uropyrhnlr~gla. Vnl . 70, No t pp 2 7-214 1742 1'rintui in Glcoi &naln 0728 t972-92S5011+0 .W r . 1992 Rrgamuv Pies plc DOES LETTER-BY-LETTER READING INVOLVE THE SPELLING SYSTEM? J . RICHARD HANLFYi and JANICE KAY ; tDeparmient of Psychology, rniversity of Liverpool, Eleanor Rathbone Building, PO Box 147, Liverpool . [69 3BX ; and ;Department of Psychology . University of Exeter, Washington Singer Laboratories, Parry Road, Exeter, EX4 4Q6, U .K . (Received 6 March 1991 : accepted 3 October 1991( Abstract-WAkaiNCmN and SHALLICS (Brain 103, 99-112, 1980) proposed that letter-by-letter cadets can no longer access the visual word-form system, and read by using the Spclhng system 'in reverse' . Contrary to their suggestion, this paper presents the case of a letter-by-letter reader who appears to use quite different strategies in word reading and spelling . Reading and spelling are both impaired, but they appear to be impaired in different ways . Most strikingly, the patient's spelling errors consist largely of phonologiwl regularizations- indeed there is little evidence that he has any lexical spelling knowledge available-whereas there are virtually no regularizations amongst his reading errors . These arc largely visual paralcxias and letter misidentifications . It is suggested that this pattern is more easily explained in terms of compensatory tnechanisms that access the reading lexicon than by use of the spelling system in reverse . INTR ODT JCTTON Lhr[ER-BY-LETTER reading is a form of acquired dyslexia in which patients appear to read single words by identifying one letter at a time . Consequently, reading is extremely laboured, and reading time increases in line with the number of letters in a word . For example, patients have been reported who take over 16 sec on average to read three- and four-letter words, and almost 50 sec to read seven- and eight-letter words [24] . 'There are at least two important questions that have been asked about letter-by-letter reading . The first concerns the nature of the reading impairment : at what stage has the normal reading process broken down? Several recent papers provide interesting discussions of this issue [4, 15, 18, 28, 30. 33] . The second question concerns the compensatory strategies used by letter-by-letter readers that produce such a distinctive pattern of performance . Although these two questions are obviously related . It is the nature of the compensatory strategies used in letter-by-letter reading that provides the main focus for the present investigation . First of all, it is important to point out that the compensatory strategies employed in letter-by-letter reading involve the use of letter names rather than letter sounds . In a shallow' orthography, there is no need to distinguish letter names from letter sounds . In a `deep' orthography such as English, however, a particular phoneme may be associated with a aunlber of different letters in the context of different words . Consequently, since we generally cannot make unique reference to a particular letter by a sound, each letter is given its own name . There are, of course, acquired dyslexics whose compensatory strategies involve the use of letter sounds . Such patients, who are generally referred to as 'sound-bysound' readers or `surface dyslexics', attempt to read words on the basis of sublexical 237 238 J. R . HANLTY and J . KAY phonology [21] . Letter-by-letter readers, in contrast, employ compensatory strategies that resemble those which a normal reader might use when attempting to recognize a word that is spelled orally . According to WARRINGTON and SHALLICF [35], patients employ a letter-by-letter strategy because they arc unable to use the visual word form system during reading . In order to compensate, they employ a radically different set of procedures from those normally used in visual word recognition . Once a representation of the names of all the letters that comprise the word has been built up, reacting proceeds via the spelling system . Exactly how this might take place has never been fully specified, although the most explicit account can be found in SHAta .ICE and McCARTtty [31] . Normally, of course, it is the job of the spelling system to produce a sequence of letters as output . However, these authors suggest that the spelling process can also work `in reverse, and can accept sequences of letter names as input . Although not explicitly stated, it appears that the authors assume that the link between the letter names and a word's representation in the spelling system is bidirectional . Then ; once the appropriate entry in the spelling lexicon is located, it can be used to address an entry in the phonological output lexicon if a spoken response is required . 7n contrast, PATTF.RSON and KAY [24] argue that letter-by-letter readers are capable of using the visual word-form system . What they have lost is the ability to access this system in the normal way . Skilled readers can pick up letter information across several letter positions simultaneously . Letter-by-letter readers, however, are obliged to access the word-form system via serial rather than parallel letter identification . This process, like the reverse use of the spelling system in Warrington and Shallice's account, is seen as being a time-consuming operation which is directly related to the number of letters in u word . The ability to recognize a word from its letter names is a skill which normal readers can apply without undue ydifficulty . Consequently, the compensatory strategics that letter-byletter readers employ do, almost certainly, reflect a normal meta-linguistic function . Nevertheless, PATTERSON and KAY [24] do not provide any a priori justification for the claim that the word form system can be accessed via serial letter recognition, and WARRINGTON and SH .ALL,WE [35] provide no a priori justification for the claim that the spelling system can operate in reverse . Perhaps the main reason for the popularity of Warrington and Shallie&s account are the parallels between the reading performance and the spelling performance of individual letter-by-letter readers . If these patients are indeed using reverse spelling to enable them to read, then one would expect that the way in which they read words should reflect the way in which they spell words . lip until recently . this has always proved to he the case [24, 26] . Patterson and Kay, for instance, showed that letter-by-letter readers could be separated into Type I and Type 2 readers . Type I letter-by-letter readers such as patients M .W . and CH . were good spellers, but experienced severe letter recognition difficulties . Since they are good at spelling, it follows from Warrington and Shalliee's account that they should be able to read words accurately whenever they can correctly identify the letters . [his turned out to he the ease . Type 2letter-by-letter readers such as T .P_ and K .C, by contrast, are generally better at letter recognition than Type 1 patients, but often read words incorrectly even when they have identified the letters appropriately . Consistent with Warrington and Shallice's view, T .P . and K .C . were also poor spellers who produced the same types of errors in reading that they made in spelling . Their spelling errors generally preserved the phonological features of the words spoken form, and are thus characteristic of surface dysgraphia [2] . For LETTER-nY-LETTER RrADIFC 239 example, T .P ., whose spelling performance is described in detail by HATFTET .D and PATTERSON [14] spelt 'nephew' as NEEFUE, and `biscuit' as BISKET . Moreover, these two patients also made phonological errors in reading . T .P ., for instance, read `gone' as JOAN, and 'city' as KITTY . Reading errors such as these are, of course, characteristic of surface dyslexia [2]] . In order to explain the pattern of impairment observed in T .P . and K .C ., Patterson and Kay were forced to argue that both Type 2 patients were suffering from two distinct deficits, me of which led to phonological errors in reading, and one of which led to phonological errors in spelling . They acknowledged (p . 