a change," so this person is not much concerned about offending peoples' aesthetic sensibilities.

Footnote [2] A correlation matrix forms part of the computer output of the principal components analysis.

There are two particularly interesting comments here. A clear distinction is made between reforms that are "easier to read" and those that are "easier to spell". At the same time, "easy to read" is strongly positively correlated with "spelling words as they sound", and "easy to spell" is strongly positively correlated with "economy of symbols". For example, "revishon" is rated as "easier to spell" and "economical", but not "easier to read" or "spelt as it sounds". The reform "appeer" is rated in the converse fashion - not "easier to spell" or "economical" but "spelt as it sounds" and "easier to read." The failure to acknowledge the distinction between readability and spellability has been one of the fundamental criticisms of the initial teaching alphabet (Haas, 1970).

Figure 3 shows a Grade A Linguist. This person is aware of childhood problems being caused by ambiguous spellings, referring mainly to words containing "x" and "q". He is obviously very concerned about the effects of his reforms on other people's pronunciation. This applies, for example, to "aw" (awe) and "cokett" (coquette). "Aw" may be particularly relevant here because this person is a Scottish dialect speaker and may be aware of the Anglo-Scottish discrepancy in the vowel phoneme system, centering on the /ɔ:/=/ɔ/ distinction (see Lass, 1974). "cokett" preserves some version of a Chomskyan stress placement rule (Chomsky and Halle, 1968), by preserving the final double consonants. However, it is more realistic to suggest that his concern is not with the native speaker's phonological awareness of this rule system, but rather with a general indeterminacy about the pronunciation of recent borrowings from foreign languages.

Original words with superflous letters are rated quite highly on "rationality". This applies mainly to words with final "-e", which, according to his original ratings, he is not too happy about dropping, especially where they have a syntactic function and/or may influence pronunciation. e.g. "hors" (horse).

In the centre of Figure 3 we have the typical "linguist's" statement "non-phonological considerations apply in the original" (not actually phrased this way in the interview - see notes to Appendix A). This statement applies particularly to the original spellings "burlesque," "coquette," "bomb," "horse."

Figure 4 shows a Grade A Phonetician. Here we see that words that have "sounds that are difficult to represent" cause problems; for example "worl" (wall). Here we have another example of the troublesome /ɔ:/ phoneme, though in this case the reformer is a Northern Englishman living in Scotland, not a Scot.

Encouragingly "spelling words as they sound" not only helps foreigners but also improves the appearance of the words. Examples here are "lej," "hav," pleez," "swet," (sweat).

On the other hand, this person is fully conscious that some people's pronunciation may differ from his own, e.g. "wick" and "paw" = "poor" = "pour", and this may be one good reason for rating the original spelling of these words quite highly.

Conclusions.


The analysis of these data is still continuing, but it is possible to draw some conclusions. Firstly, it is clear that the business of reforming English spelling is extremely complex. Many principles are involved and some of these are frequently in conflict. In practice, even relatively naive people will have a great deal to say about spelling reform. Even if some of their attitudes about spelling reform reflect their naivity rather than their sophistication, expert spelling reformers would be wise to take them into consideration. An analysis of the attitudes of a more representative sample of the general population should be undertaken.

Secondly, many of the problems highlighted by the analyses above are also found in the spelling reform proposals of the experts. A few examples will be given to illustrate this point.

For example, some expert reformers show a strong bias towards improving readability rather than spellability. Pitman's i.t.a. treatment of the schwa vowel in "legal, rebel, civil, symbol" is a good example. He prefers to retain the full vowels in these words, as in the derived forms "legality, rebellious, civility, and symbolic." This is unlikely to cause problems for the reader, at least as long as he/she knows the words; but the speller must make use of knowledge of the derivational relationships in order to spell the words correctly.

Among expert reformers, we also find "Linguists," preserving certain grammatical conventions in spelling (but which and why?), "Phoneticians," aiming for one symbol/ one sound consistency, and "Levellers," biassing their reforms towards the most frequent representations in traditional orthography (type frequency or token frequency?).

On a related issue, some reformers (e.g. Beech, 1980) recommend preserving "irrational" spellings in a few very frequent words in the interests of minimal disruption of traditional spelling. It might also be argued that reform will get off the ground more quickly if one homes in on precisely the same frequent irrational spellings from the start.

There are many problems in the details of reform. Some reformers are more worried than others about homophones. One reformer (Gassner, 1978-79) suggests doubling consonants in one of each pair of homophones in order to avoid homography. How would the speller fare with this convention?

Many reformers allow certain grammatical conventions to hold in spelling, but do not make it clear how far one should go. Wijk (1959), for example, permits us to preserve the unitary spellings of the third person singular present tense morpheme in "begs" (begs) but not in "is" (iz ).

Variations in pronunciation cause many difficulties. Pitman recommends that we should all distinguish between "law" and "lore" for the benefit of those speakers who pronounce them differently. But should we expect consistently "correct" spellings from those people who make no distinction in pronunciation? Many reformers state that Southern British speech should be the standard on which base spelling reform. Is there room for ethnocentric and parochial attitudes in a radical cause? Others recommend that in certain instances "expert" pronunciation should be relied upon. For example, "iodine" with an /i:/ vowel in the second syllable when pronounced by chemists, is apparently pronounced with an /aɪ/ vowel by many (Australian) laymen (Lindgren, 1969). To what extent should minority pronunciations be given precedence? This issue becomes more vexed when we consider spelling reform in an international context, where the majority of people use English as a second or third language.

