The Relation between Reading and Spelling in ... - Science Direct

JOURNAL OF MEMORY AND LANGUAGE ARTICLE NO.

39, 264 –289 (1998)

ML982583

The Relation between Reading and Spelling in Skilled Adult Readers V. M. Holmes and J. Carruthers University of Melbourne, Parkville, Victoria, Australia University students made spelling recognition judgements on conventional spellings and misspellings of words. Regardless of how confident they were of their spelling, they could distinguish reliably between the conventional spelling and misspellings when they could spell the word, but they could not distinguish between the conventional spelling and their own misspelling when they could not spell the word. Students also read words that they could not spell as rapidly and as accurately as words that they could spell in a silent reading task. Finally, subjective confidence in spelling was associated with consistency of spelling across trials. Overall, there was no evidence for the claim of a separate reading representation containing orthographic information superior to that in a spelling representation. Instead, the results support the view that a common representation underlies both reading and spelling. Some representations may be incomplete, thus preventing precise spelling, but allowing identification by partial cues. © 1998 Academic Press

In alphabetic languages such as English, progress in acquiring reading and spelling proficiency normally depends upon a firm understanding of the correspondence between the individual sounds of spoken words and the graphemes used to represent them (e.g., Bradley & Bryant, 1983; Byrne & Fielding-Barnsley, 1989; Snowling, 1987). While reading and spelling ability rely on the same basic principle, mastery of English spelling presents a greater challenge than learning to read. The number of different graphemes that can represent a phoneme in English is more varied than the number of pronunciations for particular graphemes. For example, consonants can often be spelled in different ways without changing the pronunciation, as in egg vs log, packs vs tax. Vowels This research was partly supported by a grant from the Australian Research Council. Experiment 1 was conducted by the second investigator as part of a fourth-year honors course. We are grateful to Naomi Brown, Alisha Cooper, and Monica Vomero for research assistance. We thank David Howard for his indirect encouragement to pursue this line of research, and Anne Castles and Chris Davis for helpful discussion and comments on an earlier version of the paper. Chris Barry is also thanked for his extremely thorough and constructive review. Address correspondence and reprint requests to V. M. Holmes, Department of Psychology, University of Melbourne, Parkville, Victoria, 3052, Australia. E-mail: [email protected]. 0749-596X/98 $25.00 Copyright © 1998 by Academic Press All rights of reproduction in any form reserved.

present an even greater problem, as all can be rendered by many different graphemes (e.g., hey, say, weigh, mate, gait, straight) (Barry, 1994; Invernezzi, 1992; Venezky, 1970). As children’s reading vocabulary expands, they encounter an increasing range of possible graphemes symbolizing a given phoneme, including those that are word-specific, that is, used in very few words or just one word (e.g., sign, yacht). The wider choice of elements in spelling than in reading contributes substantially to the greater difficulty of spelling than reading, most people being able to read more words than they can spell accurately. This is true even for advanced learners (Burden, 1992; Fischer, Shankweiler, & Liberman, 1985), and in some cases the gap between reading and spelling proficiency appears sufficiently marked that the person is considered as an ‘‘unexpectedly poor speller’’ (Frith, 1980, 1984). Current models of spelling assume that after repeated exposure to a printed word people develop an orthographic representation in memory which codes the word’s spelling (cf. Brown & Ellis, 1994; Seymour, 1992). People make reference to such a knowledge base when spelling, even in languages with very regular phoneme– grapheme correspondences, rather than assembling spellings exclusively from these

264

READING AND SPELLING IN SKILLED ADULT READERS

rules (Barry, 1992). How the memory representations that people develop for use in spelling relate to the processes involved in reading is a matter of debate. In some models, the orthographic representation formed for use in recognizing words is the same as the one that is accessed for producing spellings. Other models, however, propose that reading and spelling evolve in a sufficiently different manner that two different memory systems are created. The present research aims to provide evidence that might differentiate between these classes of models and, in so doing, clarify how it is that people can read words successfully without necessarily knowing how to spell them. The study focuses upon young adults who are well advanced in literacy learning. According to one model incorporating a single-representation view (Ehri, 1980, 1986, 1991), children’s early orthographic representations contain information, not just about letter sequences, but about the pairings of salient phonemes and their corresponding graphemes. As children increase their printed language experience, their representations typically come to be supplemented by information about words’ morphological structure, analogies with other spellings, any spelling exceptionalities, etc. On this type of view, how could advanced learners read words that they could not spell? For many words, especially short ones, each letter is crucial for distinguishing it from other words in the language. Failure to code the l of the word fold would mean that it would be confused with fond, food, and ford. Words like this would have to have fully specified representations for accurate reading and spelling. However, long words are usually relatively distinctive, containing letters whose identity is not essential for their identification. Someone who did not know whether there was an a or an e in the interior of the word separate would be unlikely to confuse it with other words during reading. Thus, for mature readers, the orthographic representation stored in memory for a word might be indistinct or incomplete, making it inadequate for perfect spelling, but adequate to support identification in reading. The notion that an analysis of only partial

265

cues in the printed stimulus can be sufficient for people to identify words during reading has received some empirical confirmation. Frith (1980) assessed the reading performance of twelve-year-old good and poor spellers who were classed as equally good readers. Although the poor spellers were less adept than the good spellers when reading aloud ‘‘phonetically misspelled’’ text, they lost their disadvantage when reading aloud text with many letters substituted with arbitrary symbols. The poor spellers’ lack of difficulty in reading text with missing letters is compatible with the notion that they were accustomed to recognizing words on the basis of partial information. Poor spellers of this type have particular difficulty spelling words requiring word-specific orthographic information, but their misspellings are usually phonologically appropriate (Burden, 1992; Frith, 1980; Holmes & Ng, 1993), suggesting that they fill in any indeterminate elements from knowledge of familiar phoneme– grapheme correspondences. These findings are consistent with the claim of a common representation that is contacted by the act of spelling as well as the act of reading. Contrasting models of spelling production assume that there is an input orthographic lexicon developed for use in reading, and an output orthographic lexicon developed for use in spelling (e.g., Ellis, 1993; Morton, 1980). The primary impetus for putting forward such a distinction has come from studies of adults known as acquired dyslexics and dysgraphics—people who have suffered brain damage that has impaired their reading and spelling. These individuals sometimes appear to use quite different procedures for reading and spelling. For example, Hanley and Kay (1992) investigated a letter-by-letter reader who could read a word only by laboriously identifying and naming individual letters. However, because his reading was characterized by ‘‘visual’’ errors, such as reading feather as ‘‘further’’, he apparently contacted stored memory representations when reading, and did not merely apply grapheme– phoneme conversions. His spelling was very much worse than his reading, and his spelling mistakes were of a different type. He generally produced ‘‘regularization’’ errors based on pre-

266

HOLMES AND CARRUTHERS

served knowledge of phoneme– grapheme correspondences, such as writing serkel for ‘‘circle.’’ Thus, he seemed not to make use of stored representations in his spelling. A converse pattern was observed by Patterson (1986), in a patient who appeared to have access to stored spelling information, as he could spell irregular words such as yacht just as accurately as regular words such as capsule. However, he was very poor at matching printed words to appropriate pictures, and could not read any of them aloud, suggesting that he did not use the information he had about spelling when he had to read words. An acquired ‘‘surface’’ dyslexic whose major problem was in reading irregularly rather than regularly spelled words has recently been investigated by Weekes and Coltheart (1996). They targeted with a reading program irregularly spelled words that the patient could not read prior to treatment. While his reading of these words improved, his spelling did not, prompting the researchers to favour a model assuming two orthographic lexicons. However, they conceded that a one-lexicon position could explain this result in terms of a single orthographic representation being improved to a level that could support reading by partial cues, but without being adequate for fully precise spelling. They bolstered their preference for a two-lexicon model with other findings indicating apparently different response characteristics for reading and spelling. While both the reading and spelling of two patients were influenced significantly by frequency of occurrence but not imageability, word length did not influence reading, but did influence spelling significantly. In addition, in an evaluation of consistency of reading and spelling for these two cases, in five out of nine analyses that attempted to predict reading from spelling with possible mediating factors such as frequency of occurrence and length of words partialed out, there was no significant relation. However, these additional results are not very damaging to a one-lexicon account. From the word lists provided it is not clear exactly which words were included in the analyses of word length, but they were probably seldom longer than seven letters. A person

could possibly read such words using partial cues without a major impact of word length, whereas attempting to reproduce all letters in a word would evidently be easier the shorter the word was. Finally, since just under half of the consistency analyses did show a significant relation between spelling and reading, the results can hardly be regarded as strong support for the claim of independent systems. When they occur, neuropsychological dissociations between impairments are more striking than associations. However, there have been cases reported whose patterns of responding across reading and spelling have been highly consistent. Friedman and Hadley (1992) studied a patient with letter-by-letter reading and poor spelling, who showed very strong regularity effects and similar regularization errors in both modes. His impairment was considered to result from difficulty in accessing a single orthographic lexicon underlying both tasks. Two other pertinent cases have been described by Coltheart and Funnell (1987) and Behrmann and Bub (1992). Both patients showed a high degree of consistency in the words they could read and spell, even when factors such as frequency of occurrence, word length, and imageability were partialed out. These authors argued for a single orthographic lexicon shared by both reading and spelling. Thus, while there are intriguing dissociations reported for a small number of brain-damaged patients, for other cases a unified interpretation appears sufficient. It is not the present intention to try to resolve this on-going conflict in the neuropsychological domain. It suffices to note that, in attempting to understand the mechanisms underlying normal adult processing, there may be a danger in relying too strongly on analogies from sufferers of brain trauma. For instance, letter-by-letter readers are not ‘‘reading’’ by a process that would be used by normal readers. By definition, they have to gain access to their stored information about printed words in a highly unusual manner. In addition, braindamaged patients often have deficits that extend beyond their difficulties in reading and spelling, many having spoken language and higher cognitive impairments of differing kinds. Thus,


