The Quarterly Journal of Experimental Psychology
Section A

ISSN: 0272-4987 (Print) 1464-0740 (Online) Journal homepage: http://www.tandfonline.com/loi/pqja20

Letter-by-letter reading: Psychological descriptions
of a neurologial syndrome
Karalyn Patterson & Janice Kay
To cite this article: Karalyn Patterson & Janice Kay (1982) Letter-by-letter reading: Psychological
descriptions of a neurologial syndrome, The Quarterly Journal of Experimental Psychology
Section A, 34:3, 411-441, DOI: 10.1080/14640748208400852
To link to this article: http://dx.doi.org/10.1080/14640748208400852

Published online: 29 May 2007.

Submit your article to this journal

Article views: 159

View related articles

Citing articles: 75 View citing articles

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=pqja20
Download by: [Central Michigan University]

Date: 22 October 2015, At: 06:21

fhuxrterly Journal of Experimental Psychology (1982)3@, 411-44’

LETTER-BY-LETTER READING :
PSYCHOLOGICAL DESCRIPTIONS OF
A NEUROLOGIAL SYNDROME
Downloaded by [Central Michigan University] at 06:21 22 October 2015

KARALYN PATTERSON AND JANICE KAY

MRC Applied Psychology Unit, I 5 Chaucer Road, Cambridge
CB2 2EF, England
Following a general description of the characteristics of letter-by-letter reading
and a summary of previous explanations of this reading deficit in both neurological
and psychological models, four single-case studies of patients with this syndrome
are presented. One central topic, addressed by experimental investigation, concerns comprehension of written words. Despite the use of multiple techniques,
no evidence was obtained for the hypothesis that comprehension of a word could
occur prior to or in the absence of the letter-by-letter analysis required for oral
reading. It appears that these patients must do sequential letter identification of
a word in order both to understand it and to report it. A second central topic,
addressed through analysis of reading errors, concerns procedures for word recognition. Two of the four patients showed a “pure” letter-by-letter syndrome,
with no difficulty in word recognition once the component letters had been identified. For the other two patients, an additional lexical deficit often prevented a
correctly identified sequence of letters from achieving recognition as the correct
word. Alternative interpretations of these patterns are discussed in terms of a
process model.

Introduction
Acquired dyslexias (adult reading disorders resulting from neurological damage)
present a variety of specific syndromes which have been differentially salient within
different research traditions. Given a reference to acquired dyslexia (or alexia)
with no further identifying label, a research neurologist would probably assume
that one meant “Dkjerine’s syndrome”, that is, alexia without agraphia (or letterby-letter reading, as it will be called here*). A research psychologist, on the other
hand, might be more likely to think of deep dyslexia or surface dyslexia. Apart
from historical precedent, there may be coherent reasons for this difference in
emphasis. Of all the known reading disorders, letter-by-letter reading probably
has (a) the most consistent set of lesion sites and (b) the most ready account in
*This syndrome (at least we think it is all the same basic syndrome, though it is difficult to be
certain) has had a rather bewildering variety of different designations, including alexia without
agraphia, pure alexia, word-form dyslexia and letter-by-letter reading. We prefer the latter
because it seems both the most descriptive and the most neutral.

+

0272-4987/82/03041
I 31

$03.00/0

@ 1982The Experimental Psychology Society

Downloaded by [Central Michigan University] at 06:21 22 October 2015

4’2

K. PATTERSON AND J. KAY

terms of simple disconnection between anatomically separate “centres” of brain
function. Such factors would obviously enhance compatibility between this
syndrome and a neurologically based approach. None the less it is slightly puzzling
that cognitive psychologists have largely ignored this syndrome, at least until very
recently (for exceptions, see Gardner and Zurif, 1975; Staller et al., 1978;
Warrington and Shallice, 1980). After all, the existence of a plausible neurological account does not preclude accounts at other levels of description. Indeed
it has been argued (Mehler and Morton, 1981) that neurological data such as
lesion site place no constraints whatsoever on psychological or functional descriptions.
One reason contributing to the relative neglect of this syndrome within cognitive
psychology may be this: letter-by-letter reading is not, in the terminology of
Shallice and Warrington (1980), a form of central dyslexia, but is rather one of the
peripheral dyslexias. T h e functional locus of the deficit in letter-by-letter reading
is probably relatively “early” in an information-processing sense. Questions
about reading likely to be of interest to many cognitive psychologists, such as the
organisation of the internal lexicon or the semantic system, may be addressed more
clearly by syndromes of central dyslexia (deep dyslexia, phonological dyslexia and
surface dyslexia) than by the peripheral dyslexias (attentional dyslexia, dyslexia
involving neglect, and letter-by-letter reading).
Furthermore, regarding peripheral dyslexias, there might even he some doubt
concerning a critical assumption in the rapprochement between neuropsychological
research and cognitive psychology. Patients with an acquired dyslexia clearly
read in a way that is different from normal reading. It is generally assumed that
this difference stems from the loss or impairment of one or more functional components of the normal reading system, rather than from reliance on wholly different
procedures (Shallice, 1979, 1981; Coltheart, 1981; Patterson, 1981). Suppose,
however, that there is disruption of a relatively early stage of processing; might
this mean that the patient no longer has access to the rest of the normal system?
Peripheral dyslexias, in other words, might not only be of less compelling interest
to cognitive psychologists; in principle, they might even fail to warrant status as
legitimately relevant to many issues of normal reading. Were this the case, of
course, it would not eliminate interest in the nature of such syndromes; but the
interest might be found primarily amongst neurologists, speech therapists, even
neuropsychologists-but not amongst cognitive psychologists. This, then, might
explain in part why most studies of letter-by-letter reading derive from a neurological tradition rather than a psychological one.
Whether this analysis is valid, or whether psychologists know little of letter-byletter reading simply because it has not been brought to their attention, we believe
that the syndrome deserves attention and that it does legitimately address questions
of interest to those who study normal reading.
T h e organisation of this paper will be as follows. Firstly, we shall present a
brief, qualitative description of the syndrome, a description assembled from previously published accounts. Secondly we will summarise the way in which the
syndrome has been interpreted, in a neurological framework and in a psychological framework. T h e third and major section will present data from four

