[Spelling Progress Bulletin, Fall 1982 pp2-4]

SSS Conference 3: Teaching and Learning Spelling continued.

"Spelling difficulty in school leavers and adults,"

by Dr. Dolores Perin*

*City Univ. Grad. Center, Devel. Psychology Program, New York, N.Y.
*A paper presented at the third International Conference "Spelling: Research and Reform", sponsored by the Simplified Spelling Society, Edinburgh, Scotland, 31st July-3rd August, 1981.
*A fuller discussion of the research presented here is available in Perin, Dolores (1982). "Spelling Strategies in Good and Poor Readers," v. 2: 1-14, Applied Psycholinguistics.


This study was concerned with the use of phoneme-grapheme correspondence rules in the spelling of sixteen-year-olds and adult literacy students of varying reading skill. A task was designed which was intended to be conducive to a phonological spelling strategy: this was a word completion task where gaps in words, representing a consonant and an adjacent vowel, had to be filled in to dictation. Two experiments were carried out using this paradigm. The first involved real words and the effect of "phonetic ambiguity" on spelling strategy was studied. The second task used nonsense words. The errors made in the real word task showed that the tendency to make phonetic spelling errors, i.e., to choose plausible graphemic options, increases with reading skill. The effect of phonetic ambiguity was not significant as a function of reading skill: subjects at all reading levels tended to make more errors on phonemes with relatively ambiguous relationships to graphemes. Nonsense word results supported the hypothesis that better readers are more able than poor readers in using phoneme-grapheme correspondences in spelling.


Educationalists and cognitive psychologists have found it useful to examine spelling errors for the clues they give about processes underlying the production of written language. Early large-scale surveys (Masters, 1927; Mendenhall, 1930) indicated that spelling errors tended to be "phonetic," that is, they retained the sound structure of the attempted word. Information processing models of spelling indicate how a phonetic error can occur: Simon's (1976) model has a level where phoneme-grapheme correspondences are employed in cases of doubt as to a word's correct spelling and Morton's (1980) logogen system allows for the use of phoneme-grapheme rules as an alternative to the retrieval of the orthographic structure of a word solely in terms of its component letters. Frith (1980) has suggested that the development of spelling ability requires a stage where phonemes are translated into graphemes.

That most errors have been reported to be phonetically accurate appears to serve the movement for spelling reform. It could be argued that phonetic errors imply the use of sound, and spelling difficulty could be seen in terms of the complex relationship between sound and symbol in English. If our writing system were completely "regular," if each phoneme were mapped by one and only one grapheme, spelling problems would not normally occur at all. Spelling reform might possibly benefit the speller or some spellers, but the cost of such reform would probably be enormous for the reader who prefers the direct (lexical) route (Frith, 1979) and would thus benefit from orthographic relationships among words which are related in meaning but not necessarily in sound (Chomsky & Halle, 1968). The possible effects of spelling reform are not further considered here. The present concern is with the relationship of phonetic accuracy in spelling, and reading.

Reading and spelling are usually strongly associated (Horn, 1969): a good reader normally spells well and the poor reader is usually a poor speller. However, reading skill does not by any means guarantee perfect spelling Frith, 1978) and extremely good readers will occasionally be troubled by poor spelling skill. Studies of spelling error patterns comparing good and poor readers /spellers have not been entirely consistent. For example, Nelson and Warrington (1974), Sweeney & Rourke (1978), and Frith (1980) have reported that spelling error patterns differ with reading skill while Holmes and Peper (1977) found that good and poor readers differ only in quantity, not quality, of spelling errors. Studies of spelling have in general been concerned with children. In the present investigation, school leavers and adult literacy students were studied in order to discover whether reading ability was related to the use of phoneme-grapheme rules in spelling in these older age groups.

In two experiments, real and nonsense words were spelled. In the first experiment, besides investigating the relationship between phonetic errors and reading ability, the effect of phoneme-grapheme ambiguity was studied.

The graphemc representation of four specific phonemes was examined. Two of these were relatively ambiguous as they could each be mapped by four separate graphemes. The other two phonemes were relatively unambiguous with only two plausible mappings each. Table 1 shows the correspondence rules for the four critical phonemes.

Table 1.

Phonemes used in word completion task.
/f/ → < f >, < ff >, < ph >, < gh >
/ʤ/ → < g> ,+ e, <g> + i/y, < dg >, < j >

/b/ → < b >, < bb >
/t/ → < t >, < tt >
It was predicted that higher reading ability would be related to the tendency to make phonetic spelling errors. Further, since better readers were expected to use phoneme-grapheme rules more than poorer readers, the former were expected to be affected by phoneme-grapheme ambiguity, a problem arising through dependence on rules.

