[Spelling Progress Bulletin, Summer 1980, pp17-20]
[Robert Baker: see Journal, Bulletins.]

Spelling reform and the psychological reality of English spelling rules,

by Robert G. Baker.*

*Dep't of Electronics, Univ. of Southampton, England.
*Presented at the 2nd International Conf. of S.S.S. at Nene College, July 29, 1979.


The starting point of this research was not a zeal for spelling reform. It was rather an attempt to gather psychological evidence for some of the counter-reformist arguments put forward by transformational and structural linguists in recent years (notably Chomsky and Halle, 1968; Albrow 1972). Table I provides some examples of English spelling conventions. These examples illustrate some of the main points made in favour of the traditional spelling system. It is not only claimed to be more regular than is generally supposed (see especially the examples on the "phonotactic level" in Table 1, where spelling conventions are seen to be regularly determined by phonological context), but also to be capable of carrying information on a number of different linguistic levels, albeit at the expense of failing to provide a straightforward representation of English phonemes. Such an analysis suggests that English is not and is not intended to be a phonemic orthography, but rather a "mixed-level" orthography.

If we look at the spelling systems of other languages, we see that they each have over time and in interaction with the spoken language, come to represent selected structural aspects of language on various linguistic levels.

Table 1.

Higher order regularities of English spelling

Graphemicgive (English words do not end in v or z)
 level freeze 
Phonotacticfetch (/tʃ/ is represented by tch when
 level vs peachpreceded by a lax vowel,
  ch when preceded by a tense vowel)
 wan wash a is pronounced /ɔ/ when preceded
 vs wag by w, except when followed by a
 velar stop)
Morphemicwalked (past tense morpheme is regularly
 levelwarned represented by -ed in spite of
 waited phonemic alternations)
 cats dogs noun plural morpheme is represented
 horses by -s in spite of phonemic alterations)
Syntacticplease vs pleas (only plural nouns and 3rd person
 levelraise vs rays singular verbs regularly end in -s)
 goose vs zoos 
 at vs add ('Content' words always have more
 in vs inn than 2 letters)
Semanticseam vs seem (Homophones are regularly kept
 level  distinct in spelling)
 sign vs signal (Derivationally related words are
  regularly similar in spelling in spite
  of phonemic differentiation)

Thus Chinese generally represents lexical items by single ideographic symbols, although many of these symbols can also be broken down into subcomponents, some of which are phonological in nature (Gleitman and Rozin, 1977). Many familiar European orthographies such as Spanish operate almost entirely on the phonemic level. German appears to have a slightly mixed orthography in which morpho-phonemes are represented by single symbols, e.g.
Bund/bunt/ "Federation"
Bunde/bundə/ "Federations"
and initial capital letters are used as syntactic form class markers (all nouns begin with capital letters).

Japanese has three quite separate orthographic components used in parallel: an ideography similar to that of Chinese; and two syllabaries, one for native syllables, representing mainly grammatical morphemes, and one for foreign loan-words.

While English is similarly a mixed orthography, the different components are not as clearly delineated as they are in Japanese, but rather merge into one another. Furthermore, higher-order regularities such as those shown in Table I appear to be somewhat haphazardly distributed. Indeed, a full description of English orthography has not yet appeared. It is therefore not surprising that children have difficulty mastering the system, and educational failure is, of course the trump card held by those who advocate reform. It has been argued (Chomsky, 1970) that the reason for educational failure may be precisely the fact that English spelling is not generally taught as a mixed-level system, but rather as a faulty phonemic representation.

In the research to be reported, it was decided in the first instance to sidestep the educational issue and to focus on the extent to which English spelling makes sense to people who have already acquired spelling competence. Are the higher-order regularities apparent to literate adults who nevertheless have had no training in linguistics? Are they highly valued or merely doggedly tolerated?

Evidence has been presented at this conference (Smith, 1979) that people can and do use such types of linguistic information when placed in the experimental situation - or, it might be argued, when necessary. Therefore, an attempt was made to find a more direct way of tapping peoples' knowledge about how English spelling works (or fails to work). One obvious method would be to simply ask ordinary people to carry out their own spelling reforms of English words.