their reading and spelling behavior may be affected in multiple ways, with compensatory strategies having little to do with normal reading and spelling determining performance. Different experimental procedures have been used to attempt to distinguish the single- and dual-representation views in regard to normally functioning literate adults. Some experiments have used a repetition priming procedure. Priming is said to occur when a stimulus which has contacted a lexical representation leaves it in an altered state, allowing an identical or a similar stimulus to contact the record more easily. If the same orthographic representation is accessed for both reading and spelling, then if someone were to write a word down, even if they were prevented from seeing what they had written, this activity should activate the common representation and produce facilitation in later reading of the word. Monsell (1985) tested this assumption by requiring students to write down spellings without obtaining visual feedback, but unexpectedly he found no facilitation from this to subsequent visual recognition. However, in further experimentation, he used a much larger set of items, obtaining significant priming from spelling to reading (Monsell, 1987). Thus, overall, the evidence from the priming procedure provides some support for the one-representation position. Another way to distinguish between the two models rests on a basic premise made by the two-representation position as it applies to normal individuals. The fact that people can often read words that they cannot spell with complete accuracy is assumed to result from the greater accuracy of information contained in the reading lexicon than in the spelling lexicon. If it is assumed that people use their reading lexicon to recognize the accuracy of spellings, then it follows that they might be able to recognize a spelling as correct, even if they could not produce the exact spelling themselves. Campbell (1987) evaluated this claim in an investigation of two university students with reading and writing difficulties. From their written work and repeated spelling tests, she established a set of words which they consistently misspelled in the same way, for example, always spelling tomor-

267

row as tommorrow. These words were then presented individually for recognition, either in their correctly spelled form or in the students’ particular misspelling. Words the students consistently spelled correctly were also presented either in their correct form or in a plausibly misspelled version. When the students could spell a word, they were able to judge the correct spelling as such, as well as being able to detect misspellings reasonably well. However, when they consistently misspelled a word, while they mostly categorized correct versions as appropriate, they were only able to class as wrong about half of their own misspellings. Campbell argued that subjecting their own misspellings to scrutiny in the reading lexicon allowed the students to judge half as wrong because of the more precise information in the reading than in the spelling lexicon. What prevented them from rejecting the other half of their misspellings as inappropriate was that consistently seeing the word written incorrectly by themselves had led to a second (incorrect) representation being established in the reading lexicon. She further maintained that the high acceptance of correct forms of words they couldn’t spell meant that ‘‘these subjects know the correct spellings of words that they consistently misspell’’ (1987, p. 494). However, rather than ‘‘knowing’’ two spellings (the correct one and their own misspelling), the students might have accepted their own misspellings simply because they were uncertain about the spelling, despite having misspelled the word the same way on several occasions. If the students had been able to reject a significant majority of their own misspellings as faulty, this would have provided more conclusive support for a superior reading lexicon. Using a similar methodology, Funnell (1992) obtained results which led her to draw exactly the opposite conclusion. She gave a ten-yearold boy who was a good reader and speller a set of correctly and incorrectly spelled words to categorize as appropriately spelled or not. He was able to discriminate at above-chance level between correct and incorrect versions when he could spell the word, but judged both versions as appropriate for all words he could not spell. Funnell also studied a girl aged twelve years,

268


who was a poor reader and speller. The child was asked to learn to read one set of words and to learn to spell a different set. Subsequently, she could distinguish very well between correct and incorrect spellings of words she learned to spell, but accepted most of the correct spellings and half of the misspellings as appropriate when she had not explicitly learned the spelling. Thus, only when these children knew how to spell a word could they tell when it was misspelled. Funnell (1992) concluded that a single orthographic lexicon accessed for both reading and spelling explained the findings, with learning to spell a word fleshing out a limited orthographic description established through reading. A result in which plausibly misspelled words are always accepted when the correct spelling cannot be produced is inconsistent with the idea of superior information in a distinct reading lexicon, and this was the result obtained for Funnell’s good reader. However, the poor reader in Funnell’s study accepted only half of the plausible misspellings of words she had been asked to learn to read but not spell, actually detecting that the words were misspelled on half the occasions. But we cannot assume that she had not learned to spell these words just because she had not been asked to do so: while her reading of the words was assessed subsequent to training, her spelling was not. It seems likely that she could have unconsciously learned the spelling of some of the items during the period of reading learning. Thus, it remains possible that the misspellings the poor reader detected were of items she had learned to spell, and that the misspellings she accepted were of items she had not learned to spell. If this were the case, her results would be consistent with those of the good reader. However, given the post hoc nature of this account, the findings for this individual are somewhat ambiguous. In the present experiments, we investigated the relation between success in spelling production and performance in spelling recognition in normal adults. We adapted the procedure used by Campbell (1987) and Funnell (1992). People were presented with words for which they could or could not produce correct spellings, and were required to make a judgement about the accu-

racy of the spelling. The goal was to determine whether a single representation was sufficient to explain the pattern of observed performance, or whether it appeared necessary to postulate the existence of two different underlying representations. Both the single-representation and the dual-representation views predict that people should be able to distinguish very well between correct and misspelled versions of words that they can spell. However, different predictions follow for words that people cannot spell. On the single-representation view, they should be unable to discriminate between the correct spelling and phonologically plausible misspellings, including their own. On the dual-representation view, however, if people have a separate reading lexicon containing even some correct representations for words they misspell, while they may not be perfect in picking correct spellings, they should be able to select correct spellings over misspellings to a detectable extent. EXPERIMENT 1 The primary aim of the experiment was to evaluate the ability of skilled adult readers such as university students to discriminate between correct and misspelled versions of words they could and could not spell. Since the intention was to draw conclusions about the performance of normal adults in general, a group design was employed. Accordingly, it was not feasible to evaluate large samples of the spontaneous writings of each individual to search for consistently misspelled words. Instead, spelling knowledge of the words was established in a spelling dictation pre-test, in which students were given a single opportunity to spell a word, the likely consistency of their spelling being assessed by asking them to rate their confidence in the accuracy of their spelling. In this way, each individual supplied a personalized set of correctly spelled and misspelled words, which they could then be asked to evaluate in a recognition task. As well as errors reflecting errors of knowledge or ‘‘competence,’’ spellers can make ‘‘performance’’ errors ‘‘due to some failure in the process of translating (accurate) knowledge


into action’’ (Houghton, Glasspool, & Shallice, 1994, p. 366). Wing and Baddeley (1980) examined all the spelling mistakes in the handwritten examination papers of 40 university students. ‘‘Slips’’ were distinguished from knowledge errors if they were corrected at the time or, if uncorrected, the word occurred at least once elsewhere in correct form. Slips comprised over 70% of the spelling errors. Even in a spelling dictation test, uncorrected performance mistakes might occasionally occur due to temporary memory lapses, motor errors, messy handwriting, and so on. Since it is generally assumed that the speller would be able to detect such mistakes and would know how to correct them if they were confronted with their misspellings, recognizing slips as wrong might be inappropriately taken as providing support for superior knowledge in a recognition system. It was therefore important in this experiment to try to distinguish slips from knowledge errors. Since students of the level being investigated here have a good command of typical correspondences between phonemes and graphemes (Fischer et al., 1985; Holmes & Ng, 1993), a misspelling was regarded as a slip if it was a phonologically implausible rendition of the spoken word, that is, if the letter sequence omitted, added, or distorted one or more phonemes when pronounced according to common grapheme– phoneme correspondences. In addition to performing the spelling recognition task, the same students subsequently saw the correctly spelled versions of the words in a silent reading for meaning task. The aim was to determine whether they could read words that they couldn’t spell as rapidly and accurately as words that they could spell. Both theoretical positions that have been considered would predict that students would be able to read words they couldn’t spell. On the dual-representation view, this would be possible owing to the more precise orthographic information contained in the reading lexicon, and on the single-representation view, students would be able to do so on the basis of partial orthographic information. However, the empirical outcome has not yet been established.

269

Method Participants. The spelling production pre-test was given to 82 undergraduate university students, who were paid a small sum for their participation in the project. All were native speakers of English, ranging in age from 18 to 26 years, with an average age of 19 years 5 months. There were 55 females and 27 males. A sample of 23 students, 18 females and 5 males, returned for participation in the second phase, and was paid a further small remuneration. Spelling production pre-test. The set of 59 words that was chosen for spelling production can be seen in the Appendix. The words were chosen to be difficult for even advanced learners to spell. Most were relatively unfamiliar, with frequencies varying from 1 to 79 per million, with an average of 11, according to Kucˇera and Francis (1967). The words were quite long, ranging in length from 7 to 14 letters, with an average of 9.5. None could be spelled completely accurately without recourse to wordspecific orthographic knowledge. After spelling each word, students rated their confidence in the spelling attempt on a five-point scale ranging from very confident to very unconfident. To check that the students knew the words, we gave them a meaning judgement task. A set of four reasonably common alternative meanings was chosen for each word, only one of which was appropriate. For example, for the target word asymmetric, the following words were presented: (a) mysterious, (b) irregular, (c) ambiguous, (d) tasteful. The target words were randomized and taperecorded. Each word was spoken clearly twice, followed by a pause of 10 seconds in which the student was to write down the spelling. Then a confidence judgement was requested, followed after five seconds by the request to make a meaning judgement. For the meaning judgement, each of the meaning alternatives was spoken once. Ten seconds elapsed before the commencement of the next item. Students were provided with a response booklet in which to record their responses. A space was left for students to attempt a spelling, with boxes alongside for them to check their confidence level.