Two samples were studied: young adults who were to leave school shortly, and adult literacy students.

The school leavers were 36 girls and 24 boys aged 15 and 16 who were in three middle-band classes in the 5th year of secondary school. They were of working class background and were native English speakers.

The adult literacy students were 10 women and 24 men whose ages ranged from the early twenties to the mid-fifties. Almost all were working class and all had English as the mother tongue.

A preliminary test of oral reading was carried out for each subject individually as follows: School leavers read a 350-word passage from a previous year's reading exam which no subject had seen previously. A percentage reading score was derived based on 116 words (functors and repetitions were disregarded).

Adults read a 265-word passage from a brochure advertising English holidays. Since the poorer readers had considerable difficulty with the text, reading score was a dichotomous measure in this sample. Good readers read quickly and accurately, while poor readers stumbled through the first two paragraphs, at which point they were told to stop.

The experimental material was in two lists each of 44 low frequency words which appeared less than five times per million in the Kucera and Francis (1967) corpus. Words were of one or two syllables and each contained one of the critical phonemes in a word-medial or final position. This was a word completion task. The printed form of each stimulus word was presented with a gap, always the same size irrespective of the number of missing letters, representing the critical phoneme plus an adjacent vowel and occasionally an adjacent consonant. Table 2 shows examples of stimuli.

Table 2.

Examples of Word Completion Stimuli.
coughs, serf, sulphur, tariff
pledged, serge, abject, baggage
tubs, lobe, hobble, proverb
stale, skit, potter, deceit

Subjects were tested in groups on the spelling task (each group spelling one list or the other) and were told to fill in the missing letters for the words, which were dictated, placed in sentences and then repeated. Guessing was encouraged in cases where subjects were in doubt as to spelling.

Spelling errors were analysed with respect to whether they were phonetically accurate and with respect to whether the ambiguous phonemes led to more errors than the unambiguous phonemes. For these error analyses, subjects who spelled four or fewer words incorrectly were excluded. This left 49 school leavers and 29 adults.

School leavers spelling the two lists did not differ in reading skill but there were different frequencies of spelling error for each list. Therefore, error analyses for this sample were carried out separately for each list. Adults' error frequencies did not differ by list and so the two lists were pooled for this sample.

Phonetic errors were productions where an incorrect choice of grapheme was made from the alternatives shown in Table 1. For each subject, phonetic errors were expressed as a percentage of total errors. Table 3 shows examples of phonetic and non-phonetic errors.

Table 3.

Examples of Errors in Word Completion Task.
Stimulus Response
 Phonetic errors Non-phonetic errors
l arf s
tar af
suffr idge
ga jiet
l arg s
tar ith
suffr agge
ga ddi t
dr abb
ra bber d
bu tta ne
pa stt el
dr ag
ra pid
bu bia ne
pa sc el

To study the effect of phoneme-grapheme ambiguity, errors for each critical phoneme were expressed as a percentage of total errors for each subject.

In the school leavers, the percentage of phonetic errors were regressed on reading In both lists, the relationship was significant, with better readers tending to make a larger percentage of phonetically reasonable spelling errors. There were 18 good readers and 16 poor readers in the adult sample. The good readers made a mean of 71.8% phonetic errors, compared with the poor readers' 48.28%. An analysis of variance showed that this was a significant difference, and that the direction of difference was the same as for the school sample. Table 4 shows the results of the error analysis.

In both samples, many more errors were made on the ambiguous than on the unambiguous phonemes. The school leavers had 70% of errors on ambiguous phonemes, while adults had 75%. In both samples, there was a significant difference between error rates for ambiguous and unambiguous phonemes. A score was constructed for each subject representing the difference between percentages of error on ambiguous and unambiguous targets. These difference scores were regressed on reading scores in the 3 school sample and subjected to an analysis of variance in the adult sample. In neither sample was the extent of difference on the two types of phoneme significantly related to reading ability. The results are shown in Table 4.

Table 4.

Error Patterns in Word Completion.

School leavers Phonetic errorsNon-phonetic errors

List 1
List 2
Adult Literacy Students
Mean %

Difference between errors on ambiguous and unambiguous phonemes.

School leavers
List 1
List 2
Adult Literacy Students
<.06 n.s.