Spelling reform and the spelling reform task.

The history of spelling reform in English-speaking countries clearly indicates that the rationalization of our spelling system is no simple matter. There appear to be almost as many suggested reformed systems as there are pleas for reform. This difficulty may be partly inherent in the structure of the English language. As J. R. Firth pointed out 45 years ago: "the main argument against phonetic spelling ... (is that) ... it removes phonetic ambiguity and creates other functional ambiguities" (Firth, 1935)

It has been argued (Yule, 1978) that, in quantitative terms, the higher order regularities discussed above can hardly be called regularities, since in many cases there are more instances of rule-breaking than of rule-following. This is not at issue here. What is at issue is whether any of these non-phonemic orthographic patterns are synchronically well-motivated. If so, the onus is on the spelling reformer to justify the necessary loss of linguistic information entailed by a phonemic orthography.

Another difficulty is simply that of doing linguistics. Different spelling reformers will come up with different phonemic analyses. It has not even been agreed among linguists precisely how many phonemes are contained in the English repertoire. This is not surprising since the status of the concept, "phoneme," is by no means firmly established (Twaddell, 1958; Chomsky, 1964; Prieto, 1969).

Is it a physically identifiable unit in the acoustic signal, a psychological construct abstracted or idealized from the acoustic signal or the articulatory complex, or an illusion induced by over-familiarity with linear alphabetic writing systems? Even if a repertoire of English phonemes and a policy of reform by phonemicization could be agreed upon, the problems would not be solved. We must also agree on the "domain" of phonemicization. For example, we may wish a "word" to be defined in the spelling system (e.g. bound by spaces) and phonemicization to be restricted to "words" as if pronounced in isolation. On the other hand, those who are unwilling to make assumptions about syntactic/semantic units may wish the orthography to take account of phonological processes obtaining across word boundaries, e.g. the assimilation in:
In Southampton Im Portsmouth Ing Cambridge.
Examination of the proliferation of proposals for spelling reform reveals many different approaches to the resolution of such problems and to the treatment of higher-order regularities. Thus the initial teaching alphabet preserves, in the name of concessions to traditional orthography, for the sake of an easy transfer, a large number of lexical derivational relationships in spite of phonemic differentiation (e.g. kwest, kwestion). On the other hand, i.t.a. is determined to differentiate the phonemes θ and ð, although, if syntactic considerations are allowed, the rules for their distribution are so straightforward that they may be treated as allophones of a single phoneme (e.g. "th"). [1]

It is the author's view that of all the possible solutions for reformed English spelling, those will succeed which are most in tune with the man-in-the-street's notions of how spelling should work. It was therefore decided to ask people to act as amateur spelling reformers. In analysing the results, it would be assumed that those spelling conventions which the subjects changed would be less psychologically real than the reformed versions which replaced them, and that those conventions left unchanged had comparative psychological integrity.

Twenty-three university undergraduates were presented with a list of 111 words. The students were all literate, moderately good spellers, but with no formal phonetic or linguistic training. The words in the lists were not intended to be a representative sample of written English but were representative of the kinds of higher-order orthographic regularities exemplified in Table 1. Each student was asked in the first place to give every word a rating on a five-point scale, according to how "rationally" the words were thought to be spelt. The definition of the term "rationally" was left to the students. They were then asked to provide a "more rational" spelling for those words which were not considered to be completely rational. The full instructions and rating scale are shown below (Table 2).

Table 2

. Instructions for spelling reform task.

I would like you to try and imagine that you have been employed as an Arbitrator for a government-sponsored "Committee for the reform of English spelling", i.e. your job is to find the best way of spelling, English words.

Assumption: English spelling is, at least in part, an irrational and inadequate system for representing spoken English. You may not personally agree with this assumption. Do you?
Answer YES or NO on the dotted line).......
Please look at each of the words below in turn. First of all give the word a score from 1 to 5 according to how "rationally" you think it is spelt. If you think the traditional spelling is the best possible, give it a 5; if you think it leaves much to be desired, give it a 1. Scores of 2, 3, and 4 will be intermediate points on the scale. (Try and use as many points on the scale as possible.)

