[Spelling Progress Bulletin, Winter 1982 pp2,3]

SSS Conference 3: Development of Improvement in English Orthografy.

"Semantic aspects of spelling reform,"

by Neville Brown, Ph.D.

Foundation for the Education of the Under-achieving and Dyslexic, Maple Hayes School and Research Centre, Lichfield, Staffordshire, Eng.


Traditionally, attempts to simplify English spelling have been predominantly motivated by the difficulties that a substantial minority of children experience in early written language acquisition, particularly in respect of phoneme-grapheme correspondence and the reproduction of polysyllabic words. Evidence from recent research suggests that too much emphasis on the phonological or phonetic aspects of written language in teaching may deter many children from using other available strategies, such as direct linguistic encoding and decoding in reading and writing, to the extent of impairing their learning and performance efficiency. From this, it is argued that semantic considerations should take precedence over phonological considerations in any attempt to reform the English spelling system.


The motivation for spelling reform stems predominantly from the difficulty experienced by a considerable proportion of school children in acquiring the written English language and also the apparent disparity between the acquisition of oral and reading vocabulary and between reading and spelling vocabulary in terms of word length. This paper will attempt to show that, whilst attention to such a problem is not of itself misplaced, some of the assumptions underlying traditional approaches to spelling simplification are not only untenable but perhaps counterproductive.

Whilst it will be readily recognized that Chinese ideographs, for example, constitute a written language that has little or no phonological basis but is nevertheless a language, there has been the general expectation in most Indo-European languages that a fine relationship between phoneme and grapheme is desirable and can be attained, so that much attention, for example, that of Sir James Pitman of i.t.a. (initial teaching alphabet) fame, has been directed to alter the orthography accordingly. It is, however, recognized that within our own language the notion of an absolute value for any phoneme is undermined by the infinitely wide variation in pronunciation between regions, villages and even individuals. One aspect of the usefulness of written language is that it transcends such differences to a large degree. At this time, it might be argued that any attempt at reforming English spelling along phonetic lines would involve such a multiplicity of necessarily acceptable spellings using an orthography of such extended range as to render a child incapable of using writing as a communicative tool. This, however, is not quite my theme in this paper.

In recent years, research findings in the field of information processing indicate that there are mechanisms in an analysis of which clearly supports the view that there are separate phonological and visual pathways in reading and also that there are two fundamentally different operations applicable to "information processing in general and to reading/spelling in particular which may be termed analytic-sequential and holistic-simultaneous (Das, 1973, Bever, 1975). Aaron (1977) suggests that reading involves the analytic-sequential processing of selected letters and the holistic-simultaneous perception of the salient features of the entire word and concluding that dyslexia is due to an imbalance between these processes.

The prevailing assumption of reading pedagogy and its research is that processing along the phonological and analytic-sequential pathway is not only essential but is prior to any other process. In other words, reading and spelling cannot proceed without phonological mediation. At first sight, commonplace findings would appear to support this view: Children who experience difficulty in reading do tend to sound out the words that are not in their sight memory letter by letter and build the sequence of sounds into the word and, further, their difficulty appears to increase as does the word length. This relationship is, however, not a simple one and correlation coefficients for difficulty x letter count rarely exceed .4, hence the need for elaborate calculations involving other variables in indices of reading difficulty of prose passages used for children. It is very easy to fall into the trap of thinking that because 'normal' readers do use phonological mediation in reading and spelling, then phonological mediation is necessary and prior, and that consequently a high correlation between written language disability and deficiency in the ability to analyse words into phonemes indicates the need to 'overreach' phoneme-grapheme correspondence or to render the orthography more amenable to such a teaching approach. . . or both!

In a study of preference for encoding modality in reading in 149 'apprentice' readers, it was found (Brown, 1978) that subjects who preferred the visual modality exhibited greater reading difficulty than those who preferred the auditory, but subjects who showed no preference tended to be better readers. Amongst the poor readers, the ratio of visuals to audials was approximately 2:1. The predominately auditory-preferent readers clearly tended to be underachievers in the sense that their mechanical reading level was below what one would expect from a study of their oral language ability. Aaron (1978) in a study of processing strategies in dyslexics, found that whilst normal readers had both adequate eidetic and phonetic memory, dyslexics tended to group into what her terms 'dyseidetics' and 'dysphonetics'. If one equates Brown's 'audials' with Aaron's 'dyseidetics' and Brown's 'visuals' with Aaron's 'dysphonetics', one comes to the. conclusion that the two processing strategies are not only potentially present in reading but also necessary, and that a 'mixed' approach to remediation of reading and spelling difficulties rather than a solely 'phonics' approach would be beneficial.

The 'mixed' view appears at first sight only to be supported by a study by Brown (1976) of linguistic complexity in prose passages of known difficulty. Brown took the responses to 460 applications of the Neale Analysis of Reading Ability which has 6 graded prose passages and found that the error counts for individual words did not bear a simple relationship to word-length or to familiarity. The word 'confident' accounted for 2.13% of the errors on passage no. 5, 'captives' had a 3.23% error, 1.39% for 'proceeded', 6.01% for 'persistent', whilst 'withstand' had no error count at all. When relationships between error count and a number of word-variables were explored, at the level of the Neale passage 1, which consists with one exception of short monosyllabic words, the only significant factor in passage 1 was the length of the words in letters. As the passages increased in difficulty, other variables such as syllable-count and morphographeme-count assumed significance. The variable which most affected error at the higher levels of the Neale was, however, not the length of word in terms of letters, syllables or morphographemes but the incongruence of syllable/ morphographeme boundaries.