The results of these analyses indicate that although poor readers are not as competent as good readers at using phoneme-grapheme correspondences, they are prone to error in a similar way to good readers, when there is a choice of which grapheme to choose out of a number of alternatives. This sensitivity to an effect of orthographic regularity suggests that poor readers are aware of phoneme-grapheme rules to some extent. They might prefer not to use a phonological strategy in spelling, or, on the other hand, they might be less able to use this strategy than better readers. To find out more about differences between good and poor readers in using rules in spelling, nonsense words were employed. This provides a stricter test of the use of rules in spelling since, by definition, there is no established orthography for such words.

In a second experiment, 57 of the school leavers (35 girls, 22 boys) and 32 adult literacy students (10 women, 22 men) were asked to complete nonsense words to dictation.

The stimuli were based on the real words used in the first experiment. Table 5 shows examples of the non-words used.

Table 5.

Examples of nonsense words.
grift, toaf, suffel, tebuff
ludged, gerge, tigeon, nuffrage
dobs, frab, lebuke, tadverb
stob, grat, kotive, brottis
Subjects who had spelled the first list in Experiment 1 now spelled the second, and vice versa. The paradigm was the same, with the subjects filling in gaps left in each word.

The criteria for correct responses were those indicated in Table 1. Any grapheme listed for a particular phoneme was accepted as correct. Therefore, all errors were non-phonetic errors in this task.

In both samples, the better readers were significantly more accurate than poor readers in spelling the target phonemes. This was shown by the regression of nonsense word scores on reading scores in the school sample (who spelled a mean of 91.1% correctly) and by an analysis of variance in the adult sample (good readers spelling 94.11% correctly, poor readers 76.36% correctly). Table 6 indicates the main results. There were no list effects in either group.

The results of the nonsense word spelling task provides interesting information about poor readers. Although on the first task they had made significantly fewer phonetic errors than good readers, it cannot be concluded that they cannot use phoneme-grapheme rules. Although they did not perform as well as better readers, poor readers in fact mapped a large number of the target phonemes correctly in their spelling of nonsense words. Therefore it is possible that poor readers avoid the phonological route in spelling real words although such a route is useful for generating plausible alternatives which could then be matched against word recognition memory (Simon & Simon, 1973; Tenney, 1980). The phonetic errors of better readers suggest their greater use of the phonological route. The nonsense word results show that although poor readers are not as able as good readers to translate phoneme to grapheme, they do have a sufficient rule-knowledge to employ a phonological route in spelling if encouraged to do so.

Table 6.

Nonsense Word Completion.

School leavers
Adult literacy students
Mean %


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

Frith, U. (1978). "From print to meaning and from print to sound, or how to read without knowing how to spell." Visible Language, 12, 43-54.

Frith, U. (1979). "Reading by eye and writing by ear." In Kolers, P.A., Wrolstad, M.E., & Bouma, H. (Eds.). Processing of Visible Language, vol. 1. New York: Plenum Press.

Frith, U. (1980). "Unexpected spelling problems." In Frith, U. (Ed.) Cognitive Processes in Spelling. London: Academic Press.

Holmes, D.L. & Peper, R.J. (1977). "An evaluation of the use of spelling error analysis in the diagnosis of reading disabilities." Child Development, 48, 1708-1711.

Horn. T.D. (1969). "Spelling". In Ebels, R.L. (Ed.) Encyclopedia of Educational Research, 4th Edition. New York: Macmillan.

Kucera, H. & Francis, W.N. (1967). Computational Analysis of Present-day American English. Providence, R.I.: Brown Univ. Press.

Masters, H.V. (1927). "A study of spelling errors." Iowa City: Univ. of Iowa Studies in Education, 4 (1st Series, 1938)

Mendenhall, J.E. (1930). An Analysis of Spelling Errors. New York: Columbia Univ. Teachers' College.

Morton, J. (1980). "The logogen model and orthographic structure." In Frith, U. (Ed.). Cognitive Processes in Spelling. London: Academic Press.

Nelson, H.E. & Warrington, E.K. (1974). "Developmental spelling retardation and its relation to other cognitive abilities." British Journal of Psychology, 65 (2), 265-74.

Perin, D. (1982) "Spelling strategies to good and poor readers," Applied Psycholinguistics, 2, 1-14.

Simon, D.P. (1976). "Spelling: A task analysis." Instructional Science. 5, 277-302.

Simon, D.P. & Simon, H.A. (1973). "Alternative uses of phonemic information in spelling." Review of Educational Research, 43 (1), 114-137.

Sweeney, J.E. & Rourke, B.P. (1978). "Neuropsychological significance of phonetically accurate and phonetically inaccurate spelling errors in younger and older retarded spellers." Brain & Language, 6, 212-225.

Tenney, Y.J. (1980). "Visual factors in spelling." In Frith, U. (Ed.) Cognitive Processes in Spelling, London Academic Press.

Back to the top.