270

HOLMES AND CARRUTHERS TABLE 1 Sample Items from One Student’s Tailored List for the Spelling Recognition Task (Experiment 1) Confidence in spelling attempt

Spelling production accuracy

Spelling version

Very confident

Quite confident

Unconfident

Can spell

Conventional spelling Most popular misspelling

parallel parrallel

indefinite indefinate

effeminate afeminate

Can’t spell

Conventional spelling Own misspelling

omission ommission

dissimilar dissimiliar

prevalent previlant

They could then follow the four spoken meaning alternatives in printed form. Students were given two practice items to become familiar with the procedure. For their spellings, they were instructed to write the word as correctly as they could. They were told that we were interested in their first attempt, and they were encouraged not to practice the spelling elsewhere or to go back to spellings once they were written down. This task took about 40 minutes to complete. Students were tested in small groups of from one to 10 participants. All students knew the meanings of most of the 59 words, the average number being 52. Spelling scores were based only on words for which students chose the correct meaning alternative. Students spelled an average of 57% of known words correctly, with the range being 15 to 91%. As students who spelled more than 50% of words correctly did not produce enough misspellings for reliable data to be obtained, the second-phase sample was derived from students who produced 50% or fewer correct spellings of known words. The two worst spellers were excluded, leaving 25 for potential participation. From this sample, 23 students remained after two failed to keep appointments. These secondphase students knew the meanings of an average of 50 of the 59 words, with a range from 41 to 58. On the basis of the confidence ratings, spelling attempts were sorted into very confident and quite confident categories, as well as a category termed unconfident for convenience, which comprised responses with a neutral rating or lower. Experimental phase. Tailored lists were pre-

pared for each student. For the spelling recognition task, each word was presented in a misspelled version and its correctly spelled form. To avoid confusion with the use of the term ‘‘correct’’ in other contexts, correctly spelled versions will be termed conventional forms. The misspelled version was the student’s own misspelling when he or she could not spell the word, and an experimenter-generated misspelling when he or she could spell the word. The experimenter-generated misspellings for the 59 words were determined initially from the most popular misspellings made by the 27 students who made more than 50% errors in the spelling production pre-test, and in 13 cases where this did not produce a unique result from 10 additional students scoring between 45 and 50% errors. Table 1 shows sample items for one student. Lists were prepared so that half the words in each condition appeared in their conventional version in the first half of items and half in their misspelled form in the second half of items. Within each half, items were presented in a random order within blocks of items in which each condition was represented. For the reading for meaning task, the target words for each student were the correctly spelled versions of the words that had been presented in the recognition task. Students had to read the target word silently, then read a second comparison word and decide whether it was related in meaning to the first word. Each target word was paired either with the appropriate meaning associate that had appeared in the meaning judgement pre-test (e.g., dissimilar– odd) or one of the unrelated distractors (e.g.,


chandelier– barrier). Half the item pairs within a condition were related in meaning, and half were unrelated in meaning. Students made the meaning decision on the comparison word after the target word had disappeared from view. In this way, reading the target word did not involve a decision component that might be influenced by unintentional variations in the degree of semantic relatedness of the two words. Students were tested individually in the second phase about four weeks after the spelling production pre-test. They completed the spelling recognition task first. This was followed by a 15-minute filler activity involving visuo-spatial judgements and then the reading for meaning task. Finally, the Reading and Spelling subtests of the Wide Range Achievement Test—Revised (WRAT-R) were administered (Jastak & Wilkinson, 1984). These are standardized tests requiring participants to either read aloud or spell individual words of increasing unfamiliarity and length. Each list contains regularly and irregularly spelled words. Ten seconds are allowed for each word in the reading test, and 15 seconds in the spelling test, and testing is terminated when ten erroneous responses are made. The average WRAT-R scores of the second-phase students were 107 for reading and 109 for spelling. The spelling recognition and reading for meaning experiments were controlled by the DMASTR software program developed by K. I. Forster and J. C. Forster at Monash University and the University of Arizona. Items appeared in lower-case letters in the center of a video monitor. For the recognition task, the student was instructed to press a yes button if the item was a correctly spelled word and a no button if the item was not a correctly spelled word. Items stayed on the screen until the student responded, a maximum of six seconds being allowed for responding. For the reading for meaning task, the target word stayed on the screen for a maximum of four seconds or until the student pressed a response button, at which time it was removed and the comparison word appeared in upper-case letters. The student then had to press a yes or a no button to indicate whether or not the second word was related in meaning to the

271

FIG. 1. Percentage ‘‘yes’’ judgments for words the student can spell, as a function of spelling version and confidence in own spelling production attempt (Experiment 1).

first one. Students were given ten seconds in which to respond. The reading time of the target words was the time from presentation of the target word until the button press which removed it from view. The time taken to read and make a decision on the comparison word was also recorded. For each task, several practice items preceded the test items. Students were instructed in both tasks to respond as quickly and accurately as possible. The computer recorded response times in milliseconds and erroneous responses. Response times were trimmed if they exceeded two standard deviations from the mean of all of a student’s response times, and any such times were set at the cutoff value. As each student saw a different subset of the original 59 items, different groupings of items in each condition, and different misspellings of words, item analyses were not possible. Results Figure 1 shows average percentages of yes, correct judgements in the spelling recognition task for words the students could spell. Three students did not rate any words they could spell as unconfident, so in the analyses, an estimate was derived for each score from the mean over the remaining subjects, and the appropriate error mean square and degrees of freedom were adjusted accordingly. Students judged conventional versions as correct significantly more of-

272


ten than they judged misspelled versions as correct, F(1,22) 5 110.88, p , .001, with a significant interaction showing that the difference decreased as confidence in the spelling production decreased, F(2,38) 5 4.22, p , .05. However, even for unconfidently produced spellings, there was significant discrimination between the conventional and misspelled forms, F(1,38) 5 19.22, p , .001. The overall rate of saying yes, correct to items regardless of spelling version did not vary significantly across confidence categories, F(2,44) 5 1.33, p . .05. To exclude performance spelling errors from the analyses of words the students misspelled, we distinguished phonologically inappropriate from phonologically appropriate misspellings. We set generous criteria for regarding a spelling as phonologically appropriate. For example, any combination of vowel letters could represent a schwa vowel, and either single or double letters could represent a consonant. All but one of the students produced some misspellings that were phonologically inappropriate, yielding 65 such misspelled words, with 79 distorted phonemes.1 In analyzing the characteristics of these mistakes, words were pooled across participants. Changes to a phoneme (e.g., atternay) occurred 47% of the time, phonemes were added (e.g., prevenlent) 33% of the time, and phonemes were omitted (e.g., indefin te) 20% of the time. The errors were also classified according to whether the letters involved in the distortion were present but misordered (e.g., dialouge, atteorny, chandlere), the same as others that had been written, were to be written or should have been written (e.g., nouseous, cemetrary, sihlou_te), or different from such letters (e.g., asynetric, curtiousy, flu_esant). A significant majority of the errors (76%) involved misorderings and same letters, compared with 24% involving different letters, x2(1) 5 21.28, p , .001, supporting the idea that many of these mistakes may have resulted from difficulties in place-keeping rather than errors of knowledge. In addition, students rejected their own implau1

Six misspellings of the word bourgeois, ranging from busuar to busirigore, were omitted from this analysis. These misspellings probably resulted from students’ failing to recognize the spoken word.

FIG. 2. Percentage ‘‘yes’’ judgments for words the student can’t spell, as a function of spelling version and confidence in own spelling production attempt (Experiment 1).

sible misspellings most of the time, responding yes to only 24% of the items. In comparison, they correctly accepted 70% of the corresponding conventional forms. This difference was significant, t(21) 5 4.34, p , .001. These trials were not included in subsequent analyses. For the remaining words the students could not spell, the spellings were phonologically plausible. Figure 2 shows average percentages of yes responses to these items. There was no evidence that students could discriminate reliably between the conventional spelling and their own misspelling, F , 1, and this null effect was observed across all three confidence categories, F(2,44) 5 1.20, p . .05. Overall, significantly fewer yes responses were made as confidence in the original spelling attempt decreased, F(2,44) 5 3.45, p , .05. Although the words occurring in the different conditions were different for each student, they might nevertheless have differed systematically in relevant properties such as length and frequency of occurrence. If this were the case, analyses of response times across different conditions would be difficult to interpret. These values were determined for the words of each student and then averaged across students. Table 2 shows the means, in addition to the average number of words that fell into each category. Understandably, students gave confident judgements more often to words they could spell than to those they could not. Scores were estimated for the three students who had rated

273

READING AND SPELLING IN SKILLED ADULT READERS TABLE 2

Average Frequency of Occurrence and Length in Letters of Words as a Function of Spelling Production Accuracy and Confidence in Spelling, and Average Number of Words Falling into Each Category (Experiment 1) Confidence in spelling attempt Spelling production accuracy

Very confident

Frequency

Can spell Can’t spell

18 19

(5) (10)

Length


8.8 (0.3) 9.6 (0.5)

Number of words


9 6

(4) (3)

Quite confident 13 10

(6) (3)

9.4 (0.7) 9.7 (0.4) 7 11

(3) (5)

Unconfident 8 7

(5) (3)

9.4 (0.6) 9.7 (0.3) 4 13

(3) (7)

Note. Standard deviations are in parentheses.

no words they could spell as unconfident. Analyses showed that decreasing confidence was associated with a decrease in word frequency, regardless of spelling production accuracy, F(1,22) 5 33.02, p , .001. However, frequency of words did not differ significantly as a function of spelling knowledge, F(1,22) 5 2.18, p . .05, and this was the case for each confidence level, F(2,41) 5 1.33, p . .05. By contrast, words students could spell were significantly shorter than words they could not spell, F(1,22) 5 28.41, p , .001, and very confidently rated words were shorter than less confidently rated words, F(2,44) 5 6.52, p , .01. The difference between word lengths in the can and can’t spell conditions was larger for very confidently rated words than less confidently rated words, F(2,41) 5 3.60, p , .05. Because words differed in length as a function of spelling production accuracy, and differed in length and frequency as a function of confidence, direct comparisons of response times were made only between conventional and misspelled versions of the same word. Although students did not see the two versions at the same time, the items could be placed into four categories: those for which both versions were accepted as correct (both accepted), those for which the conventional version was judged as correct and the misspelled version rejected (conventional selected), those for which the misspelled version was judged as correct and

the conventional version rejected (misspelled selected), and those for which both versions were rejected (both rejected). For words the student could spell, spellings that had been produced very confidently and quite confidently were grouped together to provide enough data. Not all students yielded responses in two of the response categories. However, as separate analyses were performed on each response category, it was not necessary to estimate missing values, merely to analyze the data from those students who contributed data. Table 3 gives the mean response times to make judgements on the conventional and misspelled versions, along with the number of students who contributed values for each category, and the mean frequency and length of items falling into each category. When students accepted both the conventional and the misspelled version as correct, they took significantly longer to respond yes to the misspelled than to the conventional form, t(22) 5 2.33, p , .05. Comparisons between judgement times for the other three response categories were not significant: when the conventional form was selected, t(22) 5 1.42, p . .05; when the misspelled form was selected, and when both forms were rejected, both ts , 1. There were not enough unconfident judgements of words the student could spell for meaningful subdivision of the items. The same subcategorization of responses was made for words students could not spell. For