A working definition of a morphographeme is a meaningful letter string that can, irrespective of any shift in sound or pronunciation be generalised from one word to another. Thus 'ed' is a morphographeme in the words 'wanted', 'killed' and 'skipped' though its pronunciation differs markedly over the three examples. As it could be argued that 'wanted' has two syllables and 'killed' and 'skipped' have but one, the difference is perhaps more quantitive. The word 'corporation', however, can be divided into four syllables and also into four morphographemes, but the boundaries are quite different:
In syllables: cor - por - a - tion
In morphographemes: corp - or - at - ion
It is in words where the syllable and morphographeme boundaries do not coincide - termed 'incongruent' here - that tend to give much greater difficulty to children in reading and spelling than words of comparable length without 'incongruence.'

Table 1:

Correlation coefficients of relationships. between Error Count and Word-level variables in the Neale Analysis of Reading Ability, Form A, p<001 throughout. (Brown 1978)

'Age' Letters
in word
Syllables Morpho-
of Syll/MG.




Aaron (1977) suggests that reading (and presumably spelling) involves the analytic sequential processing of selected letters and the holistic-simultaneous perception of the salient features of the entire word, coming to the conclusion that Dyslexia, specific written language difficulty, or whatever one wishes to term it, results from a deficit in one of these processes. Saffron and Marin can, however, accede from their clinical observations of aphasic dyslexics that reading can not only proceed independently along the two pathways but can even proceed exclusively along either. This suggests, further, that for some children in difficulty, the direction of attention to the morphographemes in longer words, irrespective of their fine pronunciation, might facilitate the processing of a word for meaning as morphographemes are by definition semantic units. In the course of research, it was found in the application of the Neale Test for Mechanical Accuracy and (subsequent) comprehension without feedback of mechanical errors that sometimes a key word for a comprehension answer would be unrecognizably misprocessed in mechanical reading and yet reproduced perfectly in the comprehension test.

In another experiment, unfamiliar words were taught under controlled conditions to children with reading and spelling difficulties but who were competent in the oral language. Whilst this work is reported elsewhere (Brown, 1978 and 1979), it was found that such children could learn to spell and read - both mechanically and for comprehension - the corpus of 'impossible' words significantly better by a morphographemic approach than by what may be termed a 'phonic' approach. For this experiment, the teaching of vocabulary was conducted in silence and there fill was evidence that attempted subvocalisation inhibited processing of the morphographemes.

It appears that a polysyllabic and polymorphemic written word can be encoded and decoded not only at different levels but by different pathways. The analytical sequential approach to our word 'cor-por-a-tion' may be at the level of grouping individual letter sounds into syllables and thence to the word. The implication of this approach is that semantic encoding is not possible until the whole word has been processed and referred to acoustic memory. It is frequently found in reports from certain Dyslexia centres that extra practice in 'chunking' sounds into syllables is required. The word 'chunking' (after G. A. Miller) appears to be entirely inappropriate to the analytic-sequential processing path as the resultant encoding or decoding need not involve meaning, hence the phenomena of 'barking at print' and the ability to spell better than to read in some children, recorded by Carbonell de Grampone. 'Chunking' does seem to be appropriate for the simultaneous-holistic processing of letter strings or morphographemes for meaning-cum-recognition.

The next step in the argument is that meaningful encoding and decoding is preferable to meaningless, and that anything which prevents the processing of morphographemes in a word is more serious than that which inhibits the mere pronunciation of the word by the analytic-sequential pathway.

In lieu of a conclusion, it may be useful to give examples where angels should fear to tread:

The integrity of the morphographeme 'vis' should be maintained across the words 'divisive' and 'division' so that a spelling change which differentiates 'divisive' and 'division' should be avoided. The spelling pattern 'rupt' is better maintained over 'disruptive' and 'disruption' and 'rig' is better maintained over 'rigor', 'rigid', and 'incorrigible.'

On the other hand, 'x' may well be regarded as not only a redundant letter but one that interferes with the generalisation of letter strings. Without 'z' it may be easier to relate 'ecsample' to 'sample' 'ecsamine' to 'same' (and all four together), and even perhaps the 'ecs' morpheme as in 'ecstract' to 'ec' in 'economy' and 'eco-system.' There is perhaps a case for changing 'build' to 'bild' and so bring us into line with at least one other EEC language, though this is better seen as a bonus when it occurs rather than a prime aim.

Our energy at Lichfield is currently directed towards enabling the dyslexic child to come to terms with the existing orthography and our teaching will be considerably helped by the forthcoming publication of a dictionary of morphographemes which we shall call a 'Wordbitbook'. From this paper, it should be apparent that we also have clear views on the direction that spelling reform should take if it is to assist children with unexpected difficulty or failure.


Aaron, P.G. (1978) Dyslexia, an imbalance in cerebral information processing strategies. Perceptual and Motor Skills, 1978, v. 47, 699-706.

Bever, T.G. (1975) Cerebral assymmetries in humans are due to differentiation of two incompatible processes: holistic and analytic. Annals of the New York Academy of Sciences, 1975, v. 263, 251-262.

Brown, E.N. (1978) Attentional style, linguistic complexity and the treatment of reading difficulty. In Eds., R.M. Knights and D.J. Bakker, Treatment of Hyperactive and Learning Disordered Children: Current Research. Baltimore, University Park Press.

Brown, E.N. (1979) Coding strategies and reading comprehension. In Eds., M.P. Friedman, J.P. Das & O'Conner, Intelligence and Learning. New York: Plenum.

Das, J.P. (1973) Structure of cognitive abilities: evidence for simultaneous and successive processing. Journal of Educational Psychology, 1973, 103-108.

Saffran, E.M. & Marin, S.M. (1977) Reading without phonology: evidence from asphasis. Quarterly Journal of Experimental Psychology, v. 29, 1977, 515-525.

Back to the top.