The use of handwritten misspellings to examine reading strategies can be justified for two reasons. First, they represent an enduring record of visible language which is relatively convenient and inexpensive both to collect and to analyse, and which involves none of the high technology equipment, such as the tachyscope, which the collection and analysis of direct data on reading might require.

Second, it is not intuitively unreasonable to suppose that there is some relationship between the two sides of visible language, reading and writing. According to Frith (1980a), “It is well established that there is a positive correlation between reading and spelling performance” (p. 498), and Bruck and Waters (1988) refer to “the high correlations traditionally found between reading and spelling abilities” (p. 78). Frith cites a review of large-scale studies of reading and spelling by Malmquist which found correlations from 0.50 to 0.80. The extent of this range can be accounted for in two ways: the relationship between spelling and reading may vary from individual to individual, and different studies may have involved dissimilar test instruments.

Therefore, the present study examines two types of data: data on student achievement in English from several types of test, and a corpus of misspellings. The collection of this data is first described in order to show how and to what extent it may reflect reading proficiency. Then, it is analysed quantitatively to establish the extent to which it correlates with this proficiency. Finally, the misspellings themselves are analysed in an attempt to reveal those features of writing which may reflect reading strategies.

Of course, the analysis of misspellings is not completely unmotivated. I have argued elsewhere (Pond 1990) that the transposition in writing of pairs of Arabic digits (the digits used within Roman script) by Arabic speakers, for example 49 for 94, is a representation of the verbalisation of those numbers in Arabic. However, when an advanced Arabic student, asked to find the title of a novel by Hemingway in a library card catalogue, writes The OdI Man and the Sea (personal communication), this striking, though not untypical, transposition error can hardly be phonologically mediated: it must reflect visual processing. Inasmuch as such visual processing errors may reflect Arabic orthographic right-left orientation, they may be evidence of a reading problem at the word level. Therefore, the analysis of misspellings is approached in terms of phonologically or visually mediated errors.

This analysis of misspellings must be informed by theories of word recognition within the context of English and Arabic orthographies. Central to theories of word recognition is the role of visual and phonological mediation in the translation of visible language to meaning (Barron 1986). This has led to the most widespread, the dual-route, theory of word processing (e.g. Downing & Leong 1982), according to which the meaning of a word in the mental lexicon is accessed either directly or indirectly.

Direct access is a purely visual process which makes no use of any visible clues to the phonological actualisation of the word. In an extreme example from environmental visible language, when I see the international road sign “STOP”, I understand its meaning principally from the shape and colour of the sign rather than the word ‘STOP’. If I were to see the same sign in Saudi Arabia, the same shape and colour would tell me to stop, even though the Arabic word would have neither meaning nor, as far as I am concerned, possible phonological actualisation. However, if, after seeing the sign many times, I were to see the Arabic word out of its road sign context, I may well understand the implication of the word purely from its configuration as stored in my visual memory. Consider then a less extreme example, the homograph ‘lead’. If I read a sentence beginning “The metallic element lead ...”, then my schema may disambiguate the homograph without implicit or explicit reference to the correct phonological actualisation. Such direct access to the mental lexicon is potentially available for all words.

The indirect access route, on the other hand, interposes a phonological step in the path from visible language to the mental lexicon. Graphemes are converted implicitly or explicitly by the application of a set of rules into phonemes to produce a pronunciation, which in turn is used to access the word from the mental lexicon. Such access, too, is potentially available for all words. Consider the phrase “the lead battery”. It is probable that you implicitly pronounced the homograph as /led/, but only after recognising the word from the collocational frequency of ‘lead (acid) battery’; access to meaning was by the direct route without pre-lexical phonological mediation. On the other hand, indirect access processing would have offered the prelexical pair /li:d/ and /led/. In this case, a possible context for /li:d bateri/ would have to be considered (perhaps, “air cover is important for the lead battery, which will suffer higher casualties than the rear batteries in battle”).

The classical dual-route theory of word processing implies that one route is followed to the exclusion of the other. Theoretical implausibilities have led to modifications of the theory, the most popular of which is the ‘horse-race’ model, in which both routes are followed in parallel until one is successful (Henderson 1982). Tentative alternative models have also been suggested, but have not yet gained any measure of acceptance, and “for the time being, the situation is certainly fluid” (Bertelson (ed) 1987).

Research in the field of word recognition has tended to be Anglo-centric, and has made much use of native-speaker pathology. However, there is now growing interest in other languages and in non-Roman orthographies. This is of present interest because of the different English and Arabic orthographic systems, which may affect the use of visual or phonologically mediated word processing. Orthographies are commonly placed on a cline from shallow, or transparent, to deep, or opaque (e.g. Downing & Leong 1982). An orthography is considered shallow to the extent that it unambiguously and systematically represents the sound of words. Relatively shallow orthographies are found in European languages such as Spanish. Chinese serves as the usual example of a deep orthography, where the characters are commonly thought to have little relationship to the sounds of the words they represent.