274

HOLMES AND CARRUTHERS TABLE 3

Mean Spelling Recognition Response Times (in ms), Number of Contributing Students, and Mean Frequency and Length of Words as a Function of Response Category for Words Students Could Spell, Produced Confidently (Experiment 1) Spelling version Response category Both accepted Conventional selected Misspelled selected Both rejected

Conventional 1861 1828 2427 2169

(491) (546) (905) (678)

Misspelled 2160 1947 2487 2050

Number of students

(787) (485) (975) (1018)

23 23 15 10

Frequency 20 15 16 19

(14) (5) (16) (15)

Length 9.1 9.1 9.1 9.8

(0.5) (0.4) (0.9) (1.7)


these words, very confidently and quite confidently produced spellings were again classed together, and the mean response times, number of contributing students, and mean frequency and length of items falling into each category are given in Table 4. By contrast with the results for words the student could spell, students took no longer to accept the misspelled form than the conventional form when they accepted both versions, t , 1, but when they selected their own misspelling, they did this more quickly than they rejected the conventional form, t(21) 5 2.50, p , .05. There was no significant difference between times to make judgements about the conventional and misspelled forms when the conventional spelling was selected, t , 1. The apparently longer times to reject the misspelled rather than the conventional form when both versions were rejected was far from

significant, t(13) 5 1.47, p . .05, as students rejected both forms relatively infrequently, and with a large amount of variability in response time. Response times to conventional and misspelled versions of words the students had misspelled unconfidently were also analyzed. The variability in these times was substantial, and the means for the conventional and misspelled forms within the four response categories did not differ significantly. For the silent reading for meaning task, students took 1468 ms on average to read the comparison words on correct trials. Students classified comparison words incorrectly on 9% of trials, even though they had chosen the correct alternative in the spelling production pretest. Most of the errors were failures to respond yes, typically occurring on trials when the alternative provided appeared to be less closely as-

TABLE 4 Mean Spelling Recognition Response Times (in ms), Number of Contributing Subjects, and Mean Frequency and Length of Words as a Function of Response Category for Words Students Could Not Spell, Produced Confidently (Experiment 1) Spelling version Response category Both accepted Conventional selected Misspelled selected Both rejected

Conventional 2093 2141 2385 2150

(610) (634) (813) (919)


Misspelled 2089 2122 2019 2438

(649) (659) (686) (1009)

Number of students 23 23 22 14

Frequency 12 11 16 10

(6) (10) (11) (10)

Length 9.7 9.4 9.6 9.9

(0.3) (0.7) (0.7) (0.7)

275

READING AND SPELLING IN SKILLED ADULT READERS TABLE 5

Mean Percentage Classification Errors and Mean Reading Time per Letter (in ms) on Correct Trials as a Function of Spelling Production Accuracy and Confidence in Spelling in Silent Reading for Meaning Task (Experiment 1) Confidence in spelling attempt Spelling production accuracy Percentage errors


Reading time


Very confident 4.2 (6.5) 6.0 (10.9) 119 117

(43) (45)

Quite confident 10.0 (10.8) 10.0 (10.7) 119 119

(40) (37)

Unconfident 12.7 (16.4) 14.1 (10.6) 131 120

(49) (46)


sociated to the target word than most of the other items. Table 5 shows the average percentages of occasions on which students misclassified comparison words. In the analysis of these values as well as in the subsequent analysis, scores were estimated for the three students who did not rate unconfidently any words they could spell. Students’ misclassification rate of comparison words increased significantly as their confidence in the spelling of the target word declined, F(2,44) 5 6.40, p , .001. This result was probably due to the lower frequency of occurrence of words with less confidently rated spellings. Importantly, students did not misclassify comparison words following words they could spell any more often than words they could not spell, F , 1, and there was no interaction of this null effect with confidence category, F , 1. Students identified prior target words with equal success regardless of knowledge of the precise spelling. Although words the students could spell were no more frequent than words the students could not spell, they were significantly longer. Because of this, the reading time for each target word was divided by the number of letters in the word to yield reading time per letter, and these values were averaged for the items in each category for each student. The means across students, which are given in Table 5, were virtually identical for words the students could and could not spell for each confidence category. Nonsignificant main effects of confidence, F(2,44) 5 1.11, spelling knowledge, F(2,44) 5

1.32, and their interaction, F , 1, were obtained. This result shows that students took no longer to read words whose precise spelling they did not know than words whose spellings they did know. Discussion These results demonstrate that students’ ability to recognize the appropriateness of spellings depended on whether or not they could produce the correct spelling themselves. For words that they could spell, students could discriminate quite well between conventional spellings and experimenter-generated misspellings. When they were confident of their spelling, and they did fail to discriminate between the conventional version and the experimenter-generated misspelling, they took longer to accept the misspelled than the conventional form. The results for words that they could not spell were quite different. There was no sign that they could recognize conventional forms over their own misspelling, except when their misspelling led to a phonologically implausible rendition of the spoken form, where the error was probably due to faulty translation or place-keeping rather than a knowledge error. In addition, for confidently misspelled words, when students accepted both the conventional form and their own misspelling, they took the same amount of time to reach decisions in both cases, consistent with the idea that they felt equivocal about the correct response. Further, when they showed ‘‘reverse’’ discrimination, that is, by preferring their own

276


misspelling over the correct version, they were slower to decide upon the conventional form than they were to accept their own misspelling, suggesting some residual hesitation in deciding that the conventional form was wrong. Because these findings did not reveal any ability on the part of the students to recognize the correct spelling of words they could not spell, they are clearly more supportive of the one-representation view than the two-representation view of the organization of spelling and reading. Nevertheless, when students could spell a word and were confident of their spelling, although they could discriminate reliably between conventional and misspelled versions, their performance was not as good as might have been expected on the one-representation view. Students frequently accepted the most popular misspelling even when they apparently knew how to spell the word. A possible explanation is that the single-word presentation conditions, in conjunction with the requirement to make a speeded response, may have underestimated the capacity of the students. They may have adopted a less stringent criterion for acceptance of an item than if they had had more time to make a decision or been forced to discriminate between simultaneously presented different spellings. Finally, the results of the silent reading for meaning task were in accordance with expectation. In conditions that placed no emphasis on knowing the exact spelling of a word, students could identify a word when they did not know the exact spelling just as rapidly and accurately as when they did. While this finding is explicable by both theoretical conditions being considered, it is of interest on the one-representation account that a partial-cue reading strategy apparently can be effective for words such as those used here. EXPERIMENT 2 In Experiment 1, students discriminated reliably between conventional and misspelled versions of words they could spell. However, they accepted plausible experimenter-generated misspellings of these words at a higher rate than might have been anticipated. In this experiment

we attempted to evoke even better discriminatory performance by two means. One was to take away any time constraints on their decisions, and the other was to allow them to compare directly conventional spellings and misspellings on the same trial. Consequently, students’ spellings of difficult-to-spell words were gathered and their confidence in the spelling attempts was assessed in the same way as before. Words were again categorized as those they could and could not spell, and grouped into three confidence categories. In the second phase, students were presented with several spelling versions of a word at the same time and asked to choose which one was correct, with no time pressure on responding. While we expected that students would discriminate even better than in Experiment 1 among versions of words they could spell, we still expected that they would not be able to select the conventional version over their own plausible misspelling when they could not spell a word. Method Participants. A spelling production pre-test was administered to 97 Psychology 1 students, all of whom were native speakers of English, aged between 17 and 26 years, with an average of 18 years 11 months. There were 68 females and 29 males. A sample of 20 students, 15 females and 5 males, completed the second phase of testing. Students received course credit for their participation. Spelling production pre-test.Of the 59 words comprising the spelling production test used in Experiment 1, three words were discarded as being too easy to spell, leaving 56 items, as indicated in the Appendix. The list was recorded in the same way as before, with one modification. To reduce the amount of time spent on the meaning evaluation task, the meaning alternatives were not recorded onto the tape, students being left to read the alternatives for themselves in the response booklet. Students were tested in small groups of from 1 to 12 participants. As before, spelling performance was based only on words for which students checked the correct meaning alternative. Across all students,

277

READING AND SPELLING IN SKILLED ADULT READERS TABLE 6 Sample Items from One Student’s Tailored List (Experiment 2) Words the student can spell Confidence in spelling attempt

Conventional

Most popular misspelling

Very confident Quite confident Unconfident

chandelier occurrence zucchini

chandalier occurance zuchini

Other misspellings chandellier occurence zuccini

chandeleir occurrance zuchine

Words the student can’t spell Confidence in spelling attempt

Conventional

Own misspelling

Very confident Quite confident Unconfident

silhouette plagiarism rendezvous

sillohette plagarism rhondevou

the average number of words whose meanings were known was 48. The average percentage of correctly spelled known words was 47, with a range of 12 to 83. The three worst spellers were excluded from second phase participation. A total of 60 students made between 45 and 77% spelling errors on known words. From these, 36 made themselves available for further testing, 20 for the present experiment, and 16 for another experiment being conducted in the laboratory. Allocation to a particular experiment was random. The misspellings of the 60 potential participants were recorded and used to determine a ranking of popularity of misspellings of each of the 56 words. Spelling recognition. The 20 students who were tested in this phase knew an average of 47 words, ranging from 40 to 53. They spelled correctly an average of 40% of known words, ranging from 25 to 53. On average, for words they could spell, students spelled 6 very confidently, 6 quite confidently, and 5 unconfidently; for words they could not spell, they misspelled 6 very confidently, 9 quite confidently, and 13 unconfidently. Tailored lists were again constructed for each student. Table 6 gives some sample items from one student. A list was prepared in which each conventional version was associated with the most popular misspelling and the two misspellings next on the popularity list. (Sometimes, for words for which only one