LETTER-BY-LETTER READING

473

individual case studies of letter-by-letter readers. I n particular we will focus on
(a) an evaluation of word comprehension by these patients, and (b) an analysis of
their reading errors which suggests that there are two distinct varieties of letterby-letter reader.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

Qualitative description of letter-by-letter reading
A patient with this syndrome appears to read a word in the following way:* he
identifies each letter of the word in succession, often but not invariably naming the
letter aloud as it is identified; when he has reached the end (or nearly the end) of
the sequence of letters, he produces the word. Not surprisingly, the time required
to read a word is abnormally long, and it is also a monotonic function of word
length (Staller et aE., 1978; Warrington and Shallice, 1980). Although cases have
been described where accuracy of letter identification is perfect (e.g. one of
Warrington and Shallice’s patients), the majority of letter-by-letter readers have
a letter-identification deficit. This response from one of our patients (who will
be described shortly) would be typical of many letter-by-letter readers : given the
word men, he responded “h,e,n, hen”. There is some indication that the impairment in dealing with symbols is exclusive to alphabetic symbols: a frequently
reported feature of the syndrome is preserved reading of Arabic numerals (e.g.
HCcaen and Kremin, I 976). Letter-by-letter readers do not show significant
word-class effects like those found in the central dyslexias (see Shallice and
Warrington, 1980; Coltheart, 1981; and Patterson, 1981 for summaries). For
example, oral reading by phonological dyslexic patients shows a severe deficit for
regularly spelled non-words as contrasted with real words : deep dyslexic patients
are severely impaired on reading abstract words relative to concrete words; and
surface dyslexic patients read irregularly spelled words significantly less well
than regular words. None of these major contrasts is found in letter-by-letter
reading.
The above paragraph gives a rough impression of reading performance. As
for other characteristics, the most common neurological sign found in letter-byletter readers is a right homonymous hemianopia (RHH), that is, blindness in the
right half of the visual field due to a lesion in the occipital cortex of the left
hemisphere. There are, however, cases of the syndrome without RHH, that
is, with intact visual fields (e.g. Greenblatt, 1976; Vincent et al., 1977).
As for general language skills, typically it is only reading which shows any
significant deficit (thus the designation “pure alexia”). Speech is often
completely normal, though sometimes one finds fluent speech with anomia.
Object naming and, even more commonly, colour naming can be impaired.
*It is apparently the case that the same deficit can occur in so severe a form as to render the
patient completely alexic (e.g. Caplan and Hedley-White, 1974). When this occurs, of course, the
description given here of how these patients read words is inapplicable, because totally alexic
patients do not read. As we will note again later, there is a problem here: if a patient does not read
words at all, then even if other features of his performance match the syndrome of letter-by-letter
reading, it is difficult to describe the patient’s reading in this way. One is saying that the patient
would read words letter-by-letter if he did read. This problem does not arise with respect to our
patients, all of whom read words; but it seems necessary to mention the issue since some of the
“classical” cases were totally alexic.

414

K. PATTERSON AND J. KAY

Writing is often completely normal (thus the term “alexia without agraphia”)
though it is not uncommon to find some writing difficulties at least in the
early stages. The contrast between reading and writing performance is probably the most striking feature of this syndrome. Asked to write a sentence
to dictation, a patient may do so easily and without error. Given this written
sentence to read a few minutes later, however, the patient will have to decode it
in a painfully slow and error-prone fashion and indeed may be incapable of
reading it correctly.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

Neurological interpretation
The standard interpretation of letter-by-letter reading, formulated originally by
Dkjerine (1892)and more fully Geschwind (1965),attributes the deficit to a disconnection between the angular gyrus in the left hemisphere and visual input from
both hemispheres. The angular gyrus is considered an essential structure for both
reading and writing. Since this structure is itself intact in these cases, the patient
is able to produce written output; but because of the disconnecting lesion(s),
written input cannot reach the angular gyrus and therefore reading is grossly
abnormal. The most common anatomy of the disconnection involves damage in
two separate locations. A lesion in the left occipital lobe (producing RHH) prevents any visual information from reaching the left hemisphere directly. The
right hemisphere is intact and therefore the right occipital cortex receives normal
visual input; but a lesion in the splenium of the corpus callosum prevents transfer
of the necessary information from the right hemisphere to the left angular gyrus.
Although many of the cases verified by post-mortem examination have shown this
precise pattern of pathology, the syndrome also occurs with other loci, such as a
unilateral left-hemisphere lesion in the occipito-parietal (Greenblatt, 1976), temporal (Vincent et al., 1977)or temporo-parietal (Staller et al., 1978)region. It is
hypothesized that damage in these areas can also have the effect of blocking access
to the left angular gyrus from both hemispheres. Greenblatt (1976)in fact summarises four possible patterns of neuro-pathology resulting in alexia without
agraphia, but interprets them all in terms of the basic disconnection already
described.