434) that WARRINGTON and SIIALI.TCE [35] could cffer a much more parsimonious explanation of this pattern of performance . According to the reverse spelling hypothesis, a letter-by-letter reader who produces phonological errors in spelling would be expected to show a similar pattern of impairment in reading . As Shallice and McCarthy state : TP and KC spelled with an over-reliance on the typical sound-to-spelling correspondence of English .11 this spelling process were "reversed" then one would indeed expect that patients with the appropriate double deficits would try out pronounciations rather than words per se. [30, p .367] What Warrington and Shalliee would find difficult to explain, therefore, would be a letter-by-letter reader whose impairment in spelling was not reflected in his or her reading performance. On the other hand, the existence of such a patient is exactly what would be p'cdicted from PATTERSON and KAY [24] . Recently, RAPCSAK et al . [27] have described a Type 1 letter-by-letter, P .T ., whose spelling is severely impaired, and who makes phonologically appropriate spelling errors . When she makes errors in reading, however, they come about as a result of letter misidentifications and are visual errors (e .g . 'aisle' > `assail') rather than phonological regularizations . This is not consistent with the reverse spelling hypothesis [3l, 35] ; if P .T . makes phonological errors in spelling then she should also make phonological errors in reading . Furthermore, the reverse spelling account also has problems in explaining why, when she is asked to identify orally spelled words, P .T . makes hardly any errors even though her oral spelling is as severely impaired as her written spelling . Rapseak et al . conclude that Patterson and Kay's account of letter-by-letter reading provides a much better explanation of PT's performance than the reverse spelling hypothesis . Nevertheless, P .T . was a Type 1 letter-by-letter reader whose word recognition problems reflected problems in visual letter recognition . It therefore remains possible that the reading of Type 2letteo-by-letter readers, who find it difficult to recognize words even when they have successfully identified all the letters, may be dependent upon the spelling system . In order to investigate this issue, we report below a detailed examination of the performance of a type 2 letter-by-letter reader whose reading and spelling also appear to be impaired in very different ways . CASE REPORT P .D . is 32-years-old, right-handed, and lives in the north-east of England . In November 1982, he suffered a left thalamic haemorrhage which partly ruptured into the lateral ventricle lsee Fig . 1) . It was caused by a very small arlerinvenous malformation (AVM I lodged in the posterior thalamus, extending into the brain stem, and fed from the left-posterior cerebral artery . The AVM was not excised because of the technical difficulty and extreme risk of the procedure . His clinical notes at that time show that he had a right homonymous hcmianopia with sparing of the macula and a right hemiparesis with right hemianasthesia . On follow-up, I year later, these problems had not resolved, and mild expressive dysphasia was also recorded- We began to see P .D . in November 1987 . His visual difficulties and hemiparesis remain . but he has only residual word-finding difficulties, affecting proper names in ,I . R . HANrsv and 3 . KAY 240 particular, in spontaneous conversation . Despite the visual field deficit, his contrast sensitivity is normal [25] . as is Sncllen acuity (6,t6) . His performance is also within the normal range on tests of object recognition, memory and spoken language . His score o143 on the Warrington Recognition Memory' rest for Faces [34] is normal . On tests devised by H ANLF:Y et at . [13], he was good at recognizing the faces and names of celebrities who have become famous since 1985, as well as those who were famous before 1980 . He has an auditory digit span of six items, and he shows good overall performance, and a strong effect of phonological similarity, in serial recall ofauditorily presented Idlers . He made six errors out of 80 on the non-word repetition lest from the PALPA battery [19], which is within the normal range, and he made only one error out of 80 on a test of auditory lexical decision . He also scored 30/30 on a test of phonological awareness [3] . His score of 15 on the Graded Naming Test [20J represents average' performance, and he scored 40,%40 on the Spoken Word Picture Matching Test front the PALPA battery . He also scored 36/36 on Humphreys and Riddoch's object decision task .5 Taken together, these results suggest that P .D . does not suffer from visual agnosia, amnesia, a short-term memory deficit, or a disorder of spoken language . INVESTIGATION OF READING AND SPELLING Letter recognition P,Dn was given several tests ofletter recognition from the PALPA battery [l9] . He made I error out of 26 in lowercase to uppercase matching (matching a lowercase `l' to 'I' rather than to `L') . He made one error, which he self-corrected, in spoken-letter to written-letter matching . He also made one error out of 26 in naming uppercase letters (H'> 'F') and one error in naming lowercase letters ('('>1') . He also produced the appropriate sound for all the letters that he correctly named, with the exception of the letter `X' (which he sounded out as if it were `Zuh') . Errors in letter recognition are consistent with previous studies of letter-by-letter reading . Nevertheless, P .D .'s impairment is slight and is reminiscent of the relatively mild deficit shown by the Type 2 letter-by-letter reader 'I-P . rather than the severe problems experienced by the Type 1 patient C .H, who made letter recognition errors on approximately 50% of trials [24] . Although we have not made any formal comparisons, we have never noted any differences in PD's ability to recognize either letters or words when the font size increases . Is P .D . a letter-by-letter reader :' P .D, reports that since his illness he finds reading extremely difficult and no longer reads for pleasure . This is because of the length of time that it takes him to read each word . As a result of these spontaneous complaints, we have given him several hundred single words to read aloud under formal testing conditions . On average, he takes 6 .3 sec to read three- to four-letter words . 11 .7 sec to read five- to six-letter words, and 25 .9 sec to read seven- to eight-letter words . Both the overall latencies, and the relationship between reading time and letter length, are consistent with letter-by-letter reading . Table I provides a direct comparison with PArrenSDN and KAY'S [24] cases . Table 1 indicates that it is the performance of T .P . that most closely resembles that of P .D . Furthermore, T .P- frequently made errors in reading words even when she had correctly identified the letters . This is also the case with P .D . Whenever he is specifically asked to name theletters in a word, he is almost always accurate, even when he subsequently comes up with the wrong word . The reading problems that both patients encounter go well beyond letter recognition difficulties, therefore, and confirm that like T .P . . P .D . should he thought of as a Type 2 letter-by-letter reader . We have ascertained that the effect of word length on reading times is not clue to the *G . Humphreys ndly allowed us to use this test LETTER-BV-LFTTFR RFADWG Fig . 