Other misspellings sillouette plagerism rendevous

silohette plaigarism rhondevous

silloette plagurism rondezvou

or two different misspellings had been produced, it was necessary to invent a plausible misspelling.) When students could spell a word, they were presented with these four alternatives. When students could not spell a word, they were presented with five alternatives: the conventional version, their own misspelling, the most popular misspelling (if this was different from their own misspelling), and either two or three other misspellings which contained elements of the correct spelling. Each spelling alternative was typed in large lower-case font on a strip of paper, with a random identifying letter at the side. The relevant combinations of alternatives for each student were sorted into individual piles and placed in numbered sleeves in a folder. Students returned three to four weeks after phase-one testing. They were tested individually in a quiet room, seated at a table. On each trial, they took a set of paper strips from the folder and compared all spelling alternatives at the one time. They were encouraged to ‘‘shuffle them around and try out different orders,’’ then to ‘‘rank the spellings according to how closely they resemble the correct spelling of the word.’’ They were instructed that the most important one to get right was the correct spelling, and not to worry too much about the order of the other alternatives. When they had placed the spellings in a final ranked order, they recorded the corre-

278


FIG. 3. Percentage of times item is ranked first in recognition for words the student can spell, as a function of different spelling versions and confidence in own spelling production attempt (Experiment 2).

sponding letters on a response sheet. Two practice items were given for students to familiarize themselves with the procedure. The task took from 40 to 60 minutes to complete, after which time students were given a sheet on which the words they had just judged were typed in two double-spaced columns, and they were asked to read each one aloud as clearly and as accurately as possible. This task was tape-recorded for later correction. Any mispronounced words were discarded from the analyses of recognition performance. Students mispronounced on average just over one word. Results In Figure 3, the average percentages of times students judged a spelling as correct are shown for words they could spell. The results indicate that students vastly preferred the conventional version over any misspellings, particularly when they were very confident and quite confident of their spelling. It was not appropriate to conduct parametric statistics on the very confident and quite confident categories, as almost all of the two misspelling conditions contained zeros. For these categories, all students chose the conventional version more often than the most popular or other misspellings, both of which were rarely chosen. For unconfidently produced spellings, analysis of variance was

possible.2 Two students did not rate any of their correct spellings as unconfident, so analyses were performed on the remaining 18 students. These students chose the conventional version significantly more often than the most popular misspelling, F(1,17) 5 15.32, p , .01. However, they did not choose the most popular misspelling significantly more often than any other misspelling, F(1,17) 5 1.98, p . .05. For words the student could not spell, trials with phonologically implausible spellings were first isolated and analyzed separately. Of the 20 students, 14 produced four phonologically implausible spellings on average. This resulted in 55 phonologically implausible misspellings, with 61 distorted phonemes. The types of mistake and their distribution were very similar to those in Experiment 1. Phonemes were changed in 28% of mistakes, added in 46% of mistakes, and omitted in 26% of mistakes. As before, many more of these errors (74%) involved misordered letters or letters that either were elsewhere in the spelling or should have been in the spelling rather than different letters (26%). In a pooled analysis of frequencies, this difference was significant, x2(1) 5 13.79, p , .001. For these phonologically implausible spellings, students selected their own misspelling only 9% of the time, choosing the other three misspellings combined 23% of the time, a difference which was significant, F(1,13) 5 5.04, p , .05. They chose the conventional spelling 68% of the time, also significantly more often than their own misspelling, F(1,13) 5 22.48, p , .001. For the trials on which the students’ misspellings were phonologically plausible, the percentages of times the conventional version, the student’s own misspelling, and any other misspellings were deemed the correct spelling are shown in Figure 4. In an analysis comparing the conventional spelling and the student’s own misspelling, there was no significant overall effect of spelling version, F , 1. However, students chose both these versions significantly less often as their confidence declined, F(2,38) 5 3.84, p , .05. As the interaction between 2

Note that the three conditions cannot be compared in the one analysis, as each student’s values add up to 100.


FIG. 4. Percentage of times item is ranked first in recognition for words the student can’t spell, as a function of different spelling versions and confidence in own spelling production attempt (Experiment 2).

between spelling version and confidence was not far from significance, F(2,38) 5 2.75, p . .05, it was analyzed further. Students did not discriminate significantly between the conventional version and their own misspelling for all confidence categories: for very confident misspellings, F(1,38) 5 2.39, p . .05, quite confident misspellings, F , 1, and unconfident misspellings, F(1,38) 5 2.56, p . .05. An analysis was then performed comparing a student’s misspelling with other misspellings combined. Overall, students significantly preferred their own misspelling over other misspellings, F(1,19) 5 12.54, p , .01, and while there was no main effect of confidence, F , 1, the interaction between spelling version and confidence was significant, F(2,38) 5 5.72, p , .01. Contrasts showed that the interaction resulted primarily from students’ preference for their own spelling over other misspellings being significant for very confidently produced misspellings, F(1,38) 5 16.33, p , .001, but nonsignificant for quite confidently produced misspellings, F(1,38) 5 3.33, p . .05, and unconfidently produced misspellings, F , 1. Discussion The results of this experiment show that in a forced-choice procedure with no time restrictions, students were able to match the presented letter sequences very closely with their stored representations. When they could produce the

279

correct spelling of a word, especially when they were confident of their spelling, students subsequently exhibited almost perfect discrimination between the conventional form and any misspellings. We cannot be certain which of the two changes in procedure from Experiment 1 was more important in causing students to improve their ability to reject experimenter-generated misspellings. Nevertheless, the implication for the performance of students investigated in Experiment 1 is clear. Evidently, in evaluating the presented spellings, they did not always access all the information their orthographic representations contained. When students could not produce the correct spelling of a word, the results were the same as those obtained in Experiment 1. They showed no sign of being able to select the correct version over their own misspelling, except when the misspelling was probably the result of an uncorrected slip which resulted in the production of a phonologically implausible spelling. Indeed, for very confidently produced misspellings, students tended to prefer their own misspelling over the correct one, and certainly preferred it above any other plausible misspellings. Taken together, these findings reveal a very intimate correspondence between the spelling production and spelling recognition systems. EXPERIMENT 3 In Experiments 1 and 2 the procedure of obtaining confidence ratings was used to provide a quick guide to spelling consistency. We assumed that when students displayed high confidence in the correctness of a spelling, this was because they were likely to spell the word in the same way on other occasions. The aim of Experiment 3 was to see whether this assumption of a correspondence between confidence and consistency was justified. Students spelled a list of words three times in the one experimental session, rating their confidence in the spellings as before. The trials were separated by a period of intervening activity. It might be argued that students would produce more consistent spellings in this situation than if the trials were far apart in time. However, while this is a possibility that should be investigated, we felt that the

280


current procedure would afford at least a preliminary answer to the question. For the intervening activity, students had to perform a difficult reading comprehension task. The act of reading many words in different passages and answering questions about them should provide some interference with temporary memory for any suddenly invented spellings. None of the test words occurred in the reading comprehension passages. Method Participants. A sample of 44 students was obtained from the Psychology 1 Subject Pool. All were native speakers of English, and were aged between 18 and 23 years, with an average of 19 years 5 months. There were 10 males and 34 females. None had participated in the previous experiments. Materials. To ensure that the duration of the testing session did not overtax students’ attentiveness capacities, the spelling list comprised only a subset of the words used in the previous experiments. Data from the 82 students who completed the pre-test in Experiment 1 were used to help choose the items. First of all, any words whose meanings were not known by at least two-thirds of the sample were discarded. Then any words misspelled by only a quarter of the students or fewer were excluded as too easy, and any words misspelled by three quarters or more students were excluded as too hard. From the remainder, 30 words were selected, whose meanings were known to 92% of the students on average, and which, when known, were misspelled by 50% of students on average. The words averaged 9.6 letters in length and had an average frequency of occurrence of 9 per million. The words selected are given in the Appendix. The reading comprehension task chosen as the filler activity was Scale Three of the 1957 version of the Australian Council for Educational Research Reading Comprehension Test (Form Y). This scale comprises 7 passages of literary prose followed by multiple-choice questions which assess the content, probable theme, and likely inferences of a passage. Only three students completed all of the total of 30 questions.