Psychological interpretation
As noted by Kurachi et al. (1979), “several features of this syndrome were
explained by the callosal disconnection theory. . ., but underlying impairment of
the neuropsychological processing is not yet clearly defined” (p. 297). From the
point of view of cognitive information-processing models of reading (see e.g.
Marcel, 1980;Morton and Patterson, 1980;Shallice, 1981))we would want to ask
at least two questions. Firstly, what is the functional locus of the impairment
which causes reading to be so grossly abnormal? Secondly, what functional processes or components are operating when the patients read? The majority of
letter-by-letter readers are dys-lexic rather than a-lexic. Grossly abnormal their
reading might be, in speed if not in accuracy; but most patients have some reading
ability, and for many the retained ability is considerable. Such retention is some-

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

4’5

times acknowledged in the neurological literature : “However, since some measure
of reading ability is retained in most alexic patients, one must assume that unaffected splenium or anterior callosal fibres are of some, albeit limited, importance
in the transmittal and perception of written language” (Ajax et al., 1977, p. 688).
Psychologists need not reject or ignore such neurological descriptions, but we will
still want to identify the cognitive procedures allowed via these unaffected splenial
or anterior fibres.
T o our knowledge, the only major attempt to interpret this syndrome in an
information-processing framework is by Warrington and Shallice (1980). These
authors offer the following answers to the two questions posed in the preceding
paragraph. As to functional locus, they argue that letter-by-letter readers have a
deficit of the visual word-form system; indeed Warrington and Shallice re-name
the syndrome “word-form dyslexia”. The function of the word-form system is
to segment letter strings into recognisable units ranging in size from individual
graphemes up to whole morphemes or words. Any letter string can result in
many different segmentations; however, a real word can and generally will achieve
an orthographic representation at the level of the whole word or morpheme,
whereas most nonsense words could only be represented in units smaller than the
whole letter string. In terms of the sequence of information processing stages,
the word-form system follows early visual processing and precedes both phonological and semantic processing (see Shallice, 1981, for further description). As
to the nature of the processes actually operating in letter-by-letter reading (where,
they claim, the word-form system cannot be used), Warrington and Shallice are
somewhat less explicit. Their general interpretation seems to be that once the
written letter string has been converted into a sequence of auditory letter names,
such a representation is compatible with the system used for spelling. One can
go from a word (either internally generated or auditorily presented) to its spelling,
e.g. book+b,o,o,k; the notion is that this procedure can also operate in reverse.
Thus, using procedures which form part of the spelling system rather than the
reading system, the patient moves from the letter names “b,o,o,k” to the word book.
The question of whether we agree with Warrington and Shallice’s position will
be postponed until we have presented some data.
Case reports

( I ) M.W., a right-handed postmistress, suffered a cerebro-vascular accident
(CVA) in October 1980 at the age of 58. Damage in the medial half of the left
occipital lobe was confirmed by CT brain scan. Her only major neurological sign
was a dense right homonymous hemianopia which persists unchanged. We first
saw M.W. six weeks post-CVA, when she was in the Rehabilitation Unit of
Addenbrooke’s Hospital for treatment of the reading difficulty which was her
primary complaint (reading had always been a major and favourite activity as well
as being essential to her job).
M.W.’s speech was normal. Her nominal functions were also reasonably
intact: she scored 14/15 correct in naming objects both to visual presentation and
from spoken description (tests from Coughlan and Warrington, 1978). However,

Downloaded by [Central Michigan University] at 06:21 22 October 2015

4’6

K. PATTERSON AND J. KAY

she appeared to have a mild degree of visual agnosia: in naming line drawings of
objects (materials from Patterson et al., in press), she scored 79/92=86%. Her
errors seemed largely to be of an agnosic rather than an anomic character: she
might say that she had no idea what an object was (e.g. a picture of a telescope)
or she might give a somewhat bizarre response (picture of a torch-+“a pen or a
top? a kid’s top?”). She showed no evidence of either agnosia or anomia for
colours.
M.W.’s spelling to dictation, both oral and written, was good; in fact her
occasional errors were much like those of “normal” spellers. For example, her
oral spelling of “susceptible” ended “a,b,l,e”, and her written spelling of “siege”
was sezge. Her spontaneous writing was adequate but rather untidy, and she sometimes ignored the right side of the page, filling only the left half of a line. Her
performance on identifying words spelled aloud to her was excellent.
Her reading of words or sentences was almost entirely letter-by-letter and painfully slow. She usually produced the correct word eventually, though often after
one or more incorrect attempts. Her ability to identify single letters was impaired
but not severely so. When the alphabet was presented one letter at a time in
random order, she correctly named 25/26 upper-case letters and 22/26 lower-case
letters. She was given 40 trials of matching upper and lower-case letters; half of
the trials required a positive response (e.g. D and d, presented one above the other)
and half required a negative response (e.g. D and g). She was correct on 35/40,
and her five errors were all rejections of a matching pair (Ss, gG, pP, kK
and Dd).
(2) C.H., a right-handed man who had worked mainly as a chauffeur prior to
retirement, suffered a CVA in 1975 at the age of 81. H e saw his general practitioner at the time of the stroke but was never admitted to hospital. A neurological
examination was carried out in November 1979 at the Queen Elizabeth Medical
Centre in Birmingham. A C T brain scan revealed damage in the left occipital
and temporal lobes. The only major neurological sign was a complete right
homonymous hemianopia.
C.H. was 85 or 86 when we tested him but extremely youthful and lively for this
age, and he participated willingly in our studies. His speech comprehension was
quite normal, as was his speech production (except for the high rate of jokes it
contained). His nominal abilities were good : on Coughlan and Warrington’s
tests (1978), he named 14/15 objects with visual presentation and 15/15 from
description. His naming of colour patches was mildly abnormal (blue+“purple”,
purple+“blue” and red+“pink”) ; but he showed no evidence of colour agnosia,
being able to point correctly to colours for named objects like banana, tomato,
coal, etc.
C.H.’s spelling to dictation, both oral and written, was fairly good but not
perfect. Most of the errors were either phonologically acceptable (“definite”-+
definate, in written spelling) or nearly so “yacht”-+“y,a,u,g,h,t”, in oral spelling);
but occasionally his spellings were phonologically as well as orthographically
incorrect (“capacity”+compasity). It is difficult to know whether this pattern
represents an impairment relative to his pre-morbid spelling. His handwriting
was adequate though rather shaky. He was excellent at identifying words from