1 . CT scan taken on admittance showing the effects of left thalamic haemorrhage . 241 243 LETTER-BY-LETTER READING Table I . PD's reading performance in comparison with the four letter-by-letter readers studied by PATTERSON and KAY (1982) Patient T .P . PD . M .W . K,C, CH, Letter naming errors(%) 3 4' Reading latencies (Seel 5 6" 7 V 11 .2 10 17 50 7 .6 6,3 12 .8 10 .5 16 .9 15,5 25,9 24 .3 26,0 47 .5 It .? 18 .4 18 .6 32 .4 "Number of letters in wards . number of syllables, rather than the number of letters, that a word contains . Thus P .D . was given a set of 24 five-letter words from the PALPA battery which comprise eight words of one syllable (e .g . `blood', 'watch', 'bread', 'flash'), eight words of two syllables (e .g . 'hotel' . 'metal' . 'beard', `angel'), and eight words of three syllables (e .g . 'radio' .'unioti , `ivory', `enemy') . The words are matched for imageability and lexical frequency . P .D . took an average of 13 .4 sec (SD= 14 .7) to read the words of one syllable, 8 .2 sec on average (SD =4 .9) to read the twosyllable words, and 11 .2 sec (SD=6 .4) to read the three-syllable words . This indicates that there is no effect of number of syllables on the length of time he takes to read a word when number of letters is held constant . P .D,'s performance on a same-different matching task [12] provides more specific evidence about the way in which he processes sequences of letters . In this task, one word is presented immediately above another and the subject must decide, as quickly as possible, whether or not the two words match . When the two-words differed, the different letter was either at the beginning (e .g . cork-work), middle (e .g . firm farm) or end (e .g . plum plug) of the words . Unlike normal subjects, P .D . took significantly longer to match words which contained six letters (e.g . health) than words which contained five letters (e .g . lemon) or four letters (e,g, cork) . Furthermore, P .D . took significantly longer to detect a mismatch when it occurred at the end of a word than when it occurred at the beginning or in the middle . Normal subjects [12], and our control subject, by contrast, took longest when the mismatch occurred in the middle of the word, l'hese results show that P .D . processes the letters in a word from left-to-right in serial order, and support the view that letter-by-letter readers have great d .fficulty in processing letters in parallel [24] . KAY and MANLEY [18] provide complete details, including means and standard deviations, and a fuller discussion of P .D : s performance on this task . A further characteristic of letter-by-letter readers is that they encounter greater problems when confronted by a cursive script. For example, WARRINGTON and ST1ALT .1cu [35] report that J .D .C . and M .L . both found it more difficult to read words when the letters were joined together . P .D, shows a similar pattern of performance . On two separate occasions, he was gven the same list of 72 words to read from ELLi5 era! . [11] . The first time they were typed, aid the second time they were handwritten in a cursive style . Twelve months separated the two presentations . PD's performance dropped from 54/72 (75"/ fl ) correct with the typed list to 39/72 (54% ) correct with the handwritten words . In addition, there were 29 words that he read correctly on both occasions . Of these, 24/29 were read more quickly when typed, whereas only 5/29 were read more quickly when handwritten . P .D, shows no difference in the accuracy with which he reads high and low frequency 244 J . R . HASLEY and 1 . KAY words . On ELLIS et al.'s [II] list where the words are matched for number of letters and imagery, P .D . was correct on 18/24 words of `high' frequency (e .g . 'gun', 'land', 'river' . 'doctor'), 17/24 words of 'medium' frequency (e .g . lea', 'scat','hread', `pencil'), and 19/24 words of low' frequency (e .g . 'zoo', 'plum' . 'tulip' . 'bucket') . He did, however, read the high frequency words slightly quicker (mean=5 .5 sec, SD =6 .1) than medium frequency words (mean =7 .8 see, SD =6 .1) and low frequency words (mean=8 .6, SD =9 .4) . He showed similar effects with high imagery and low imagery words when asked to read 30 high imagery (e .g . 'drum', 'onion', 'radio', 'school') and 30 low imagery words (e .g . 'clue','rca)m','moment' . 'tribute') from the PALPA battery . The words were controlled for frequency and phonological regularity . He was correct on 21/30 high imagery words, and 19/30 low imagery words, lie was, however, slower to read the low imageability words . One final point is that P .D . showed no evidence of being able to access the meanings of words that he was unable to identify overtly, when they were presented for short durations . Unlike some letter-by-letter readers [9, 32], P .D . performed no better than chance at categorizing written words as Bring or non-tiring, when they were presented to him for approximately 1 sec each . In this respect, P .D . resembles the four letter-by-letter readers of PATTERSON and KAY [24], and five letter-by-letter readers examined by J3ttn ei at . [4], who also failed to show any evidence of rapid comprehension of written words . To summarize, P .D . takes an extremely long time to read each word . His performance is directly influenced by letter length, and his reading speed is consistent with the number of letter recognition errors that he makes when compared with other letter-by-letter readers . He finds reading more difficult when the letters are joined together in cursive script . Performance on matching tasks suggests that he perceives letters serially from left to right, rather than in parallel . Therefore, P . D_ appears to conform to virtually all of the criteria that are typically associated with letter-by-letter reading . Since his letter recognition performance; is only mildly impaired, and since he fails to identify words even when he can accurately report the letters, it is clear that P .D . is a Type 2 rather than a Type I letter-byletter reader . The nature of PD's reading errors P .D . was able to read approximately 70% of all the words that we gave him accurately . We have categorized a total of 105 of his reading errors . Of these, 94 can be regarded as visual paralexias, 10 are neologisms, and one might be considered a possible semantic errur('bridge'-*PATH). Twelve of the paralexias arose from letter misidentifications, 10 of them involving the letter 'b' (e.g . 'bread'-'DREAD', 'rib'-*'RID', 'bump'-*'PUMP', 'bury'-'JURY') . Five of the paralexias were difficult to classify, but may be cases of letter misidentifications leading to visual paralexias . The remaining 77 were visual errors . Examples include 'canal' read as CANDLE, 'feather read as FURTHER, 'music' read as MUSHY, 'slope' read as SLICK, `victor' read as VISITOR . Only one of the 10 neologisms was a phonological regularization . This was the word `halve' which was read as HALV . The appendix contains further details of all P .D .'s reading errors . The visual errors that P .D . makes means it is extremely unlikely that he is using phonological strategies during reading . Since a sound-by-sound strategy might also lead to reading times that increase in line with the number of letters in a word (e .g . SAFFRAN [29]), the virtual absence of phonological regularizations in PD's corpus of reading errors confirms that he should be classified its a letter-by-letter reader rather than as a surface dyslexic . The critical LETTER-BY-LETTER READING 245 question for WARRINGTON and SHALLICGS account of letter-by-letter reading is therefore whether or not the way in which P .D . reads is reflected in the way in which he spells . Spelling As a result of his hemiplegia, P .D . is forced to write with his left hand . This means that his writing is very slow and not particularly tidy (though the letters are clearly legible) . However, his main spontaneous complaint is that, since his illness, he has forgotten how most words aree spelt . We have confirmed this in formal written spelling tests conducted during several sessions over a 2-year period . In these tests, a word was read out and P .DD was asked to write it down . There was no time limit, and the word was repeated if this was requested . Out of 200 words that we have given him to spell in this way, P .D . was correct on only 63 (32%) . Spelling was no better when he was asked to write down the names of 20 of the 40 common pictures (e .g . 