Design, procedure, and analysis. Spelling lists were presented via tape recorder to small groups of students. On trial one, each target word was preceded by a sentence in which it appeared, usually as the last word, for example, ‘‘To serve you has been a privilege—privilege.’’ On trials two and three, the target words were presented in a different random order, spoken twice with no sentence context. Students were given ten seconds to write the spelling of the word on a response sheet, and then to check one of five boxes to the right to indicate their confidence in the accuracy of their spelling. After the first trial, they spent ten minutes reading passages and answering questions from the reading comprehension task. Then the second spelling trial was conducted, followed by ten more minutes for performing the reading comprehension task, and finally the third spelling trial was completed. Each spelling trial used separate response sheets, with the experimenter removing earlier sheets. Students were told that both their reading and their spelling were being assessed. The spelling responses of each student were classified into those that were correct on the first trial and those that were incorrect on the first trial. The confidence ratings were given numerical values ranging from 1 for very confident ratings to 5 for very unconfident ratings. Subsequently the responses were grouped into very confident, quite confident and unconfident categories, as before. The data from the two worst spellers, who spelled fewer than a quarter of the words correctly, were removed. The data from three more spellers at the poorer end of the distribution were also rejected, owing to their producing no spellings very confidently. On the basis of the percentage of correct spellings on trial one, the remaining 39 students were divided into three groups of 13, labeled for simplicity good, average, and below average. The average and below average groups correspond roughly to the samples used in Experiments 1 and 2. To gain an idea of how well a spelling produced at any slice in time predicted later spelling, spellings on trials two and three were evaluated relative to the spelling on trial one. They could either be the same on all three trials,

281

READING AND SPELLING IN SKILLED ADULT READERS TABLE 7 Mean Percentage Correct Spellings and Mean Confidence Ratings as a Function of Trial and Spelling Group (Experiment 3) Spelling group Percentage correct

Good Average Below average Mean

Confidence rating


Trial one 58 42 30 43 2.0 2.4 2.6 2.3

(11) (4) (3)

(0.5) (0.4) (0.4)

Trial two 57 47 29 45 1.9 2.2 2.4 2.2

(11) (6) (5)

(0.5) (0.4) (0.4)

Trial three 55 47 31 44 1.8 2.1 2.2 2.0

(13) (6) (6)

(0.5) (0.4) (0.4)


different once from trial one spelling, or different twice from trial one spelling. (The latter category included trials on which the same new spelling was given on both trials two and three.) In the analyses, any missing scores were replaced by the means of the other subjects, and the error mean squares and degrees of freedom were adjusted.3 Results Table 7 shows the mean numbers of correctly spelled words for each spelling group across trials. Naturally, the good spellers outperformed the average and below average spellers significantly, F(1,36) 5 62.01, p , .001, and average spellers outperformed the below average spellers significantly, F(1,36) 5 28.32, p , .001. Across the three groups, the number of correctly spelled words did not change significantly over trials, F , 1, although there was a significant interaction between spelling group and trial, F(2,72) 5 3.67, p , .01. Good spellers became slightly less accurate across trials, while average spellers became slightly more accurate, and below average spellers remained the same. These changes in overall accuracy were very small. Table 7 also gives the mean confidence 3 There were 8 students with a particular confidence category containing no responses: 5 students had no unconfident correct spellings, one had no very confident correct spellings, one had no very confident incorrect spellings, and one had no unconfident incorrect spellings.

ratings on each trial for the three groups. As might be expected, good spellers were in general significantly more confident of their spellings than average and less average spellers, F(1,36) 5 6.30, p , .05, though average spellers were not significantly more confident than below average spellers, F , 1. All students became less confident in their spellings over trials, F(2,36) 5 34.06, p , .001, and there was no interaction of this effect with spelling group, F , 1. For each student, three correlations were obtained across the 30 words between their confidence ratings on trials one and two, on trials two and three, and on trials one and three. Table 8 provides means of these correlations for the three spelling groups. Analysis of variance was performed on the correlations. The overall correlation did not vary significantly as a function of spelling goup, F , 1. However, the correlation between temporally contiguous trials was significantly higher than the average of the two correlations between temporally separated trials, F(1,36) 5 13.13, p , .01. The correlation across trials one and two was nonsignificantly lower than that across trials two and three, F(1,36) 5 3.95, p . .05. These effects did not interact with spelling group, F , 1. For words the students could spell on trial one, the mean percentages of words spelled in the same way on all three trials are shown in

282

HOLMES AND CARRUTHERS TABLE 8 Mean Correlations of Confidence Ratings on Pairs of Trials as a Function of Spelling Group (Experiment 3)

Spelling group

Trials one & two

Trials two & three

Trials one & three


.66 (.21) .67 (.14) .74 (.14) .69

.71 (.22) .72 (.17) .78 (.10) .74

.59 (.27) .67 (.20) .66 (.16) .64


Table 9. There were no significant differences among these means: for spelling group, F(2,36) 5 1.25, p . .05, for confidence in spelling attempt, F(2,66) 5 1.57, p . .05, and for their interaction, F , 1. Students spelled words the same way three times on 84% of trials, a value which was significantly higher than 50, F(1,66) 5 274.83, p , .001, indicating that spelling, once correct, was more likely to stay that way than to change. Table 9 also gives the mean percentages of words that were spelled once more as on trial one, or not spelled again as on trial one. In the analysis of these data, no effects approached significance. The results for spelling consistency for words the students could not spell on trial one, shown in Table 10 contrast with those for words the

students could spell. Analyses were again performed first on words that students spelled in the same way on all three trials. Good spellers produced the same misspelling significantly more often than did average and below average spellers, F(1,70) 5 6.35, p , .05, and average spellers tended to do so more often than did below average spellers, F(1,70) 5 3.72, p . .05. For all spellers, very confidently produced misspellings were consistent more often than quite confidently produced misspellings, though the effect was not significant, F(1,36) 5 3.61, p . .05. However, these words were spelled consistently significantly more often than unconfidently spelled words, F(1,36) 5 25.99, p , .05. The interaction between spelling group and confidence in spelling was far from signif-

TABLE 9 Percentage of Words Spelled in Different Consistency Categories as a Function of Spelling Group and Spelling Confidence for Words Students Could Spell (Experiment 3) Confidence in spelling attempt Spelling consistency category

Spelling group

Very confident

Quite confident

Unconfident

Same spelling three times

Good Average Below average

91 (18) 92 (11) 85 (32)

86 (14) 86 (30) 75 (22)

77 (18) 85 (23) 80 (21)

Different once from trial one


5 (11) 5 (9) 15 (32)

7 (11) 11 (28) 15 (20)

4 (8) 10 (17) 10 (14)

Different twice from trial one


4 (11) 3 (8) 0 (0)

7 (14) 3 (8) 10 (14)

19 (21) 5 (10) 10 (17)


283

READING AND SPELLING IN SKILLED ADULT READERS TABLE 10 Percentage of Words Spelled in Different Consistency Categories as a Function of Spelling Group and Spelling Confidence for Words Students Could Not Spell (Experiment 3) Confidence in spelling attempt Spelling consistency category Same spelling three times

Different once from trial one

Different twice from trial one

Spelling group

Very confident

Quite confident

Unconfident


89 (14) 75 (30) 74 (26)

80 (21) 64 (28) 60 (36)

53 (38) 59 (20) 30 (20)


6 (12) 7 (10) 12 (21)

10 (17) 20 (17) 23 (33)

17 (31) 11 (14) 21 (20)


5 (11) 18 (30) 14 (16)

10 (13) 16 (16) 17 (22)

30 (30) 32 (19) 49 (22)


icant, F(4,70) 5 1.24, p . .05. Finally, the value required for each of the means reported in the table to be significantly higher than 50, with p , .05, was determined. A percentage of more than 64 was required. In the analysis of the percentages of words that students misspelled differently after trial one, the main effects of spelling group and confidence and their interaction were of no interest as they were simply the inverse of the effects analyzed previously. Students misspelled significantly more words differently twice than once, F(1,36) 5 6.62, p , .05. This effect interacted significantly with confidence category, F(1,68) 5 5.11, p , .01, though not with spelling group, and the threeway interaction was also nonsignificant, both Fs , 1. It is evident that for very confidently and quite confidently produced misspellings, spellings were changed once about as often as they were changed twice. It was only for unconfidently produced misspellings that spellings differed from trial one spelling more often twice than once. Discussion The average and below average spellers tested in this experiment were approximately

equivalent in spelling ability to the students tested in the second phases of Experiments 1 and 2. The students had a realistic view of their spelling ability: poorer spellers were less confident of their spellings than were better spellers. Nevertheless, both weaker and better spellers rated confidence in their spelling in a highly stable manner across trials. In addition, regardless of their confidence rating, when students could spell a word correctly on the first trial, they almost always spelled the word the same way on the next two trials, and this was true for all spelling groups. However, when students could not spell a word on the first trial, they were subsequently more variable in their spelling. It is interesting that the good spellers were actually more consistent misspellers than were average and below average spellers. If we define consistent misspelling as spelling a word in the same way on all trials over 64 percent of the time, then good spellers consistently misspelled words when they were very confident and quite confident of their misspelling, and average and below average spellers consistently misspelled words when they were very confident. Overall, the results justify using the proce-

284


dure of obtaining confidence ratings to provide an approximate estimation of spelling consistency. GENERAL DISCUSSION The findings that have been obtained in the study are clear-cut. Experiments 1 and 2 showed that when students could spell a word, they could subsequently distinguish reliably between the correct spelling and plausible experimenter-generated misspellings. However, when they could not spell a word, they were unable to discriminate between the correct spelling and their own misspelling. This was the case for spellings students had produced with very high confidence as well as spellings about which they felt less certain. Further, in Experiment 2, when students were very confident of their original misspelling, they chose their own misspelling significantly more often than any other misspellings. Experiment 3 provided results indicating that asking spellers to rate their confidence in the accuracy of their spelling provides a reasonable short-cut procedure for estimating spelling consistency. Students of similar spelling ability to those investigated in Experiments 1 and 2 consistently misspelled words about whose spellings they were very confident approximately three-quarters of the time. This suggests that in Experiments 1 and 2, students might have altered some of their very confident misspellings on another occasion. However, if such inconsistently misspelled words had somehow been excluded, our basic conclusions would not change. Given that students increasingly preferred their own misspelling over others as they became more confident of their spelling, it is hard to believe that they would have been better able to pick conventional spellings when all misspellings were consistent. Indeed, their attachment to their own misspelling would probably have appeared even stronger if only completely consistently misspelled words had been considered. If the students had information available to them for reading that was superior to the information used for spelling, they did not reveal it in our experiments. If they possessed this