The availability of phonological information in an orthography appears to influence the use of direct or indirect lexical access. For example, Turvey et al (1984) argue that the phonetically very shallow Serbo-Croatian orthographies (Roman and Cyrillic alphabets) neither encourage nor allow direct lexical access and
 

it is very different from the phonologically deep English orthography which encourages (and, perhaps, demands) flexibility. For the beginning reader and for the fluent reader of Serbo-Croatian there are few enticements to try any strategy other than one that is phonologically analytic. Such a strategy is efficient, economical, and most befitting the Serbo-Croatian orthography. (pp. 88-89)

As indicated, English orthography is phonologically deep, that is to say deeper than other alphabetic orthographies, and this appears to have been caused principally by the vagaries of attempts to standardise it for printing (Scragg 1974). As an example of the depth of English orthography, the derivational morphemes for plurals of regular nouns is ‘-s’, whereas the corresponding phonemes are /s/ or /z/. This orthographic convention encourages and repays visual rather than phonological processing. However, it is implausible that English readers consider all words as logographs and disregard the canonical phonological value of letters of the alphabet. Therefore, the orthography of English predisposes a reader to a mixture of direct and indirect access, and Chomsky and Halle (1968) claim that “conventional orthography is ... a near optimal system for the representation of English words” (p. 49). The proportions of direct and indirect access employed will depend on individual differences, but perhaps more importantly on the way reading has been learned. The great divide in the teaching of native English speaker reading corresponds to the great divide in the dual-route theory of word processing. The short-lived experiment with the initial teaching alphabet was an attempt to delay full use of visual processing, and today individual differences in use of direct and indirect access may reflect early learning experiences. The roles of orthography and learning are summed up by Henderson (1982):
 

It is unclear, therefore, to what extent differences in the reader’s treatment of scripts is due to the formal properties of the script as a system of representation or due to the way in which competence with a script is acquired and supported. (p. 209)

Despite complaints about the difficulty and inconsistency of English spelling, a modern orthography has emerged which fluent readers can access efficiently with the flexible use of visual and phonological clues. Individual readers rely to differing extents on direct and indirect access, and Baron et al (1980) describe them as Chinese or Phoenicians. Readers who favour direct access are called Chinese since the predominant route available to logographic script readers is presumed to be visual. The Phoenicians are named in honour of the inventors of the first script designed to represent the sound of the word, in this case by syllables, and this led to the development of alphabets. Therefore, it is necessary to consider what aspects of Arabic orthography, and the way competence in it has been acquired, might predispose the present students to access English words either directly or indirectly, visually or phonologically, to what extent, in other words, they are Chinese or Phoenicians.

Arabic orthography can be considered phonetically shallow in one important respect. The literal grapheme to phoneme correspondence is almost completely unambiguous (Azzam 1989) and in this respect Arabic orthography is shallower than Latin, Italian, Spanish or “the much-touted ancient Greek” (Justice 1987, p. 19). This feature provides a rich fund of information which can be exploited by indirect, phonologically mediated access. Moreover, the history and present status of written Arabic reinforce its role as a representation of spoken Arabic.

Modern written Arabic is a direct descendant of the Koran, whose divine words permeate and inspire the whole Arab world to this day. Indeed, the Koran is the word of God in a very literal and direct sense, more so than, say, the various translations and retranslations of the Bible. The Koran is a written record of the spoken word of God as revealed to the illiterate Prophet Mohammed, who transmitted it orally to his companions and through them to his followers. The exact divine phonological actualisation was preserved and transmitted orally. The actual writing of the Koran followed later when the number of followers and the danger of incorrect transmission increased. The Koran is therefore an attempted phonetic transcription of the word of Allah - of Allah, not God, because this was the word He pronounced. To this day, education in the Arab world begins, and until recently ended, with learning correct recitation of the Koran. Therefore, Arabic speakers are massively encouraged to consider written Arabic as a visual record of oral language.

However, the original Koran was written in an orthography which, in common with Hebrew and other Semitic languages, indicated consonantal phonemes only. Such orthographies have been called ‘disemvowelled’, and Henderson (1982) accounts for this lack of vowel phoneme representation, and indeed relative paucity of vowel phonemes in the spoken language, by suggesting that “The reason for the indifference to vowels in the Semitic languages is, I suspect, that they play a less critical role than in lndo-European languages” (p.21).

For some years, the lack of vowel representation in the Koran did not endanger accurate phonetic transmission since the unrepresented vowels were easily and unambiguously inferred. However, in later and subsequent copies of the Koran, diacritical marks were introduced to indicate with ever increasing delicacy not only the missing vowel phonemes but other phonological features as well; such texts are called ‘pointed’. The diacritical marks in the divine text for the three long, but not the three short, vowel phonemes were gradually introduced into profane writing first as diacritical marks and then as letters to accompany the consonants. It should be noted that primary stress in a word is not assigned automatically to long vowels, and that unsignalled ambiguous short vowels may produce homographs.