There are many other basic issues on which the experts disagree. An urgent task for the Simplified Spelling Socety should be for members to examine their own attitudes to these basic issues. If possible, the Society should agree, preferably in consultation with the general public, on an order of priorities for the principles of spelling reform. Indeed it may be a salutary exercise for members to subject their own attitudes to the kind of analysis presented above.

Bibliography.

Albrow, K. H. (1972) The English Writing System - notes towards a description. Longmans: London.

Baker, R. G. (1980) "Spelling reform and the psychological reality of English spelling rules:" Spelling Progress Bulletin, Summer, 1980, pp. 17-20.

Beech. J. R. (1980) "Some proposed principles for simplifying English orthography. Spelling Progress Bulletin, Summer, 1980, pp. 7-13.

Chomsky, N. and Halle, M. (1968) The Sound pattern of English. Harper & Row: New York.

Gassner, W. (1978-79) "The essential requirements for reformed spelling." Spelling Action, 1978-79, pp. 8-10.

Haas, W. (1970) Phonographic translation. Manchester U. Press.

Hanna, P.R., Hanna, J.S., Hodges, R.E. and Rudorf, E.H. (1966) Phoneme-Grapheme correspondences as cues to spelling improvement. U.S. Gov't Printing Office, Washington, D.C.

Ives, K.H. (1979) "Acceptability of proposed spelling reforms." Spelling Progress Bulletin, Summer, 1979, p.14-5.

Lass, R. (1974) "Linguistic orthogenesis-Scots' vowel quantity and the English length conspiracy." In: J. M. Anderson & C. Jones, Historical Linguistics, Proceedings of 1st International Conference on Historical Linguistics; Edinburgh 1973, North Holland, Amsterdam, Vol. 2, p311-352.

Lindgren. (1969) Spelling Reform - a New Approach. Alpha Books, Sydney, Australia.

Pitman, Sir J. & St. John, J. (1969) Alphabets and Reading: the Initial Teaching Alphabet. Pitman: London.

Slater, P. (1976-7) The Measurement of Intrapersonal Space by Grid Technique. (2 Vols.) Wiley: London.

Venezky, R.L. (1970) The Structure of English Orthography. Mouton: The Hague.

Wijk, A. (1959) Regularized English. Almqvist & Wiksell: Stockholm.


Appendix A.

Lists of comments and explanations. concerning spelling reform behaviour

Comment.Number of people (out of 23) producing comment
1. The original word could be pronounced wrong.
2. I'm not sure how I pronounce the original word.
3. There are non-phonemic justifications for the original spelling.
4. This reform would be unnecessary if people spoke more clearly.
5. I had difficulty with this word as a child.
6. This spelling should definitely be changed.
7. It isn't clear how this word should be pronounced.
8. I am aiming for economy of symbols in my reform.
9. My reform indicates the word's emphasis better.
10. My reform makes the word easier to pronounce.
11. My reform makes the word easier to learn to read.
12. My reform makes the word easier to learn to spell.
13. My reform makes the word more difficult to identify.
14. My reform gets rid of unnecessary letters.
15. My reform gets rid of unpronounced letters.
16. My reform gets rid of ambiguous letters.
17. My reform gets rid of confusing letters.
18. My reform improves the order of letters.
19. My reform produces a more straightforward rule.
20. I have spelt this word the way it sounds.
21. I have spelt this word the way I pronounce it.
22. My reform is phonetically accurate.
23. My reform shows a more direct relationship between spelling and sound.
24. My reform is more straightforward than the original.
25. I am trying to avoid producing a homograph.
26. I am trying to maintain relationships between related words.
27. There are non-phonemic considerations in my reform.
28. My reform is better than the original.
29. I'm not sure if my reform is an improvement.
30. I'm not sure if my reform is the best possible.
31. I can't think how to improve my reform further.
32. This word would be better left unchanged.
33. I'm not sure why I made this change.
34. I don't like my reform.
35. My reform obscures the word's origins.
36. My reform looks odd.
37. My reform is cumbersome.
38. My reform is aesthetically offensive.
39. My reform could be pronounced wrong.
40. My reform has altered the pronunciation.
41. My reform is too complex for children to learn.
42. I am exchanging one irrationality for another.
43. I could have been more consistent in my reform.
44. Some British people may pronounce this word differently from the way I've spelt it.
45. It is difficult to represent some of the sounds in word without introducing new letters.
46. I am finding it hard to break old spelling habits.
47. My reform looks better than the original.
4
1
5
1
1
1
1
6
3
6
2
2
2
17
5
7
4
4
2
8
12
4
3
3
1
1
5
2
1
2
4
5
4
14
1
4
1
1
6
1
1
1
1
15
1
3
2

Notes:

 Some of the comments and explanations have been rephrased for the sake of brevity. For example, the explanation "There are non-phonemic justifications for the original spelling" (No. 3) was in fact originally given in forms such as "There are good reasons for spelling the original word this way although it doesn't represent the sound accurately."

In cases where two or more comments seem remarkably close to one another in content (e.g. Nos. 2 and 7), the justification for representing them separately is that both were produced and used separately and distinctly by a single individual.

Acknowledgements:

 This research was carried out while the author was employed on an SSRC supported research project at the University of Stirling. The principal components analysis was provided by the MRC service for analysing repertory grists for which Dr. Patrick Slater is responsible.

Back to the top.