'fork', 'arrow' . 'cow', 'iron') from the PALPA battery . Only five of the names (2 :5%) were spelt accurately, even though it was clear that he had the correct word in mind in 19 out of 20 cases . He named the remaining 20 pictures from this set without error . We have classified the first 95 of his written spelling errors . Fourteen of these contained a missing letter or grapheme (e .g . 'pencil' was spelt PECIL,'potato' was spelt POTTO, 'sister' was spelt SlI ER) . Thirteen contained the wrong letter or grapheme (e .g . 'beard' was spelt RFARB,'shoe' was spelt SHOW, 'tiger' was spelt TIKER) . Only three could be described as errors of `partial lexical knowledge' [l0] . These were `night' spelt NIHT, 'onion' spelt OUNYEN, and 'toe' spelt TOEW . The overwhelming majority of P .D : s spelling errors (65/95) arc phonological errors or re,7ularizarions . Some examples are 'circle' spelt as SERKEL,'elephant' spelt as ELAFUNT, 'member' spelt as MEMRA,'photograph'spelt as FOTERGRAF,'queen' spelt as CWEN, 'wallet' spelt as WHOLET, and `water' spelt as WORTER . This is very similar to the pattern of spelling errors made by letter-by-letter reader T .P . [24], which are, as we noted earlier, characteristic of surface dysgraphiu [14] . P .D . appears to have very line lexical knowedge available when attempting to spell words, and is forced to rely on fairly rudimentary sublcxical phonology . This was confirmed by a direct comparison of his ability to spell regular and irregular words . He was given 20 high-frequency regular words (e .g . 'cat', 'banana', 'potato', 'canal', 'swim') and 20 high-frequency irregular words (e .g . 'move','knock','ghost' .'photograph', 'sledge') from the PALPA battery spelling test . The words in this test arc matched as closely as possible for imageability, number of letters and syllables .. part of speech and morphemic complexity . P .D . spelt 12/20 regular words correctly, but only 2/20 irregular words ('egg' and 'move') were correct . A Fisher's Exact test showed that this difference was significant (P=0 .003) . A similar pattern emerged when he was asked to spell 48 words from a list devised by PARKIN cr al . [23] in which the exception words are all orthographically regular and contained familiar lettersequences (e .g . `front' . 'pint', 'monk', 'have') . P.D . spelt 13/24 of the regular words correctly, but only 1/24 exception words ('have') was spelt correctly (F=(1 .0005) . Finally, P .D . shows a very similar pattern in oral spelling as in written spelling . Once more, he has great difficulty with irregular words and frequently produces phonological regularizaUOns in place of the correct spelling (e .g . 'blood' was spell B-L-U-D, and 'pretty' was spelt P-R-1-T-Y) . When asked to spell orally the same set of 40 regular and exception words from the PALPA battery that he had spelt in writing 12 months previously, he was able to produce 7/20 correct spellings of regular words, but only 2/20 correct exception word 246 J . R . HA~L[Y and J . KAY spellings (a diflerenc that stops short of statistical significance (y 2 =2 .3 ; P=0 .12) . Examination of his oral spelling errors indicates that the majority are governed by a strong resemblance to the sound rather than the spelling of the target, and comparison of oral and written spelling errors in Table 2 shows a close parallel between the two types of response . Since P .D . is poor at both written spelling and oral spelling, and since his spelling is impaired regardless of whether the input is from phonology or semantics, it appears most likely that P .D . has a lexical level spelling impairment . Table 2 . A comparison of oral and written spelling responses Regular words Cat Banana PoLno Canal Pet Caravan Nest Robin Spring Bump Jam Hold Holiday Bird Swim lent Sister Tiger Wind Frog Oral Written + hi nna polo knot t carovan ball na polio kanl karvan robbin + holeday + - + sprik bumpe + + hotieday had + tense sites liga wine Erg sites tikes Irregular words Oral Written Aunt Aeroplane Sword Castle Egg Elephant Ghost Giraffe Heart Knock Iamb Move Photograph Queen Shoe Sledge Soldier Squirrel Watch Yacht antiv aryoptan swod canal + ellevent goste garaf brat + tam nmve footergraf kwene snow siege solger sgirl woth yoth Boot aro plan sorde eassole Blatant goast grath hart nark lame fotergraf ewer show sleje sogher gist woche vole Comparison of reading and spelling errors If WARRINGTON and SHALUCE s [35] explanation of letter-by-letter reading is correct, then it should be the case that the way in which P .D . spells should be reflected in the way that he reads . In fact, though, P .D .'s reading and spelling are strikingly different . The first thing to note is the overall level of accuracy . P .D . was able to read approximately 70% of all the words that we gave him correctly . This figure is twice as high as the number of words that he is able to spell accurately (32%), despite the fact that the characteristics of the words are very similar--in many cases the sets of words are identical . It is difficult to see how the spelling system could be mediating accurate reading of words that P.D . cannot spell! In fact, P .D . is only able to spell accurately 38% of the words that he can read . Most important of all, the phonological errors that he makes in spelling are very different from the visual errors that P .D . makes in reading . This difference in the pattern observed in reading and spelling occurs even when performance on the same lists of words is compared . The appendix provides details of the way in which P .D . reads words that he spells incorrectly . It can be seen that words which receive regularized spellings are sometimes read correctly (e .g . CONCEPT, MONEY), are sometimes read as visual errors ('body'>BABY . 'father' > FU RTI II?R), and are sometimes given no response (e .g . `theory','valour') . In short, LETTER-0Y-LETTER READING 247 there seems little relationship between the way in which P .D . spells a word and the way in which he reads it . This means that it is extremely unlikely that his reading strategy involves the spelling system operating in reverse, contrary to the views of WARRINGTON and SHALLICE [35] and SHAT .LICP. and MCCARTHY [31] . Reading regular and exception words Further information about the possible use of phonological strategies in reading emerged from a comparison of PD's ability to read regular and irregular words . Initially, he was given a list of 40 regular (e .g . `take', 'free', `market', `effort', `plant') and 40 exception words (e .g . `come', 'sure', 'island', 'blood', 'colonel') from the PALPA battery . These words are matched for frequency, number of letters and syllables, number of morphemes, imageahility, and part of speech . P .D . performed much better on the regular words (24/40) than on the exception words (10,/40) (y'-=8 .65 ; P=0 .003) . This was somewhat surprising given the nature of his reading errors, which contained no regularizations . However, no attempt had been made to balance the regular and exception words for orthographic regularity in this list . Thus the exception words contained items such as subtle, bouquet, yacht, and tsar, which contain unusual letter combinations . It may well have been this rather than the fact that these words have exceptional grapheme/phoneme relations that made them relatively difficult for P .D . to read . In order to test this explanation, P .D . was given PARKIN et al .'s [23] list of words to read in which the regular and exception words are controlled for orthographic regularity (e .g . 