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

417

their oral spelling; indeed this was probably his favourite test, and usually even
difficult words like rescind caused no problems.
His reading was almost entirely letter-by-letter, enormously slow, and full of
errors. He showed a severe deficit in identifying individual letters: 16/26 correct
naming on upper-case and 10126 correct on lower-case letters. For an evaluation
of letter identification apart from naming, C.H. was given an upper-lower case
matching test (from Warrington and Shallice, 1979) in which he had to choose
the “odd-man out” in IOO triplets like f F e. Working very slowly, he managed
to score 75% correct on this task, which is significantly better than chance (33%)
but quite impaired.
(3) T.P., a right-handed female chief radiographer, suffered a left-cerebral
haemorrhage in March 1979 at the age of 50. A left temporo-parietal haematoma
was removed four days after her initial collapse; two weeks later, a further operation was performed to remove an angioma from the left occipito-temporal region
and to evacuate the residual intracerebral haematoma. In the very early stages
she had both a dense right hemiplegia and a complete right homonymous hemianopia. The former cleared quickly, leaving no residual motor impairment ; the
hemianopia remains unchanged.
T.P.’s speech was fluent but with a severe nominal dysphasia; the absence of
specific content words was striking. Given the word “grape” to define, for
example, she said: “It’s one of these little ones with lots of things which come
down, and they’re very nice to eat. They’re not much in England but they’re
trying to make them more, aren’t they?” (“Why?” we asked). “Because I suppose they want to make the drink.” A multi-modal anomia was confirmed by
tests of naming objects (a) presented visually (2115 correct), (b) from description
(2115 correct) and (c) presented tactually (3/15 correct). There did not appear
to be any agnosia for the objects themselves, as she could generally make appropriate gestures or comments (shown a clothes peg, for example, she said “when I
have to do the wash. . . ’ I ) . Likewise, in attempting to name objects from description, she appeared to have no difficulty in understanding the description (asked
“What is the name of the net or canvas bed that we hang up in order to lie on it?”,
she correctly demonstrated its shape with her hands). With colours, T.P. showed
not only anomia but agnosia, performing very poorly when asked to point to the
colour of a banana or a strawberry.
Although T.P.’s writing was better than her reading, a definite and reasonably
severe deficit was apparent in both written and oral spelling. Her spelling problem, which has been reported elsewhere (Patterson and Hatfield, 1980), might be
called “surface” dysgraphia. That is, her spelling of a word was usually phonologically acceptable, but was very often wrong, especially on irregular words (e.g.
“answer”+amer).
Samples of her pre-morbid writing (professional lecture notes)
show normal spelling and thus establish that her spelling deficit resulted from the
neurological impairment. The relevance of this dysgraphia to the pattern of
reading deficit will be discussed later.
In the earliest phase of her illness, T.P. was almost totally alexic in terms of
both oral reading and reading comprehension, Gradually she began to use a
letter-by-letter reading strategy, though she named the letters of the word aloud

Downloaded by [Central Michigan University] at 06:21 22 October 2015

418

K. PATTERSON AND J. KAY

rather less often than many letter-by-letter readers. At the level of individual
letters, her identification performance was good. She correctly named 25/26
lower-case letters, though occasionally a correct identification followed one or more
incorrect attempts. I n 28 trials of upper-lower case matching (Ddvs. Dg), she
made no errors.
(4) K.C., a right-handed woman who, together with her husband, owned and
ran a shop, suffered a CVA in December 1980 at the age of 65. The problem was
diagnosed as a left occipital haematoma, and a low-density area in the left temporooccipital region was confirmed by C T scan. The only major neurological sign
was a right homonymous hemianopia.
K.C.’s speech comprehension and production both appeared reasonably normal.
She correctly named 14/15 objects presented visually and 15/15 from description
(tests from Coughlan and Warrington, 1978). Her performance with colours was
like that described for C.H.: a mild abnormality in colour naming (e.g. blue+
maroon”, red+“brown”, others correct) but normal ability to select appropriate
colours for named objects like “tomato” and “grass”.
K.C.’s spelling performance showed the same pattern as T.P.’s (case report 3),
that is, a “surface” dysgraphia. This deficit was perhaps less severe for K.C., but
the predominant type of error was the same (e.g. “sign”+“s,i,y,n,e” in oral spelling and scyne in written spelling).
K.C.’s reading, employing a letter-by-letter strategy, was not only slow but also
error-prone. She showed some impairment in identifying single letters : 25/26
correct (though requiring some corrections of initially wrong responses) for uppercase and 18/26 correct with lower-case letters. On 40 trials of matching upper and
lower-case letters, she made five errors, all acceptances of non-matching pairs
(rC, tB, Qd, bK and Li).
(1

Experimental investigations
We have claimed that these four patients were letter-by-letter readers. The
major evidence underlying this claim, the existence of a monotonic function
between a word’s length and the time required to read it, is presented in the top
half of Table I. The 234 words contributing to these measurements came from
three sources: (I) the IOO words of the Schonell(1942) graded reading test; (2) the
word list published by Coltheart et al. (1979) composed of 39 words with regular
spelling (e.g. grill) and 39 words with irregular spelling (e.g. gauge) matched for
length, frequency and part of speech; and (3) an unpublished list (Coltheart,
personal communication) composed of 28 nouns of high imageability (e.g. blister)
and 28 nouns of low imageability (e.g. essence) matched for frequency, number of
letters and number of syllables. The numbers of observations vary from cell to
cell (the smallest being about 10 and the largest about IOO), for three reasons.
Firstly, there are unequal numbers of words of these various lengths in the lists.
Secondly, we did not collect data on all three entire lists for all four patients.
Thirdly, only correct reading responses are included in the time measurements
and, as can be seen in the lower half of Table I, the proportion correct also varies
from cell to cell. The varying N no doubt means that some of these values are

419

LETTER-BY-LETTER READING

TABLE
I

Downloaded by [Central Michigan University] at 06:21 22 October 2015

Reading latency (top half of table) and reading accuracy (lower half) as a function
of word length, for four letter-by-letter readers

Mean time (s)
to read words
of letter-length:
3-4
5-6
7-8
9-10
Proportion correct in
reading words of
letter-length:
3-4
5-6
7-8
9-10

M.W.