Today, normal modern written and printed Arabic includes long vowels but is unpointed. Pointed texts are generally restricted to the Koran and books for children, as well as monumental inscriptions and parts of text demanding a certain amount of gravitas. The general use of unpointed orthography causes difficulty for non-Arabic speakers; according to the International Organization for Standardization (ISO 1 961), “Transliteration therefore requires a good knowledge of Arabic, and it may be necessary to use dictionaries for an indication of vowelling” (p. 3). Therefore, to the extent that Arabic orthography does not represent full vocalic information, native speaker readers may be encouraged to process words visually. Thus, the history and status of the Koran has led to an orthography which encourages both direct and indirect access.

The sanctity of the Koran, and the consequent sanctity of its orthography, militated against creative, profane uses of the script. However, human creativity, aided no doubt by the plethora of diacritical marks, could be channelled into ever more fanciful configurations of the words of Allah, and principally the word Allah itself. There was thus born the long and venerable tradition of Arabic calligraphy, whose influence can still be seen in the modern orthography. Whereas English orthography differentiates between handwritten, cursive script and printing, Arabic print is essentially a mechanical replication of the written hand. This is illustrated by Al-Asadi and Rueda (1984), who state that
 

When the print process appeared the typography inherited some features from the calligraphy. This has resulted in one of the most distinctive characteristics of the Arabic script: its cursiveness. Also, elongation strokes are used to justify lines to give a more artistic appearance than with the simple space. (pp. 5-6)

In their design of an Arabic font for computers, Al-Asadi and Rueda specify parameters which are common to both handwritten and printed orthography. To begin with, each letter requires four distinct shapes, for word initial, final and medial, and for isolated citation form, the latter being used almost exclusively in alphabetic enumeration. The three forms occurring in words require shapes which allow obligatory tautolexical juncture with adjacent letters, with the exception of the six letters which do not allow juncture to the left.

Therefore, Arabic print, like handwriting, may aim to appeal to the eye as much as to convey sound or meaning. Given the individual initial and final forms of letters which signal the word unit much more graphically than the meagre spacing, and the tendency for the letters in the word to form a continuous, elaborate line, it would not be surprising to find words approached logographically, as a visually appealing picture representing a word. studies of how Arabic and other speakers scan Roman letters (Green & Meara 1987; Randall & Meara 1988) showed that Arabic and Chinese speakers exhibited the same scanning pattern, and that a different pattern was evident in English and Spanish speakers, who both have a Roman alphabet orthography. Moreover, the scan pattern used by the Arabs and the Chinese was the same pattern used by the English and Spanish to scan pictures.

To sum up, there is reason to believe that Arabic speakers process words in Arabic both directly and indirectly, and that this strategy is transferred to the processing of English, which also encourages use of both routes. The analysis of misspellings will consider the relative frequency of visually and phonologically mediated word processing which was only partially successful and which produced the misspellings.

One important feature of Arabic orthography which has not so far been discussed is its right-left orientation. Writing must be read and words must be written from right to left. This could be expected to interfere in the processing of English with its left-right orientation. In a study of six Arabic speakers, ElHassan (1984) found evidence of a general right-left orientation which “extends beyond graphology to the layout of non-verbal information” (p. 83). It appears that orientation extends considerably further. Henderson (1984) cites work on the click location task which suggests that there is a
 

“tendency to listen, as it were, from left to right [which] appears to be related to reading processes since bilingual subjects showed the left-right effect when listening to French, but those same subjects showed a reversed effect when listening to Hebrew, which is read from right to left.” (p. 22)

In a small-scale study of 467 students, of which the sample for the present study was a part, I found evidence of misorientation. A multiple-choice question referred to the contents page of a periodical reprinted in the textbook, and asked what the article on page 12 was about. Finding the correct answer involved no more than referring to the well-studied page, identifying the number, and referring to the article title to the left of the number. Nevertheless 15.7% of the sample evidently referred to the article title to the right of the number in the second column of the contents page. This was the first, and the easiest of ten questions, yet, as can be seen in the distribution of students choosing this incorrect answer (Appendix B6), it was by no means only the least successful students who were misoriented.

The effect of misorientation in the present sample is clearly significant. The particular case and cause of Arabic misorientation is highlighted by a study of Israeli Arabic and Hebrew speaking school and college students carried out by Kugelmass and Lieblich (1979), in which they found that the right-left tendency was much more strongly ingrained in Arabic speaking than Hebrew speaking subjects. This difference is attributed to proposed, but unspecified, differences in acquisition of reading skills in Hebrew and Arabic. The most obvious difference between Arabic and Hebrew orthography is that, while they are both disemvowelled consonantal representations of linguistic cousins, they are visually so dissimilar. The regular sequence of separate box-like letters in Hebrew is in striking contrast to the elegant flow of the Arabic pen. This is surely even further inducement for Arabic speakers to access words directly and visually rather than indirectly and phonologically. Therefore, in the analysis of misspellings, attention will be paid to visually inspired misspellings which might be the result of right-left orientation interference.

To sum up, the rest of this study consists of three parts. First, the data and its collection will be described. Then, the data will be analysed quantitatively to establish the correlation between reading and spelling proficiency. Finally, the misspellings will be analysed in relation to evidence of visual and phonological mediation.

 
NEXT