knowledge, they should have been able to reject their own misspellings in favor of conventional spellings. Rather than bolstering the claim of distinct information in two separate lexicons, the results provide strong support for the idea that normal adult readers access the same orthographic representation for both reading and spelling, in line with the theorizing of Ehri (1980, 1986, 1991) and Funnell and colleagues (Coltheart & Funnell, 1987; Funnell, 1992). The results of the silent reading task of Experiment 1 are also consistent with this model. Students were able to read words that they could not spell as rapidly and as accurately as words that they could spell when they did not have to evaluate spelling accuracy. This result can readily be interpreted in terms of successful identification being based on partial information only—at least on those letters that would allow the reader to distinguish the word from any similarly spelled words in the lexicon. Of course, we recognize that the two-representation model would also expect this outcome on the assumption of more detailed orthographic information in the reading lexicon. A possible problem with the silent reading results could be thought to stem from the fact that students performed the task not long after they had seen both correct and incorrect forms of the target words. Perhaps this experience had the result of blurring all their orthographic representations to such an extent that none remained correct by the time of testing. Although this prospect seems to us unlikely, it would be worthwhile for this experiment to be conducted on a separate sample of participants. The one-lexicon framework also accounts for the finding that students were able to discriminate much better in Experiment 2 than in Experiment 1 between conventional spellings and plausible misspellings of words they could spell. In Experiment 1, when just one item was presented on a given trial and students were under time pressure, they frequently accepted the most popular misspelling as well as the conventional spelling as correct. However, in Experiment 2, when they were allowed to compare alternative versions, and work at their own pace, discrimination between conventional and


incorrect forms was extremely high. Taken together, these results suggest that when students make reference to their orthographic representation to perform a spelling check, they may sometimes fail to retrieve all the information it actually contains. Given that the spelling recognition procedure is basically the same as a difficult lexical decision, the present result is analogous to criterion effects found in lexical decision tasks when the degree of similarity of nonwords to real words is varied. For example, semantic effects obtained on word decisions when the distractors comprise orthographically legal nonwords are eliminated when nonwords are orthographically illegal (James, 1975; Shulman & Davison, 1977). More recently, Stone and Van Orden (1993) showed that word frequency effects are significantly larger when nonwords sound like words than when they are pronounceable but do not sound like words. Additionally, unpublished data from our own laboratory have shown substantial effects of orthographic complexity and word length on lexical decisions when the nonwords are derived from real words with a single letter substitution or transposition, but the effects are markedly reduced when the nonwords, while orthographically legal, do not resemble any particular word closely. What these results suggest is that the lexical decision procedure can be more like a reading task or more like a spelling recognition task depending on the difficulty of the discrimination required. A reader would check the identity and order of letters against their stored knowledge only to a level that was necessary according to some subjective criterion level of accuracy. Ehri (1980, 1986, 1991) has contended that the orthographic representations of mature readers do not simply comprise linearly ordered sequences of letters, but also contain stored information about spelling patterns and morphological relations shared with other words, as well as notes about unusual characteristics. It is these word-specific orthographic ‘‘details’’ of the spelling of long words of the type investigated here that readers would optionally access, depending on the depth of processing

285

required by the task. In normal reading, when people have to discriminate words only from other potential words, they may be able to adopt a shallow procedure for identifying long words in order to enhance identification speed. The fact that readers could read words for meaning in Experiment 1 regardless of exact spelling knowledge is consistent with this possibility. Indeed, readers may be able to control strategically whether they read using a rigorous ‘‘proofreading’’ mode or a more liberal mode of reading for meaning. Presumably, better spellers, with more elaborate and entrenched ‘‘optional’’ spelling information, would be able to switch modes more easily, as they would find the detailed letter-by-letter analysis required by proofreading easier to perform. An objection might be raised about investigating the relation between reading and spelling in a situation in which students could see what they were writing when they produced their spellings. Monsell’s (1985, 1987) solution to this problem was to have people produce spellings with a screen obscuring what they were writing. However, spelling long words such as those used in the present experiments without visual feedback might be expected to involve burdens on temporary memory processes which are not present in normal spelling. A similar memory component is introduced in oral spelling. Jorm and Schoknecht (1981) showed that oral spelling was inferior to written spelling at least partly because spellers had no external record to remind them of what they had already written and where they were up to in the spelling. Since we were interested in obtaining a reflection of student’s knowledge of spellings, we followed Campbell’s (1987) and Funnell’s (1992) procedure of obtaining spellings in the normal way. When mature learners produce spellings under optimal and careful conditions, reading and spelling processes are occurring simultaneously. Treiman (1993) has noted that young children rarely verify their spellings by determining whether what they have produced matches what they meant to write, whereas for adults this process typically be-

286


comes a fundamental part of the spelling process. Adult spellers monitor their spellings visually because they are aware that they sometimes make slips in translating what they know into an accurate output. Problems occurring at this peripheral level are said to involve disruptions to a ‘‘graphemic buffer’’ in which information about the identity and order of subsequent letter shapes is held while current letters are being produced (e.g., Ellis, 1982; Caramazza, Miceli, Villa, & Romani, 1987). Students of the type we studied in the present experiments typically have a good command of common phoneme– grapheme correspondences (Holmes & Ng, 1993). Nevertheless, our participants still sometimes produced phonologically implausible misspellings. The mistakes in these misspellings were much more likely to involve misorderings of letters, or deletions, substitutions, and insertions of letters that had already been produced or were about to be produced, than entirely different letters, suggesting that the mistakes usually were due to problems of supervising the production of elements from a graphemic buffer. Even some of the mistakes involving different letters, for example, omission of letters, can also be interpreted in this way. When students were presented with these misspellings in typed format, they were able to reject the vast majority of them as wrong, further indicating that the mistakes were indeed usually performance errors rather than competence errors. The participants of the present study were highly educated students who were not poor readers or spellers as conventionally defined. They were at the lower end of the spelling distribution compared with their peers, otherwise they would not have provided enough misspellings for the ‘‘can’t spell’’ conditions in the recognition experiments. However, they were above average compared with the general population of the same age, according to the standard reading and spelling tests. Would less advanced readers be more likely to show evidence for a dissociation between reading and spelling? The relation between reading and spelling changes over the course of literacy development

(Ehri, 1986; Frith, 1985; Goswami, 1992). Very young children may even sometimes be able to spell simple words that they can’t read (Bryant & Bradley, 1980). This could happen if the child were able to generate a spelling from their knowledge of phoneme– grapheme correspondences, rather than accessing any information from a representation established in memory. However, as children progress in both their reading experience and their learning of conversions between phonemes and graphemes, they begin to store spellings of particular words in memory. In what is often termed the orthographic stage, children continue to accumulate specific word spellings as well as to develop general knowledge about properties of the orthographic system (Frith, 1985; Marsh, Friedman, Welch, & Desborg, 1981). We believe that the situation of a child at this stage may not be very different in principle from that of the young adults we have investigated. After coming across a new printed word on enough occasions, the reader would develop a memory record corresponding to it, and the record would continue to be augmented either through further passive reading encounters or through deliberate analytic strategies for committing the word’s spelling to memory. Nevertheless, it would be interesting to apply the procedures we have used here to groups of younger normal readers in the attempt to detect any greater superiority of reading over spelling representations in less advanced individuals. How can the existence of ‘‘unexpectedly poor spellers’’ be accounted for on this view? These people are classified as such by comparison with individuals who read at the same level but spell much better. When reading level has been defined in terms of a text comprehension measure, poor spellers who are ‘‘good readers’’ are actually slower and less accurate word recognizers than good spellers who are good readers, both for children (Bruck & Waters, 1990a, 1990b) and adults (Joshi & Aaron, 1991). Standard word reading tests have also been the gauge of reading ability (Frith, 1980; Perin, 1983). However, because reading of individual words is not timed, these tests are relatively

287


insensitive to word recognition efficiency. Burden (1992) classified adult poor spellers as good readers on the basis of their above average performance on a standard word reading test, and then probed their reading performance more intensively, comparing them with good spellers who were also good readers. While poor spellers were as accurate as good spellers in reading aloud highfrequency words, they were less accurate in reading aloud low-frequency words. They were also significantly worse at reading and spelling nonwords, and were substantially slower at reading connected text. In short, when the word recognition processes of unexpectedly poor spellers are examined systematically, they turn out to be inferior to those of good spellers who are good readers. The unexpectedly poor speller should perhaps be regarded as someone who is unexpectedly good at disguising their relative lack of orthographic expertise. Finally, we recognize that a version of the dual-representation position could explain our results by discarding the assumption that the reading lexicon contains more precise information than the spelling lexicon. A proponent of this account could maintain a dichotomy between reading and spelling representations, but claim that mature readers have acquired and stored exactly the same information in each representation as a consequence of their extensive reading and spelling experience. Of couse, this is a position that would be difficult to confirm or refute with empirical evidence. In the absence of any unambiguous external grounds providing a substantive characterization of dual but identical orthographic representations in mature readers, the only logical—and certainly the most parsimonious— course of action seems to be to accept the unified representation view. Clearly, reading and spelling will differ in how they access and make use of information contained in the common orthographic representation, and it is these distinct processes that continue to be worthy of investigation.

APPENDIX: WORD LIST Words Used in Experiments 1, 2, and 3 accommodate annihilate asymmetric bourgeois bureaucracy cemetery chandelier chronicle

commitment consensus correspondence dissimilar effeminate embarrassed fluorescent harassing

jeopardy marijuana misshapen misspell occurrence omission plagiarism prevalent

privilege rendezvous separate silhouette superseded zucchini

Additional Words Used in Experiments 1 and 2 attorney charlatan commemorate connoisseur consummate courtesy desperate

diabetes diaphragm dissolved exchequer haemorrhage immaculate indefinite

ingratiating inoculation kaleidoscopic loquacious manoeuvre nauseous necessity

parallel primitive rarefied recurring sovereign

Additional Words Used in Experiment 1, but Not in Experiment 2 accumulate

dialogue

proceed

REFERENCES Barry, C. (1992). Interactions between lexical and assembled spelling (in English, Italian and Welsh). In C. M. Sterling & C. Robson (Eds.), Psychology, spelling and education (pp. 71– 86). Clevedon, UK: Multilingual Matters. Barry, C. (1994). Spelling routes (or roots or rutes). In G. D. A. Brown, & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 27– 49). Chichester: Wiley. Behrmann, M., & Bub, D. (1992). Surface dyslexia and dysgraphia: Dual routes, single lexicon. Cognitive Neuropsychology, 9, 209 –251. Bradley, L., & Bryant, P. E. (1983). Categorising sounds and learning to read: A causal connection. Nature, 301, 419 – 421. Brown, G. D. A., & Ellis, N. C. (1994). Issues in spelling research: An overview. In G. D. A. Brown, & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 3–25). Chichester: Wiley. Bruck, M., & Waters, G. (1990a). An analysis of the component spelling and reading skills of good readers— good spellers, good readers—poor spellers, and poor readers—poor spellers. In T. H. Carr & B. A. Levy (Eds.), Reading and its development: Component skills approaches (pp. 161–206). San Diego: Academic Press.