C.H.

T.P.

K.C.

12.8
18.4
24'3
42.1

16.9
32'4
47'5
97'3

7'6
11.7
'5'5

10.5

0.89
0.78

0.67
0.77

0.62

0.82

0.62

0'33

19'5

0.84

18.6
26.0
34'2

0.85

0.89
0.89

0.83
0.74

0.56
0.55

better estimates than others of "true" reading time; however, Table I is merely
intended to establish the basic fact that all four patients show a monotonic increase
in reading latency as a function of word length. Word length also affects the
probability of a correct response, but the function is not nearly so continuous for
accuracy as it is for latency.
Two aspects of these data are noteworthy. First is simply the extraordinary
length of the latencies: although the patients manage to read many words correctly,
the time required to do so is abnormal by several orders of magnitude. Even the
quickest patient (T.P.) needed over 7 s on average to identify a three- to four-letter
word. The second feature of the latency data is the dramatic effect of word
length, even more dramatic for these four patients than for previous patients in
whom it has been demonstrated (Staller et al., 1978; Warrington and Shallice,
1980). This word-length effect alone would be sufficient to characterise the
patients' performance as abnormal: normal readers may show either a small effect
of word length on latency of report (Frederiksen and Kroll, 1976) or indeed no
effect at all (Schiepers, 1980).
Having demonstrated the basic phenomenon of letter-by-letter reading, we turn
now to the specific questions around which our presentation of data will be organised. We note apologetically that the four patients were available at different
times and for differing periods of time; thus we do not have equal amounts of data
for all of them. Furthermore, our ideas about what to test and how to test it varied
considerably over the approximately 18 months of data collection with these
patients; therefore often we do not even have comparable data for all patients.

I. Visual perception
The first issue we address is whether the impairment in letter-by-letter reading
can be confidently identified as specific to reading, as opposed to some more

Downloaded by [Central Michigan University] at 06:21 22 October 2015

420

K. PATTERSON AND J. KAY

general deficit of visual perception. This question was considered and answered
affirmatively by Warrington and Shallice (1980), and we accept their conclusion.
T h e issue is, however, of paramount importance if one wants to draw implications
from the syndrome to issues in the study of reading. It therefore seems useful
to adduce further evidence against alternative interpretations.
(A) It seems clear that letter-by-letter reading is not due to the presence of a
hemianopia. All of our patients had a complete RHH, but some patients with
this syndrome do not (e.g. Greenblatt, 1976; Vincent et al., 1977). Furthermore,
many neurological patients with an RHH do not show this particular (or indeed
any) syndrome of acquired dyslexia. Some deep dyslexic patients, for example,
have an RHH; and reading in deep dyslexia, which shows no effects of word
length or exposure duration on reading accuracy (Patterson, 1981), is utterly unlike
letter-by-letter reading.
(B) We concur with Warrington and Shallice (1980) that letter-by-letter reading cannot be due to a general simultanagnosia, that is, a failure in either perceptual
processing (Kinsbourne and Warrington, I 962) or comprehension (Wolpert, 1924)
of the whole visual display as an entity. Three of our four patients were given
Bishop’s (1980) sentence-picture matching test, in which one of four pictures must
be selected to match a sentence (e.g. “the girl is pushed by the cow”). This test
was designed to evaluate aspects of syntactic comprehension, but it should also be
sensitive to visual simultanagnosia since the subject must choose from amongst
four different pictures. Also, the pictures are reasonably complex scenes whose
appreciation depends upon recognition not only of the individual components but
also of their relationship. In the picture alternatives for the sentence example
above, in addition to the correct picture of a cow pushing a girl, there are pictures
of a girl pushing a cow and of a cow looking at a girl. C.H., T.P. and K.C. were
given this 40-item test with (of course) auditory presentation of the sentences.
They scored 93, 90 and 95%, respectively. Response times were not measured
but none of the patients was slow to respond. Their behaviour when interpreting
a scene is simply nothing like their behaviour when faced with a word.
(C) Letter-by-letter reading cannot even be considered a failure of parallel processing which is specific to abstract symbols like letters or numbers. Warrington
and Shallice (1980) devised several relevant tests. I n one, the tachistoscopically
presented stimulus consisted of two digits placed one at either end of a word (e.g.
6STOPq) and, across the set of stimulus materials, the length of the word was
varied from four to eight letters; the patient was simply to report the numbers.
Warrington and Shallice argued that if the patients were unable to process a
sufficiently large “chunk” of visual space in parallel, their performance on this
task would decline as a function of lengthening word, They showed that there
was no such decrement for one of their patients. As Warrington and Shallice’s
other patient was not tested on this task, we thought it would be worthwhile to
present a replication of their result for one of our cases.
T.P. was given this test, using a two-field tachistoscope with an exposure duration of 150ms and with a fixation point at the right-hand edge of the area in which
the stimulus would appear.* The results in Table I1 show reasonably good
*See first footnote on following page.

42’

LETTER-BY-LETTER READING

TABLE
I1
6STOP4 task (from Warrington and Shallice, 1980):Patient T.P.’s performance
with exposure duration= 150 ms

Downloaded by [Central Michigan University] at 06:21 22 October 2015

Length of word separating numbers
Numbers correctly reported

4

7/10

5
8/10

6

7

8

~ojro

rojro

7/10

performance. The important finding is that for T.P., as for Warrington and
Shallice’s case R.A.V., performance did not decline with increasing separation
between the two numbers to be reported. It is thus unlikely that the reading
symptoms in letter-by-letter reading generally occur because the patient cannot
“see” or “attend to” a whole word.
(D) Even if a patient, given a brief presentation of 6STOP4, can report “6,4”,
one could query whether there might be a deficit of visual acuity for fine spatial
frequencies affecting the middle of a word or stimulus more than its beginning
and end.? Although we did not investigate this issue in detail, we can report
some results of visual testing for one of our patients, M.W. Her acuity as
measured by Snellen chart was normal in both eyes though (not surprisingly) she
misidentified a few letters. More importantly, her acuity as measured by contrast
sensitivity gratings was also normal, even for the finest spatial frequencies. It
seems unlikely that any explanation at the level of a visual-processing deficit will
account for letter-by-letter reading.