'front', 'pint', 'monk', `have') . Under these circumstances, the regularity effect virtually disappears . He, was correct on 17,/24 of the regular words and 13/24 on the exception words-a difference that is no longer statistically significant (y'-=8 : P=0 .371) . It will be recalled from the previous section that when P .D . was given this list of words to write (6 months separated the two testing sessions), there was a statistically significant effect of regularity (13/24 vs 1/24) . There is an alternative explanation of why P .D . does not produce phonological regularizations during reading . There are reports [8, 17] of acquired dyslexics who use sublexieal phonology in reading, but do not make phonological errors because of an editing strategy which stops them from uttering any response which they know (from consulting their speech lexicon) is not a genuine word . These authors showed that this was the case by giving their patients a mixed list of words and non-words to read . Under these circumstances, in which an editing strategy would be inappropriate, it was observed that they started to make a large number of regularization errors . We decided to investigate this possibility with P .D . P .D . was asked to read a mixed list of 30 common words and 30 legal non-words from the PALPA battery . Thee words comprised 15 regular (e .g . `need', `same','fresh')and I s exception words (e .g . `give', `broad', 'bind') . Since the exception words in this list are orthographically regular, this provides a further opportunity to look for a genuine regularity effect in PD's reading . The non-words comprise 15 pseudohomophones (e .g . `boaf, `jale','meeti) and 15 non-homophonic non-words (e.g . 'fike', 'loth', 'noast') controlled for orthographic similarity, graphemiccomplexity, and number of letters and syllables . P .D . was given this list to read on two separate occasions, 1 year apart . Table 2 provides details of his errors on the regular and exception words on the two presentations . It can be seen that P .D . made no regularizations on the first administration of the test, and while he did make one regularization on the second administration, the overwhelming majority of errors are still visual paralexias . (By contrast, as many as 12 of the 14 irregular word errors that KATZ and J . It HANLEY and J . KAY 248 SEvush's [17] patient made under similar conditions were phonological regularizations .) In addition, the difference between the number correct for the two word-sets is very small, as it was with PARKw et al .'s [23] lists . Finally, PD's ability to read non-words (at least when mixed with real words) is poor . His overall success rate was only 47% regardless of whether the items were pseudohomophones or non-homophonic non-words . Table 3 . The way that P .D . read IS regular and IS exception words from the PALPA battery on two separate occasions . "the list inetuded 30 non-words Regular words Presentation I Presentation 2 Lend leaned lined Clip + - Same Like Shine Steer Fresh Long Prune Need Mist Dump Pair Time Speed + + shin + + + + + - + + "share" "shtee" + + + + pain r + + + Total 13 ;15 IIi15 Exception words Presentation I Presentation 2 Broad Womb Touch Some Rind Bush Most Roth Give Dove Shove Bull Dread Have Sieve "prude" wham torch same + + + dutch "hrode" "womble } same "drinned" + moss + I i + + i + + save + shave + + + save 9/tS 8 ;15 It appears, then, that P .D . does not rely on phonological strategics when he reads .* His reading errors consist almost exclusively of visual paralexias rather than phonological regularizations . He performs in a similar way regardless of whether there are non-words in the list and, in this context at least, he is poor at non-word reading . One possible objection to our account is that P .D . does show a slight, though non-sigrtiheant, difference between regular and exception words on both the PALPA and PARKIN et al . [23] lists . We would suggest, however, that the small but consistent difference between these word-types is an orthographic and not a phonological effect . Even though both sets take 'orthographic regularity' into account in their construction by using either bigram frequency [23] or N [24], there are a number of different measures of this variable (e .g . orthographic neighbourhoods and transitional trigram frequencies) that were not controlled for and may be exerting an effect . In support of this explanation, we would note that the size of the `regularity' effect diminished significantly when orthographic regularity was taken into account and that P .D . produced visual rather than phonological errors in response to exception target words . Since PD's spelling, in contrast, relics almost exclusively on sublexieal phonology, this means that he is unlikely to he using the spelling system in reverse to read, contrary to WARRINGTON and SHAUtcs [35] . Instead, it seems that his reading is characterized by an +We do not wish to claim that P .D . is unable to use s'ublexieal phonology in reading aloud . Rather that he generally makes use of visual access to lexical information, even though this procedure is error prone and results in visual paralexias . LETTER-BY-UTTER READING 24 9 attempt to access the visual word-form system, consistent with the views of PATTERSON and K aY [24] . Since he continues to make visual errors even when he has correctly identified all the letters in a word, it seems likely that he is using a strategy of approximate visual access [7], which makes him produce a response that shares several letters with the word he is trying to read . Although P .D . may well use approximate visual access because of the length of time that it takes him to recognize letters in parallel [IS], it is probable that the use of this strategy reflects a secondary impairment within the visual word-form system itself . In fact, the type of acquired dyslexia that this error-pattern most resembles is visual dyslexia [6, 21] . P D . 's reading of his own mis-spellings One of the reasons why PATTERSON and KAY [24] came to question the view that letter-hy-letter reading involves the spelling system was their observation (p . 434) that the patient T .P, was sometimes quite unable to read her own mis-spellings . We investigated this issue with P .D . by asking him to perform a lexical decision task on his own mis-spellings, and on the correctly spelt version of the same word . We used a procedure that was broadly similar to that employed by CAMPBELL [5] . Twenty-six words were selected that P .D . had spelt incorrectly during a previous testing session, and each was presented to P .D . on an index card alongside the correct spelling of the same word . Examples of pairs of words used were FENCE/FENSE, GHOST/GOAST, SCHOOL,/SCOOL . A further 26 words were selected that P .D . had spelt correctly on a previous session, and these were paired with an experimenter-generated mis-spelling . Examples were DOGMA/DOGMER, HEAD/HED, MOVE/MOOVE . P .D . was told that only one of the two letter-strings on each card was a real word, and that he should go through the pack of cards putting a tick next to the one which he considered to be correct . If P .D . is using the spelling system in reverse when he is reading, then he would be expected to select his own mis-spelling in preference to the real word . In