Then, if you have given the word a score of less than 5, try to suggest a possible "more rational" alternative spelling. In some cases you may not be able to think of one (then leave a blank).

Try to work consistently through the list and try not to miss any words out. You may at any time refer back to words you have already dealt with, but, if you make any alterations in your "reformed" spellings, please make it quite clear what you have done by crossing out the altered form with one line. Examples: night= 3, nite. tough= 2, tuf tuff


The number of words rated less than perfect differed greatly from student to student (from 17 out of 111 to 100 out of 111) although all the students considered that English spelling was at least in part an inadequate and irrational system for representing spoken English. In order to gain some idea of how the rating scale was being used, the average "rationality rating" for each word was computed and then correlated with an objective measure of the word's spelling regularity. This measure was derived from a frequency count of sound to spelling correspondences (Hanna et al, 1965). Thus if the sound /s/ is represented by "ss" in 442 out of 6326 occasions sampled, a score of 442/6336 for /s/="ss" was given. Scores for all phoneme-grapheme correspondences in each word were summed and divided by the number of phoneme-grapheme correspondences in the word in order to provide an average regularity score. The correlation between regularity scores and average rationality ratings was positive and highly statistically significant (Spearman's rho=0.44, n=111, p< 0.001). Thus words which are rated as highly rational were also highly regular and we may infer that the students were making meaningful judgements about the words' spellings.

The reformed spellings were analysed primarily in terms of the extent to which they maintained or destroyed the higher-order regularities. Table 3 shows the extent of rule preservation for the different types of regularities in Table 1. The average rationality ratings in, Table 3 cannot be considered to relate directly to the spelling conventions of interest since the ratings are applied to whole words. In general, however, those word types which are highly rated also show preservation of higher-order regularities.

Two results stand out particularly from the percentages of rule-preservation. The use of syntactic final "-e" after "s" to indicate single nouns is not highly valued (29.3% preservation). On the other hand, use of "s" for phonemic /z/ in plural nouns, i.e. preservation of morphemic "s" plural in spite of phonemic variation is highly valued (87.9% preservation). In the majority of other cases, rule-preservation is close to 50% and it is difficult to draw practical conclusions for the benefit of spelling reformers.

However, the overall data may be examined in a different way. We may ask to what extent there is agreement between students on particular reformed spellings. In fact, for 42 out of the 111 words (36%) there was exact agreement on the reformed version between at least one third (> 8 out of 23) of students. While this result is not world-shattering, it is likely that such "popular" spellings represent forms of high psychological plausibility and spelling reformers would do well to take cognizance of them. Examples of such reforms are "apeer" (for "appear", 8/23), "oger" (for "ogre", 13/23), "peech" (for "peach", 19/23).

Conclusions and recommendations from the spelling reform task.

'The above selection of results from a single study is not offered as a definitive guide for spelling reformers. The subjects and words which were sampled were far from representative. It is however suggested that larger scale empirical studies of popular attitudes towards spelling adequacy would provide valuable insights into the pragmatics of spelling reform. It could be that every elegant creation by linguistically or educationally sophisticated spelling reformers is bound to fail when transferred from the study or the committee room to the market place or the classroom. Or more important, will any of these creations ever get as far as the market place or classroom? On the other hand, if some degree of popular consensus on the substance of rational spelling could be achieved, then the popular view of spelling reformers as ineffectual cranks would accordingly be diminished. It may be argued that a "democratically" achieved spelling reform would not necessarily be the best to work in practice, but would merely reflect a collection of mass prejudices. On the other hand, it can be counter-argued that a reform, once adopted, will be modified and optimized by usage; but it must first be adopted! People may accept, at least for a while, what they think they want.

Table 3.