I I. Comprehension
The evidence presented by Warrington and Shallice (1980),and in the previous
section, argues that the deficit belongs within the domain of reading rather than
in the early stages of visual analysis which must precede reading. The domain
of reading itself, however, has various stages or components. All of the data
considered thus far have involved oral reading or report: letter naming and word
naming. Indeed, as Coltheart (1980) has noted, most studies of this syndrome
only report performance on oral reading. An issue which must be addressed is
whether the patients might be reading letter-by-letter merely in order to “name”
words. That is, might there be some specific difficulty of access from orthographic information to a phonological output system? If this were the case, then
the patients might comprehend printed words with relatively normal speed even
if they can only read them aloud very slowly or indeed not at all.
This question takes on added saliency due to recent interest in lateralisation of
language functions. By classical views (e.g. Benson and Geschwind, 1969),
*In all of our tachistoscopic tests we used a fixation point just to the right of where the stimulus
would appear, to try to ensure that the stimulus would be entirely within the patient’s intact visual
field. The wisdom of this procedure is debatable, as hemianopic patients may, to a variable extent,
adapt their fixation to accommodate to the hemianopia. Our use of a fixation point seemed to be
satisfactory; but we acknowledge that alternative procedures (such as providing two points or
vertical bars and instructing the patient that the stimulus will appear between them) may be equally
good or better.
?We are grateful to Dr Arnold Wilkins for raising this possibility, and to Dr Ronald Zimmern
for testing M.W.’s contrast sensitivity.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

422

K. PATTERSON AND J. KAY

language functions in a right-handed individual are represented only in the left
hemisphere. Most letter-by-letter readers have a right homonymous hemianopia,
which means in principle that visual information is received only by the right
hemisphere. According to the classical view, then, any reading skill achieved by
such a patient will require and reflect processing by the left hemisphere of information transferred from the right hemisphere. Recently, however, it has been suggested that, even in the typical right-handed person, certain aspects of language
function may be more bilaterally represented. In particular, the notion is that
while phonology for a written word is the exclusive province of the left hemisphere,
comprehension of a written word (at least if the word has a concrete, imageable
referent) may be subserved by processes represented in both hemispheres (see
Coltheart, 1980, for a review).
Much of the evidence for this claim comes from studies of split-field presentation
either with normal readers (e.g. Day, 1977; Marcel and Patterson, 1978) or with
commissurotomised patients (e.g. Gazzaniga and Hillyard, 197 I ; Levy and
Trevarthen, 1977; Zaidel, 1978). Yet neither of these populations represents an
ideal basis for such a conclusion. Normal readers, of course, have intact brains,
and it is difficult to be certain that the relevant processing is limited to the right
hemisphere. The arguments depend crucially on assumptions regarding visual
fixation by the subjects, the time required for interhemispheric transfer, etc. The
commissurotomy data on this question seem almost entirely inapplicable to conclusions about the “typical” right hemisphere : most of the split-brain patients had
neurological problems in early life which might well have resulted in abnormal
lateralisation of function. As Coltheart (1980) has observed, it is more legitimate
to draw implications from the disabilities in right-hemisphere processing by splitbrain patients than from their abilities.
Perhaps the most appropriate population with which to assess the issue of
comprehension in the right hemisphere is patients who were neurologically normal
until, as adults, they suffered a CVA affecting left-hemisphere language areas. In
analysing one such category of patients, namely deep dyslexic readers, both
Coltheart (1980) and Saffran et al. (1980) have argued for a “right-hemisphere
explanation”. These authors claim that the characteristics of deep dyslexia are
compatible with the notion that much of the patients’ reading performance-and
in particular their substantial comprehension of written words-is occurring in
the right hemisphere. It seems to us that such an argument entails the strong
prediction that letter-by-letter readers, with intact right hemispheres, should also
comprehend written words. It is obvious that the patients do understand a
written word once they can read it aloud, having worked it out letter-by-letter.
The prediction from the right-hemisphere hypothesis of deep dyslexia, however,
is much stronger. I n temporal characteristics, the letter-by-letter reader’s comprehension should significantly precede his slow oral reading response ; in terms of
accuracy, he perhaps ought to understand a written word even if he never manages
to read it correctly. This prediction may be restricted to concrete, imageable
words.
Although there are a number of anecdotal comments regarding this question,
we know of only three papers with relevant experimental observations. Two of

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

423

these (Warrington and Shallice, 1980;Landis et al., 1980)involved patients similar
to the ones presented here; however, these two publications provide only a small
amount of data and reach differing conclusions. T h e third paper (Albert et al.,
1973) involved a patient with a somewhat different pattern of deficit. We present
first our own observations, following which we will relate them to the results of
these previous studies.
The first of four techniques which we used to address this question was lexical
decision performance with exposure durations at or below I s. These durations
will generally not permit oral report of words and non-words by the patients in
this study; but, by the hypothesis under consideration, such durations might well
support the ability to discriminate between words and non-words.
Two of the patients, C.H. and M.W., performed with fair accuracy (though
very slowly and never without some errors) in uizpaced lexical decision tests given
short letter-strings. As soon as the items in a lexical decision test exceeded about
five letters in length, both patients began to show high error rates. Thus for the
tachistoscopically presented version with these two patients, we used only threeletter strings. The words were 20 high-frequency concrete nouns like dog, arm,
etc. ; the non-words were 20 CVC nonsense syllables like nad and hin. T h e letter
strings were presented in a two-field tachistoscope with a fixation point just to the
right of the right-hand edge of the stimulus. The exposure duration for M.W.
was 250 ms. For C.H., having tried 250, 500 and 750 ms (at all of which he complained that he could not even see a stimulus), we used an exposure duration of I s
(where he said that he was at least aware of an unidentifiable something). Performance for both patients, shown in Table 111, was essentially at the level of
chance.
‘rABLE 111
Lexical decision performance on three-letter words and nonsense syllables by patients
C.H. and M . W.