fact, though, P .D . selected the e acct word rather than his own mis-spelling on 20/26 occasions . This figure was very s : milar to the number of his own correct spellings that he selected when the alternative was an experimenter-generated mis-spelling (19/26) . Once again, these results are not easy to reconcile with the account of WARRINGTON and SHAT-LICE [35], and suggest that very different processes are operating in P .D .'s reading and spelling . His rejection of his own misspellings shows that he has access to lexical information when he is reading that is not available to him when he is spelling . I' D's recognition of orally-spelled words In the final experiment, we examined PD's ability to identify words that were read out to ltitn letter-by-letter . The purpose was to discover the nature of the errors that P .D . would make on this task . Would he produce the sort of visual errors that occur when he is reading, or the phonological errors that characterize his spelling? There were two reasons for wanting to know the answer to this question . First, if P .D .'s performance in identifying orally spelled words resembles his reading, then this would be consistent with the idea that he does normally try to read words by identifying the names of the letters . Second, it would suggest, contrary to W ARRINC;TON and SHALLIC [35], that identifying orally spelled words makes use of the visual word Dorm system rather than the spelling lexicon . PD, was given 20 three-letter words and 20 five-letter words from HOWARD (pers . Comm .) to identify from letter names . He correctly identified 16/20 of the three-letter words and 9/20 of the five-letter words (overall success rate=63%) . This compared with scores of 17/20 for 250 J. R . MAYctY and J . KAY the three-letter words and 16/20 for the five-letter words (overall success rate= 83%) when he was asked to read the same set of words on a subsequent testing session . Most important, P .D . produced exactly the same type of errors that he makes in reading (see Table4) . Examples of these visual paralexias include S-K-Y identified as'ski' . S-M-I-L-E identified as 'smell', T-R-O-U-T identified as 'throat', and W-H-A-L-E identified as 'wheel' . The only exceptions to this pattern were two words for which P .D . offered no response (A-N-K-L-E and II-O-N-E-Y), and two neologisms (P-I-A-N-O>'pinoh', and J-E-W-E-L >'gwill') . Table 4 . P .D : s ability to recognize words from their spoken letter names, compared with oral reading of the same words 3-letter words Dog Rox Cow Ant Ear Oar Rug Owl Leg Pig Nun Ape Ink Egg Toe Cat Sky ice Ply Gym Reeognm: Read + + - Co - + -t 1 'oh' + + are + + + + + + + + ski + + womb i I+ + ski + + 5-letter words Tower Salad Smile Piano Knife Spade Tiger Chair Shirt Ankle Cider Trout Grass Pearl Shark Whale Honey Jewel Stone Lunch Recognize Read + solid smell "pin-oh" I+ "don't know throat grace pale + solid ! t shake -lstool - wheel "don't know" "gwdl" + launch "kidder" On a subsequent testing session, in order to nullify any 'editing' strategy, P.D, was given a list of 20 irregular words such as WORD and ONE whose regularized pronunciations also form words [WARD, OWN) . P .D, was correct on 8/20 (KNOW, COME, BOWL, WORD, TOUR, HEAD- SEW and BREAD) . He made visual errors on 8/20 (E-R-A>'ear', G-A-U-G-E>'gone', G-R-E-A-T> 'grant', R-O-U-G-H> 'roach', H-E-I-R >'here' .. O-N-E >'on', C-O-U-C-H >'coach', B-E-A-R >'bird') . He made two responses which could he either visual errors or regularizations (S-I I-O-E>'show', W-H-E-R-E >'weir') . He made no response on 2/20 (CEILING and HEIGHT) . Even where an editing strategy would not prevent regularized pronunciations, then, P .D . still shows little sign of using phonological strategies on this task . Instead he typically responds with either the correct word or with a visually similar word . These findings suggest that there are indeed striking parallels between the way in which P .D, reads words, and the way in which he identifies words from their spoken letter-names . This is consistent with the views of both PATTERSON and KAY [24] and WARRINGTON and SHALLIC[ [35] that letter-by-letter readers identify written words via serial letter identification . What Warrington and Shallice cannot explain, of course, is the striking difference between the visual errors that P .D . makes in identifying orally spelled words and LFTiER-8Y-LFrra% READING 21 the phonological errors that he produces when attempting to spell words . The comparison between identifying orally spelled words and oral spelling is particularly striking . What this means is that PD's identification of orally spelled words, like his reading, seems to involve the visual word-form system rather than the spelling system . The recent literature provides two independent reasons for believing that it is best to associate the identification of orally spelled words with the reading system rather than the spelling system . The first is a priming study by MoNSST .t . [22] that used normal readers as subjects . Monsell reports that the identification of orally spelled words facilitated subsequent recognition of these items on a visual lexical decision task relative to words that had not been presented earlier . Consequently, Monsell (p . 319) concluded that "letter-by-letter identification of words does activate the orthographic lexicon normally used for reading" . Secondly, KATZ [16] specifically investigated the recognition of orally spelled words in an acquired dyslexic, NP ., whose reading and spelling were both impaired . Like PD ., the type of errors that N .P . made in recognizing orally spelled words were much more similar to the type of errors that he made in reading than to the type of errors that he made in spelling (see p . 218 for examples) . In addition, N .P . showed a similar overall level of accuracy in reading as in recognizing orally spelled words, whereas spelling was impaired more severely . There were also words that N .P . was able to spell correctly, but could not identify either from their orally presented letter names or during reading . Katz concluded (p . 216) that W ARRINgTON and SHALLICL''s [35] account of the recognition of orally spelled words did not fit his results as well as PATTeusoN and KAV's [24] alternative . What all this indicates is that the visual word-form system can accept input from serial sequences of letter names presented in either the visual or the auditory modality, as well as from sequences of letters picked up in parallel during visual perception . Consequently, we would suggest that PD's main problems in reading come about for two distinct reasons . First, he has lost the ability to pick up letter information in parallel, and instead attempts to access the visual word-form system via serial letter identification . Second, there is probably an additional impairment at the level of the lexical representations themselves . This means that P .D . must frequently rely on a process of approximate access, and makes a number of isual errors as a consequence . One issue that has not been resolved is the precise way in which the visual word form system might be accessed in either reading or the recognition of orally spelled words following serial identification of the word's letter names . One possibility is that the letter names might he converted into a visual image of the word which might then be fed into the visual word form system . We have several times asked P .D . for his own impression of how he reads . It is interesting to note that on one occasion he did come up with an account very similar to this . He said that he first tries to form a mental picture of all the letters which he then attempts to read as a word in his mind's eye . The problem that he apparently encounters when using