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

Rule level SpellingAverage rationality
% of occasions
rule is preserved
Graphemicgive not giv 3.7142.7
 freeze not freez 4.2967.8
Phonotacticfetch not fech 4.1866.7
 wash not vosh 4.0152.2
Morphemicwalked not walkt 3.2260.9
 dogs not dogz 4.7187.9
Syntacticgoose not goos 3.7829.3
 add not ad 3.5352.2
Semantic'g' retained in 3.5346.1
 sign and signal
 seem and seam 3.8057.2

A further problem concerning data from the spelling reform task is that the man-in-the-street may have traditional spelling so deeply ingrained in his mind that he cannot took beyond it. How can he be objective? However, the question remains as to how deeply engrained different aspects of the spelling system are. Surely, those aspects which are most deeply ingrained will be most resistant to change. It is as well to know what these obstacles to reform will be. Furthermore, those most resistant spelling conventions are no doubt those which made most sense to the child as he/she was learning to use the system, and they are most likely to make most sense to him as an adult.

Children's use and understanding of spelling rules.

It would be valuable to be able to trace the development of spelling rule knowledge in children. Accordingly a number of simplified pilot versions of the spelling reform task were tried out with young children (6-8 years), but without much success. In general, young children do not seem prepared to discuss the pros and cons of less or more rational spelling conventions. This may reflect cognitive immaturity or an unwillingness to question the authority of English spelling. For these children, a word is either spelt right or wrong and there is no ground for debate.

Since these young children were unable or unwilling to manipulate correct spellings, it was decided to examine their attitudes towards incorrect spellings. A diagnostic spelling test (Peters, 1970) was administered to two classes of primary school children, a Primary 3 class (n = 16) and a Primary 5 class (n = 29). The children's errors were analysed and related to test data on their reading ages made available by their teachers. After completing the spelling test, each child was given a structured interview centered on the errors he/she made in the test. The children were asked to try and explain their particular difficulties with the words and to talk, about why they thought they had made these particular errors. This approach provided considerable insights into the children's approaches to spelling and to the testing situation. The results of this study are reported in greater detail elsewhere (Smith et al, 1979), but some general points will be made here.

Table 4

Reasons given for spelling errors made by primary school children

Class of ExampleFrequency of occurrence:
explanation Primary 3Primary 5Total
(Un)familiarity"I don't know that word" 374481
Perception"It looks OK my way" 294675
Test situation"I'd have got it right if I'd had more time"253661
Difficulty"It's a hard word" 183553
Performance"My pen slipped" 141630
Rule"I got mixed up about the rule" 02222
Phonic strategy"I tried sounding it out" 21012
Bad speller self image"I'm just careless" 044

In Table 4, the classes of explanation are given in the overall order of frequency in which they occurred. Two of these classes are particularly relevant to this discussion of English orthography; the use of a phonic strategy and the use of rules. It is noteworthy that the phonic strategy explanation is not particularly frequent. This may be either because it is such a useful strategy that it does not generally lead to errors (see McBride, 1977) or because the children do not often consciously use it as a strategy. Also noteworthy is the status of "rules" in this situation. Only the older class of children referred to rules at all. This may merely reflect the stage they had reached in their reading and spelling schemes. In the Primary 5 class, the rule explanation was used in three different ways. Sometimes it was used quite appropriately, for example, "I forgot the doubling rule" (for "spining" = "spinning"), sometimes apparently quite inappropriately, e.g. "I don't know the rule" (for "svicetoin" ="satisfaction"), and some times appropriately but erroneously, e.g. "I spelt it like that because it's got 'high' in it" (for "hight"). This last example is evidence of the false overextension of a higher-order lexical derivational relationship rule. The children who offered rule explanations inappropriately and/or incorrectly were predominantly in the lower third of the reading/spelling ability range in their class. Furthermore, for these children and for these children only, the rule explanation was always associated with the difficulty explanation. It appears that children with lesser reading/spelling ability will resort to rules but are likely to be led astray by them.