C.H. (I s)
M.W. ( 2 5 0 ms)

Hit rate

FP rate

d’

0.60
0.70

0.50

0.26

0‘5.5

0.40

A more extensive investigation of lexical decision performance was undertaken
with ‘r.P., whose performance on unpaced tests was rather better than that of
C.H. or M.W. First, T.P. was given a 60-item lexical decision list (strings of
three to nine letters in length) over-the-desk. T h e words were nouns of high
concreteness but medium-low frequency, for example author, throat, enamel,
gentleman. T h e non-words were constructed by changing one letter in other
words of the same general type, e.g. Jlankel (from jannel), elgineer (from engineer),
thunder (from thunder), etc. In this unpaced version of the test, T.P. was correct
on 25/30 words and 25/30 non-words. T h e 50 items on which she had made
correct judgements were then prepared on tachistoscope cards, and presented to
her on four different occasions over a 6-week period. The exposure durations

K. PATTERSON AND J. KAY

424

TABLE
IV
Lexical decision for patient T.P. at varying exposure durations
~~~

FP rate

Hit rate

T.P.

Unlimited

0.83 (1.0)

0.13 (0‘00)

2.08 (>4)

I S
500 ms

0.72
0.76
0.72
0.60
0.77
0.87

0-4.8

0.63
0.66

200 ms

IOO ms

D.E. Unlimited
zoo ms

Downloaded by [Central Michigan University] at 06:21 22 October 2015

d‘

0‘52
0.56
0.68
0.13

0.27

0.43
-0’22
I
I

-87
‘74

Data for a deep dyslexic patient (D.E.) for comparison.

for these four tests, in the order in which they occurred, were 200 ms, 500 ms,
I s and IOO ms. Her performance on all versions of the test is shown in Table 1V.
Two different values are given for the case of unlimited exposure because while
she was of course perfect on these 50 items with unlimited exposure, a fairer
estimate is probably her performance on the original 60-item set from which the
50 were then selected for tachistoscope testing. Her ability to make lexical decisions at I s or 500 ms may have been slightly above chance; by 200 ms, an exposure
duration at which normal readers would be perfect, she was essentially at chance.
For the sake of comparison, the performance of a deep dyslexic patient, D.E. (see
Patterson, 1978, for description), in lexical decision on these same 50 items is also
shown in Table IV. With unlimited duration, his performance was roughly comparable to T.P.’s; but reducing the exposure to 200 ms produced only a small
decrement for D.E., whereas for T.P. the same manipulation brought performance
from adequate to chance.
None of the lexical decision tests provides support for a hypothesis of rapid
comprehension in letter-by-letter readers. One could argue, however, that lexical
decision is not the most appropriate test for the hypothesis. For a patient with a
deficit in identifying letters, the concatenation of real words and non-words which
are minimally different in orthography might make the test unusually difficult.
We proceed, therefore, to tests involving words only.
Our second technique is based on one used by Warrington and Shallice (1979)
with a different type of patient whose impairment they termed semantic access
dyslexia. A word is presented either tachistoscopically for a brief duration or over
the desk for a longer exposure, and the patient is asked to report the word. If he
does so correctly, the test proceeds to presentation of the next word. If he is
unable to report the word, he is asked to make a binary semantic judgement about
the word (for example, “was it an animal or a fruit?”). Correct judgements at a
level above chance would constitute evidence for comprehension.
Three 24-word lists were composed of 12 words in each of two categories. In
each list, the words from the two categories were balanced for word length and
approximately balanced for probability of occurrence as an instance of that category
(Battig and Montague, 1969). One list was animal names (e.g. pony, zebra,
monkey) and fruit names (e.g. pear, grape, banana); a second consisted of varieties

425

LETTER-BY-LETTER READING

Downloaded by [Central Michigan University] at 06:21 22 October 2015

of trees (oak, spruce, cypress, etc.) and fish (cod, plaice, sardine); the third was types
of cloth (wool, denim, jlannel) and substances used for seasoning or flavouring food
(mint, curry, mustard). Each randomised list was presented to T.P. tachistoscopically with an exposure duration of 2 slword. She was asked first to report what
she had seen; if she could not identify the word, she was given the two category
names (auditorily) and asked to judge or guess the category of the stimulus.
‘Table V shows the proportion of words reported and performance on semantic
classification. Only the list of tree and fish names seems to have resulted in
TABLE
V
T.P.’s correct report on words and her binary classification performance on words
not correctly reported; exposure duration= 2 s
Correct report Correct classification
Animal/fruit
Tree/fish
Cloth/seasoning

519 =0*56

I 5/24 =0.625

I I /24= 0.46
I I 124 =0.46

I

o/I 3 =0.77
6113 =0.46

better-than-chance performance. Even that, however, is probably more apparent
than real. When T.P. could not report a word, she almost invariably reported
the first one, two or three letters (sycamore+“I only saw an s”; tuna+“t,u, I
think”; halibut+“h,a,l”). Several weeks after this test, at which time she seemed
to remember little of it, we took the 10 items of the tree/fish list which she had
failed to read but correctly classified. We presented her (auditorily) with exactly
what she had reported to us (for sycamore, “s”; for tuna, “t,u”; etc.), and asked
her to guess if she thought this was the beginning of a tree name or a fish name.
She guessed correctly for 9/10. Indeed, given how well she was able to do this
sort of guessing, it almost becomes a problem to account for her generally poor
performance in the binary classification task! We do not see any compelling evidence here that she was comprehending the words which she could not report.
The same three lists of categorised words used in the second technique with T.P.
were used in our third technique with both K.C. and M.W. In this procedure,*
as each word is presented over the desk, the patient is required first and quickly
to make a binary judgement of some kind about the word, and only then is asked
to read the word. If he produces his reading response fairly quickly (which does