288


Bruck, M., & Waters, G. (1990b). Effects of reading skill on component spelling skills. Applied Psycholinguistics, 11, 425– 437. Bryant, P. E., & Bradley, L. (1980). Why children sometimes write words which they do not read. In U. Frith (Ed.), Cognitive processes in spelling (pp. 355–372). London: Academic Press. Byrne, B., & Fielding-Barnsley, R. (1989). Phonemic awareness and letter knowledge in the child’s acquisiition of the alphabetic principle. Journal of Educational Psychology, 81, 313–321. Burden, V. (1992). Why are some ‘normal’ readers such poor spellers? In C. M. Sterling & C. Robson (Eds.), Psychology, spelling and education (pp. 200 –214). Clevedon, UK: Multilingual Matters. Campbell, R. (1987). One or two lexicons for reading and writing words: Can misspellings shed any light? Cognitive Neuropsychology, 4, 487– 499. Caramazza, A., Miceli, G., Villa, G., & Romani, C. (1987). The role of the graphemic buffer in spelling: Evidence from a case of acquired dysgraphia. Cognition, 26, 59 – 85. Coltheart, M., & Funnell, E. (1987). Reading and writing: One lexicon or two? In A. Allport, D. G. Mackay, W. Prinz, & E. Scheerer (Eds.), Language perception and production: Relationships between listening, speaking, reading and writing (pp. 313–339). London: Academic Press. Ehri, L. C. (1980). The development of orthographic images. In U. Frith (Ed.), Cognitive processes in spelling (pp. 311–338). London: Academic Press. Ehri, L. C. (1986). Sources of difficulty in learning to spell and read. Advances in Developmental and Behavioral Pediatrics, 7, 121–195. Ehri, L. C. (1991). Development of the ability to read words. In R. Barr, M. L. Kamil, P. Mosenthal, & P. D. Pearson (Eds.), Handbook of reading research, Vol. 2 (pp. 385– 419). New York: Longman. Ellis, A. W. (1982). Spelling and writing (and reading and speaking). In A. W. Ellis (Ed.), Normality and pathology in cognitive functions (pp. 113–146). London: Academic Press. Ellis, A. W. (1993). Reading, writing and dyslexia: A cognitive analysis. Hove, UK: Erlbaum. Fischer, F. W., Shankweiler, S., & Liberman, I. Y. (1985). Spelling proficiency and sensitivity to word structure. Journal of Memory and Language, 24, 423– 441. Friedman, R. B., & Hadley, J.A. (1992). Letter-by-letter surface alexia. Cognitive Neuropsychology, 9, 185– 208. Frith, U. (1980). Unexpected spelling problems. In U. Frith (Ed.), Cognitive processes in spelling (pp. 495–515). London: Academic Press. Frith, U. (1984). Specific spelling problems. In R. N. Malatesha & H. A. Whitaker (Eds.), Dyslexia: A global issue (pp. 83–103). The Hague: Martinus Nijhoff. Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. Patterson, J. Marshall, & M. Coltheart

(Eds.), Surface dyslexia: Neuropsychological and cognitive studies of phonological reading (pp. 301–330). London: Erlbaum. Funnell, E. (1992). On recognising misspelled words. In C. M. Sterling & C. Robson (Eds.), Psychology, spelling and education (pp. 87–99). Clevedon, UK: Multilingual Matters. Goswami, U. (1992). Annotation: Phonological factors in spelling development. Journal of Child Psychology & Psychiatry, 6, 967–975. Hanley, J. R., & Kay, J. (1992). Does letter-by-letter reading involve the spelling system? Neuropsychologia, 30, 237–256. Holmes, V. M., & Ng, E. C. (1993). Word-specific knowledge, word-recognition strategies, and spelling ability. Journal of Memory and Language, 32, 230 –257. Houghton, G., Glasspool, D. W., & Shallice, T. (1994). Spelling and serial recall: Insights from a competitive queueing model. In G. D. A. Brown, & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 365– 404). Chichester: Wiley. Invernezzi, M. A. (1992). The vowel and what follows: A phonological frame of orthographic analysis. In S. Templeton & D.R. Bear (Eds.), Development of orthographic knowledge and the foundations of literacy: A memorial festschrift for Edmund H. Henderson (pp. 105–136). Hillsdale, NJ: Erlbaum. James, C. T. (1975). The role of semantic information in lexical decisions. Journal of Experimental Psychology: Human Perception and Performance, 1, 130 –136. Jastak, S., & Wilkinson, G. S. (1984). Wide Range Achievement Test—Revised. Wilmington, Delaware: Jastak Associates Inc. Jorm, A. F., & Schoknecht, C. (1981). The role of visual word-recognition checks in children’s spelling. Australian Journal of Psychology, 33, 393– 403. Joshi, R. M., & Aaron, P. G. (1991). Developmental reading and spelling disabilities: Are these dissociable? In R. M. Joshi (Ed.), Written language disorders (pp. 1–23). Dordrecht: Kluwer. Kucˇera, H., & Francis, W.N. (1967). Computational analysis of present-day American English. Providence, RI: Brown Univ. Press. Marsh, G., Friedman, M. P., Welch, V., & Desborg, P. (1981). A cognitive-developmental theory of reading acquisition. In T. G. Waller & G. E. MacKinnon (Eds.), Reading research: Advances in theory and practice, Vol 3 (pp. 199 –221). New York: Academic Press. Monsell, S. (1985). Repetition and the lexicon. In A. W. Ellis (Ed.), Progress in the psychology of language, Vol. 2 (pp. 147–195). London: Erlbaum. Monsell, S. (1987). Nonvisual orthographic processing and the orthographic input lexicon. In M. Coltheart (Ed.), Attention and Performance XII: Reading (pp. 299 – 323). Hillsdale, NJ: Erlbaum. Morton, J. (1980). The logogen model and orthographic

READING AND SPELLING IN SKILLED ADULT READERS structure. In U. Frith (Ed.), Cognitive processes in spelling (pp. 117–133). London: Academic Press. Patterson, K. (1986). Lexical but nonsemantic spelling? Cognitive Neuropsychology, 3, 341–367. Perin, D. (1983). Phonemic segmentation in spelling. British Journal of Psychology, 74, 129 –144. Seymour, P. H. K. (1992). Cognitive theories of spelling and implications for education. In C. M. Sterling & C. Robson (Eds.), Psychology, spelling and education (pp. 50 –70). Clevedon, UK: Multilingual Matters. Shulman, H. G., & Davison, T. C. B. (1977). Control properties of semantic coding in a lexical decision task. Journal of Verbal Learning and Verbal Behavior, 16, 91–98. Stone, G. O., & Van Orden, G. (1993). Strategic control of processing in word recognition. Journal of Experimental Psychology: Human Perception and Performance, 19, 744 –774.

289

Treiman, R. (1993). Beginning to spell: A study of firstgrade children (pp. 277–293). New York: Oxford Univ. Press. Snowling, M. (1987). Dyslexia: A cognitive developmental perspective. Oxford: Blackwell. Venezky, R. L. (1970). The structure of English orthography. The Hague: Mouton. Weekes, B., & Coltheart, M. (1996). Surface dyslexia and surface dysgraphia: Treatment studies and their theoretical implications. Cognitive Neuropsychology, 13, 277–315. Wing, A. M., & Baddeley, A. D. (1980). Spelling errors in handwriting: A corpus and a distributional analysis. In U. Frith (Ed.), Cognitive processes in spelling (pp. 251–285). London: Academic Press. (Received December 22, 1997) (Revision received April 16, 1998)

The Relation between Reading and Spelling in ... - Science Direct

The Relation between Reading and Spelling in ... - Science Direct

Suggest Documents

The Relation between Foreign Direct Investments ... - Science Direct

The Relation between Demographic Factors and the ... - Science Direct

The relation between mobbing behaviors teachers in ... - Science Direct

Perspectives on the relation between information and ... - Science Direct

Non-Fickian diffusion affects the relation between the ... - Science Direct

Relation between Power and Linear Model of ... - Science Direct

Relation between Increasing Spinal Curve and Anxiety - Science Direct

Exploratory Study on the Relation between Urban ... - Science Direct

The relation between DIBELS, reading comprehension - NCSIP

DIRECT MEASUREMENT OF THE RELATION BETWEEN ... - CiteSeerX

Relation Between Behavior Pattern, Job Satisfaction ... - Science Direct

Education, Cultural and Intercultural Relation - Science Direct

The relation between reading ability and morphological skills ...

Literacy predictors and early reading and spelling

The relation between foreign direct investments and some economic ...

The direct relation between safe hemorrhoidectomy and anal ... - OAText

Review of the Critical Reading Education in the ... - Science Direct

Cognitive mechanisms underlying reading and spelling development ...

Performance of The Slovak Economy in Relation to ... - Science Direct

Math Reading Phonics Spelling Grammar Science/Social Other ...

Math Reading Phonics Spelling Grammar Science/Social Other ...

Are Reading and Spelling Phonologically Mediated? - CiteSeerX

Cognitive mechanisms underlying reading and spelling development ...

Decision - Making, and Their Relation to The ... - Science Direct