Overall, this study did not provide much evidence for or against the psychological reality of particular spelling rules in young children. There were occasions on which statements such as "e usually has an 'a' next to it (for "neaver" = "never") were made, but these were too sporadic to be given much weight in the main analysis. This study has, however, given a clearer picture of the child's general approach to spelling and a framework for further research. ne kinds of explanations given by children to account for their own failures in spelling tests could be of use to teachers in assessing their teaching procedures and in evaluating the testing situation. Tests always take place in a context. In particular, different children will not be equally familiar with the words they are asked to spell. They may have experience in spelling a word, or they may have only encountered it in reading, or they may not know the word at all. One Primary 3 child made a further distinction between having spelt a word on his own initiative and having copied its spelling as part of an exercise. These different degrees of experience are likely to be associated with different types of spelling error. Furthermore, the effects of imposing time constraints on a rest will differ from child to child. The explanations classed as "Perception" will relate to those classifications of spelling errors made in terms of information encoding (e.g. Avakian-Whitaker and Whitaker, 1973) and may help to validate such classifications.

Self-characterizations as a "careless speller" are fortunately rare in this group of children. The dangers of allowing a child to continue with the notion that he/she is a bad or careless speller have been pointed out (Peters, 1967). Members of this society will of course be amongst the first to stress the role that English orthography plays in generating spelling difficulties and to press the more urgently for reform. In the meantime, however, we would do well to explore ways of putting across the complexities of our spelling system, taking into consideration the child's own expectations and intuitions about the task. Such explorations will in turn guide spelling reformers towards the most highly motivated and usable alternative system.


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

Avakian-Whitaker, H. and Whitaker, H. A. (1973) "The spelling errors of children with communication disorders: a preliminary classification." Linguistics, 115, p.103-118.

Chomsky, C. (1970) "Reading, writing and phonology." Harvard Educational Review, 40, 2 p.287-309.

Chomsky, N. (1964) Current Issues in Linguistic Theory. The Hague: Mouton.

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

Firth, J. R. (1935) "The technique of semantics." Transactions of the Philological Society, p.36-72.

Gleitman L. R. and Rozin, P. (1977) "The structure and acquisition of reading." in A.S. Reber & D.L. Scarborough: Towards a Psychology of Reading. Hillsdale, N. J. Lawrence Erlbaum.

Hanna, P. R., Hanna, J. S., Hodges, R. E., & Rudorf, E. H., Jr. (1966) Phoneme-grapheme correspondences as cues to spelling improvement. Washington, D.C.: U.S. Government Printing Office.

McBride, J. F. (1977) Phonics in the teaching of reading. Paper presented to the Reading Assoc. of Ireland.

Peters, M. L. (1967) Spelling: caught or taught. London: Routledge & Kegan Paul.

Peters, M. L. (1970) Success in Spelling. Cambridge: Institute of Education.

Prieto, L. J. (1969) "La découverte du phonème: Interprétation epistemologique." La Pensée, 148, p.35-53.

Smith, P. T. (1979) In defence of conservatism in English spelling. Paper presented at SSS Conference, Northampton.

Smith, P. T., Baker, R. G. and Groat, A. (1979) Spelling as a source of information about children's linguistic knowledge. Paper presented at "Beyond Description in Child Language Research," Nijmegen, Holland.

Twaddell, W.F. (1958) "On defining the phoneme," in M. Joos. Readings in Linguistic 1. New York: American Council of Learned Societies.

Yule, V. (1978) "Is there evidence for Chomsky's interpretation of English spelling?" Spelling Progress Bulletin 18, 4, p.10-12. [Yule: see Bulletins.]

Acknowledgements: This research was carried out while the author was working on an SSRC supported project in the Dept. of Psychology, Univ. of Stirling. Thanks are due to Dunblane Primary School, Perthshire, for their help and cooperation in the project.

[1] Note:

1. th= ð in word-medial position and in initial position in function words, (e.g. this, there, etc.) but not in content words (e.g. thin, theme, etc.)

2. th= θ elsewhere.

The exceptions are the set of minimal pairs, "wreath, wreathe, sooth, soothe, etc" Here it is noteworthy that the orthographic representations place the voiced alternant in non-final position in the word, thus conforming to rule 1.)

Back to the top.