this strategy is that he cannot `see' all the letters simultaneously in his visual image. On all the other occasions that we have asked him, however, P .D . has been much more vague about the way that he reads and appears to be quite unclear about precisely what strategy he is attempting . CONCLUSION The primary aim of our case study of P .D . has been to evaluate two conflicting accounts of the compensatory strategies that are involved in letter-by-letter reading . In so doing, we have 252 l . R . HANLF:v and 7 . KAY come up with a variety of reasons why the explanation put forward by WARRINGTON and SHALLICE [35] and SHALLICE and MCCARTHY [311 cannot explain the reading strategies that this patient uses . Their hypothesis that letter-by-letter reading involves the use of the spelling system in reverse cannot account for the different type of errors that P .D . makes in reading and spelling, the absence of any strong correlation between words that he can spell and words that he can read, and the way in which he chose real words over his own incorrect spellings in lexical decision tasks . On the other hand, the theory put forward by PATTERSoN and KAY [24] provides a much more plausible account of the strategies that this patient seems to use during reading . Evidence has been presented which suggests that P .D ., like normal subjects, reads words by accessing the visual word-form system rather than the spelling system . What makes PD's reading so laborious is the fact he can no longer identify letters in parallel, and must access the visual word-form system by serial identification of letter forms . KAY and HANLEY [18] provide further discussion of the precise point at which the normal reading process has broken down in the case of P .D . The findings that we have obtained with P .D . complement those that have recently been reported by RAPCSAK et al . [27] in their investigation of patient P .T . Like PD ., P .T . makes phonological errors in both written and oral spelling, but visual errors in reading . Furthermore . P .T . is a Type 1 letter-by-letter reader who has a severe letter recognition deficit, while P .D . is a type 2 letter-by-letter reader who has problems in word recognition even when he reports letters accurately . It therefore follows that the reverse spelling hypothesis is unable to provide an adequate account of either type of letter-by-letter reader . Is there any way in which the reverse spelling hypothesis might be altered so as to accommodate the pattern of performance that P .T . and P .D . present? One possible argument might be that the poor spelling of these two patients comes about because of an inability to access representations in the spelling system for the purposes of spelling . It might then be claimed that lexical representations are nevertheless still stored in the spelling system and can be accessed for the purposes of letter-by-letter reading . In other words, P . D . and P .T . do have lexical representation available in the spelling system, but they can only access them whilst reading . not whilst spelling . It is important to note that this new account would constitute a significant weakening of the hypothesis that letter-by-letter reading involves the spelling system . Perhaps the main reason for the popularity of WARRINWON and SnALLtcE's [35] and SHALLICE and MCCARTttv's [31] account relative to PATTERSON and KAY [24] is that it predicts the parallels that up until recently have been reported between the reading and spelling of individual letter-by-letter readers such as T .P . and K .C . [24] . An adherent of the new account might chose to argue that these links are purely coincidental and that, as a consequence, the cases of P .D . and .T P . (where the relationship breaks down) does not contradict the reverse-spelling hypothesis . Nevertheless, they would then have to acknowledge that cases such as T .P . and K .C . no longer provide any support for it . Under such circumstances, there would not seem to be any compelling reason to continue to advocate the spelling hypothesis . Such a modification of the reverse spelling hypothesis faces a more fundamental problem, however . It will be noted that the argument is analogous to the one that ALLPORT and FUNNELL [I] made when they argued that dissociations between reading and spelling performance do not necessarily force the conclusion that separate lexicons are involved in reading and spelling . These dissoaiations might occur, they argue, because a single set of LErTER-RY-LFTTER REAI)INO 253 le_cical representations can be accessed for the purposes of reading but not for spelling (or vice-versa) . Whatever the merits of this argument, it is clearly not one which could be advocated by a supporter of the reverse spelling hypothesis since the logic of the reverse spelling hypothesis demands that there are separate lexicons for reading and spelling . It w ~uld therefore be quite paradoxical if a supporter of the reverse spelling hypothesis were to use the same argument to claim that dissociations between reading and spelling performance in letter-by-letter reading reflect access impairments . Finally, if the Warrington and Shallice account now seems less plausible than it did, so Patterson and Kay's theory seems rather more attractive today than in 1982 . This is because of recent studies that have investigated the recognition of orally spelled words in normal subjects [22], and in an acquired dyslexic who is not a letter-by-letter reader [16] . These two studies have greatly increased our knowledge of what is involved in this task, and both come to a strikingly similar conclusion . This is that the identification of words from their letter names makes use of the reading system rather than the spelling system, consistent with the Yews of PATTERsoN and KAY [24] . Furthermore, a specific model of how serial-letter identification might address the visual word-form system has been provided [16], whereas the idea that the spelling system can he accessed in reverse' [31 ; 35] still awaits detailed specification . We therefore believe that the theory put forward by PATTERSON and KAY [24] is now firmly established as the dominant account of the reading strategies that letter-byletter readers employ . .lcknowfedgementaThe authors would like to express their gratitude to P .D . for giving up his time so rreely to he tested . We would also like to thank PD's spouse and Norma Pearson, both of whom enabled us to make and maintain contact with P .D . M r Kal bag Consultant Neurosurgeon at the Regional Neurological Centre, Newcastleupon-Tyne, allowed access to neu rodiagnostic information . A version of this paper was presented to a meet ing of the Experimental Psychology Society at Cambridge University in July 1989 . REFERENCES 1 . ALLPORT, D . A . and FIINNrLL, E . Components of the mental lexicon . Pllil . i'ans . Royal . Soc . B295, 397 410, 1981 . 2 . BEAuvois, M .-F, and DEROUTSNE, J . Lexical or orthographic dysgraphia . Brain 104,21 50, 1981 . 3 . BRADLEY, L . Assessing Reading Difficulties : A Diagnostic Approach . Macmillan, London . 1980 . 4 . Bun, D . N . . BLACK, S . and Howaa .L, J. Word recognition and orthographic context effects in a letter-by-Letter reader . Brain Lang -36, 357-376, 1989 . 5 . CAMPBELL, R . One or two lexicons for reading and writing words ; can misspelling shed any light? Cognll . .Neuropsychnl.4,487-499,1987 . 6 . CASEY, T . and ETTLINGER, C. 'the occasional `independence' of dyslexia and dysgraphia front dysphasia . J . Neurol . Neurosurg . Psychiat . 25, 228 236, 1960 . 7 . CoLrHaART, M . Disorders of reading and their implications for models of normal reading . Visib . Lang . 15, 245-286 . 1981 . 8 . COLTItEART, M . and FUNNEL . E . Reading and writing : One lexicon or two, h Language Perception and Production : Shared Mechanisms in Listening . Reading and Writing, D . A . ALLRJRT, D . C . MnCKAY, W . PRINz and E . SCHEERER (Editors) . Academic Press, London, 1987 . 9. Cost.ETT, II . B . and SAPFRAN, P .M . Evidence roe preserved reading in "pure alexia" . Bruin 112, 327-339,1989 . 0 . ELLIS, A . W . Spelling and writing land reading and speaking) . In Normality and Patholngp in Cognit ire Functions, A . W . ELI-1S (Editor) . Academic Press, London, 1982 . L ELLIS, A . W., ELUDE, B . M . and YOUNG, A . W . "Neglect Dyslexia" and the early visual processing of letters in words . Cognit . Neerapsycho! 4, 439 464 . 1987 . 2. FRIEDRICII . F . J ., WALKER, J . A . and PGSNER, M . r . Effects of parietal Lesions on visual matching : Implications for reading errors . Cognit- Neuapaecho! . 2, 253-264, 1985 . 13 . HANLEY, J . R ., PEARSON, N . P . and Youx A . . w .lmpaired memory nor new visual corms . Brain 113,1131-1148 . 1990 . 14 . HA'rFIEI,n, F . M . and PATURSON, K . F . Phonological spelling . Q . J . exp . Psycho!. 354, 451-468, 1983 . 15 . HOWARD, D . Letter-by-letter readers : Evidence for parallel processing . In Basic Processes in Reading : Visual Recognition, D . BLSNFR and C . W . HOMPHRFYS (Editors) . Lawrence Edbaual, Hillsdale, NJ . 1991 . 254 J . R . HANLEY and J . KAY 16. KATZ,R .B .Recognizingorallyspelledwords : Ananalysis olproceduressharedwith readingandspelling . Brain Lang . 37,201-219,1989 . 17 . KATZ, R . B . and SEVUSH,S . Accurate readingby non-lexical means : Acasestudy . Brain Lang . 31,252 266.1987 . 18 . KAY, L and HASLEY, J . R . Simultaneous form perception and serial letter recognition in a case of letter-by-letter reading . Cognit . Neuropsychol . 8,249-273, 1991 . 19 . KAY, J ., LF sER, Rt and CoernEART, M . Psvcholinguisric Assessments of Language Processing in Aphasia (PALPA) . Lawrence Edbaum . London, 1991 . 20. MCKLNNA, P . and WARRINGmN, E . K. The Graded-Naming Test. NFER-Nelson, Windsor, 1983 . 21 . MARSNALI, L C . and NF:wtvmau, F . Patterns of paralexia : A psycholinguislie approach . J . Psychofmg . Res . 2, 175-199,1973 . 22. MONSEt, S . Nunvisual orthographic processing and the orthographic input lexicon . In Attention and Performance XO, M . Col tixRART (Editor) . Lawrence Eribaum, Hillsdale, NJ, 1987 . 23 . PApICIN, A . J ., MCMULLEN, M- and GRAYSTONE, D . Spelling-to-sound affects pronunciation latency but not lexical decision . Psycho!. Res . 48, 87-92, 1986 . 24. PArrrRSON, K. E . and KAY, l . Letter-by-letter reading : Psychological descriptions of a neurological syndrome . Q . J. rep . Psycho! . 34A, 411-441, 1982 . 25 . PELLI, D . G., Rousos, J . C . and WILKINS, A . J. The design of a new letter chart for measuring control sensitivity . din . V'is . Sd . 2, t87-199, 1988 . 26. PRIOR M . and MCC0RRISTIIN, M . Acquired and developmental spelling dyslexia . Brain Lang . 20, 263-285, 1983 . 27 . RAPOSAK, S . Z ., RUnras, A . B . and LAGUNA, J . F . From letters to words : Procedures for word recognition in letter-by-letter reading . Brain Lang . 38, 504-514, 1990 . 28 . REOreR-LORENZ, P . A. and BRONN, J . L . A prelexical basis for letter-by-letter reading. Cognit . Neuropsychol .7. I-20, 1990 . 29 . SAFERAN, E . M . Lexiealisation and reading performancc in surface dyslexia . In .Surface Dyslexia, K- PATTERSON, J . C . MARSHALL and M . COLTHEART (Editors) . Lawrence Erlbaum, London, 1985 . 30 . SCHACTER, D . L ., RAVCSAK, A . B ., RERUNS, A. B . . TnARAN, M . and LAOLNA, .f. Priming effects in a letter-byletter reader depend upon access to the word form system . Neuropspchologia 28, t079-1094, 1990. 31 . SHALLICE, T. and MCCARTHY, R . Phonological reading: From patterns of impairment to possible procedures . In Surface Dyslexia, K . PATTERSON, J . C. MARSHALL and M . Cot.ruaaRT (Editors) . Lawrence Earlhmlm, London, 1985 . 32 . SHALLICL, 7. and SAEPRAN, E . Lexical processes in the absence of explicit word identification : Evidence from a letter-by-letter reader . Cognit. Neuropsydloi . 3, 429-458, 1986 . 33- SHALLICE, T . From Neuropsychology to Cogniriue Structure . Cambridge University Press, Cambridge, 1988 . 34 . WARRINGTON, E. K . Recognition Memory Test . NFER-Nelson, Windsor . 1985 . 35 . WARRINGTON, E. K . and SHALLICE, T . Word-form dyslexia. Brain 103,99 112, 1980 . 255 LETTER-BY-LETTER READING APPENDIX (t) A classification of P .D .'s oral reading responses (a) Visual paralexias . TARGET RESPONSE TARGET RESPONSE TARGET RESPONSE answer barge h -idge hooch circle cord engine feather hobby kg night peril s roe snit smog soot tomb tractor rictor weasel wrath insider bark burglar brook scissors cod ending further happy ledge neither pearly shove shirt smoke set Tony presenter visitor whistle with attitude beard brief canal colonel cull fair flannel indict lend nurse priest shrewd sky soldier stench touch tsar wallet wedding wrong attained board brave candle cooler cute fire final tndesit leaned nuts pressed Swede ski scissors sense torch teaser weather wording one audience beret broad choir context curb father foot infest music opium scarce sieve slope some sure touch uncle watch womb iron admittance Bert board cheer contest cub further fort m first mushy open shake slave/save slick same shirt took uncivil water wham around taster usual unusual Ic) Visual poralexia (but possible violation of final 'e' rule) . dose doss duke duck osc loss shine shin ,;make snack hate smoke hat smock Ib) Visual poraleeia (but possible morphemic errorl . nervy taste nerve d) Visual paralexia (but possible 'regulaiization) . oeard :gown bird clone heart hurt sweat sweet bread rib lamb square dread rid lamp spare dread curb bury gun bread curd jury bun (f) Letter nosidentifieariwt+ visual paralevio . hat dead body both dutch break baby drink bird drink castle seep yacht /efa'stl,,/ le) Letter ntisidentifieation . boy debt hump queen day bet pump gene (g) Possible semantic error . bridge path (hl Weoioylsnrs . ape knock sword halve /pa/ /nook/ /swnd/ /h~lv/ hat pot tongue /tnngir ;jsktd .r 256 J . R HANLEY and 1 . KAY (2) A classification of P.D.'.s writ Ien spelling responses (a) Reyularizarions (phonic spellings) . TARGET RESPONSE READ AS TARGET RESPONSE READ AS body bread castle child circle clown clue concept effort I-elephant elbow fact father fence I--fire ghost gravy heart holiday kitten letter member money pact photograph pillow plea potato pupil queen radio road robin rib school session sledge sword system theory valour wallet water weed bodes bred cassole chilld serkel cloven slow concept eful elafunt ellbaw racket farther fence fihe ghost gravi hart holeday kiten Icler memba muney packs fotergraf pilo ploy pottatow piupel oven radco rode robhin ribb scoot sesyon steje cord sistem therey solo whales worsen wend baby dread ehastle + scissors clone + {+ skull seat realm scone selc relme sortie woche wise aeroplan eosin ketel marug plan snack villas wow bath T + + 1 + + + ending "don't know" "don't know" snack "don't know" "don't know" f "don't know" further "don't know" + + hurt + + + + f "don't know" + + + gene + + + rid "don't know" + "swodd" `don't know" "don't know" weather + + (6) Regnlaricat ions ussociated with fmal'e . heel hotel lamb lunch pill plum hele howtele lame lunche pile plume + -j lamp + + + sword watch wig aeroplane engine kettle marriage plane snake village woe battle (c) Missing let icr . aunt arms karma banana knot canal caravan karvan church cerg pencil pecil picture picker potato potto principle prinapul liter sister sky lye squirrel quiral tongue tune character karitor Id) Wrong grapheme . beard cow doctor episode mother shoe square surf tea tiger uncle yacht knock bearb caw dorcktor epesold mucer show sgcr serve lye tiker huncel vote nuns (e) Partial lexical knowledge . night onion toe nihl ounven toew