sometimes happen, especially when the patient has relevant category information
about the word), the word is discarded from the analysis. We are only interested
to know about classificationjudgements made at a time when the patient is unable
to name the word, as evidenced by the fact that he or she must then work it out
letter-by-letter. For those words which are subsequently read in a letter-by-letter
fashion, above-chance classification performance would once again argue that
comprehension can occur before or in the absence of oral reading for these patients.
The results with this technique for K.C. and M.W. are shown in the first two
*This procedure was suggested by Professor Max Coltheart.

K. PATTERSON AND J. KAY

426

TABLE
VI
Performance for patients K.C. and M . W. in binary classiJicationof words subsequent& read letter-by-letter

Downloaded by [Central Michigan University] at 06:21 22 October 2015

Animallfruit
Tree/fish
Cloth/seasoning

K.C.

M.W. (speaking)

M.W. (pointing)

519 =0*56
511 2 =0’42
7/11=0.64

8117 =0.47

10116=0.625

11/19=0.58
I 1/21~ 0 . 5 2

12/23=0.52

-

8118 =0.44

columns of Table VI. Although each list contained 24 words, the denominators
are always less than 24 because of discarding relatively quick reading responses.
None of the proportions in the first two columns of Table VI is notably above
chance. We also repeated this test with M.W. a month later using a slightly
different procedure. If one conjectures that the word might be understood
quickly by processes occurring in the patient’s right hemisphere, then perhaps we
are failing to observe this comprehension because we ask the patient to make a
verbal categorisation judgement, which will presumably require transfer of information from right to left hemisphere.* I n the second version of the task, M.W.
had two pictures in front of her. For the first list, she had photographs of a tiger
and of an apple; for the second, there were sketches of a tree and of a fish; for the
third, she had a photograph of some bolts of cloth (from a colour advert) and an
(actual) pair of salt-and-pepper shakers. As each word in a list was given to her,
M.W. looked at the word and then had to point to the appropriate picture or
object to make her classification judgement. Neither she nor the experimenter
said anything until after the judgement was made, and then she attempted to read
the word aloud. As shown in column three of Table VI, this procedure produced
no greater evidence of comprehension than did the verbal judgements.
I n case it is not abundantly clear, it should be emphasised that none of the
patients had any difficulty with such semantic classifications once they had read
the word aloud. An incorrect pre-reading judgement was invariably corrected
after the patient had identified the word in a letter-by-letter fashion.
It is still just possible that some information might be quickly available to the
patient, of a nature not sufficiently specific to permit these categorisation judgements. Using the third technique, we asked M.W. to make a more general type
of binary judgement. The list consisted of 60 words, 30 names of living things
(e.g. bird, doctor, wasp, sister) and 30 names of non-living things (e.g. tent, hanzmer,
flag, bottle). M.W.’s initial classification performance (living or not) on those
words which she subsequently read letter-by-letter was 17/3I =0.55 correct, which
is not significantly above chance. Although the procedure is not exactly parallel,
for comparison M.W. was also given a set of 73 line drawings of objects, some of
which represented living things (e.g. gout, porter) and some non-living things (tank,
teapot). She was asked to make a quick binary judgement for each picture (living
or not) and then to name it. She made correct judgements on 72/73 pictures.
Admittedly rather few (14) pictures were subsequently named either slowly or
*We thank Dr Morris Moscovitch for raising this issue.

427

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

incorrectly; but unlike the pattern with words, all of these 14 pictures were given
correct classification judgements.
T h e fourth and final technique used in our search for evidence of comprehension
was a version of the Stroop test (1938). Although there are varying accounts of
the locus of the Stroop effect, it must be partly attributable to comprehension of
the printed word, since incongruous colour words retard naming of the ink colour
(e.g. the word blue printed in red) but congruous colour-related words facilitate
naming of the ink colour (e.g. the word blood printed in red) (Dalrymple-Alford,
1972). M.W. was given four different types of list in all of which her task was
simply to name the colour of the ink in which the stimulus was drawn or printed.
( I ) I n the baseline condition, the stimulus was the outline of a small rectangle
divided into three vertical compartments ; this is non-alphabetic but in visual
complexity is not too different from a short word. (2) I n the congruent condition,
the stimulus was the word corresponding to the ink colour. (3) I n the interference
condition, the stimulus was the name of a colour different from the ink colour (e.g.
the word green printed in red ink). (4)In a condition meant to control for orthographic similarity to the words in condition (3), the stimulus was a word with the
same first few letters as the interfering colour name (e.g. greed in red ink or rest in
green ink). Four colours were used (red, green, blue and brown); all lists contained
12 items printed in a column, three items in each colour. Two different random
orders were prepared for each condition, and the test was given to M.W. on two
different occasions (three months apart) ; thus there are four observations per condition. M.W. was asked to go through each list, naming ink colours as quickly
as possible. The time required to complete a list of 12 items was measured to the
nearest second, and the values in Table VII represent means of the four observations in each condition.
TABLE
VII
M . W.’sperformance in a Stroop task: Time for list of 12items
Condition
Baseline
Congruent word
Interfering word
Orthographic control

Example

Time (s)

red
green
greed

13’5

m

15.0
15.0

16.0

*Response should always be ink colour (for example above “red”).

We see no indication in these results that M.W. had rapid access to the meaning
It is true that she appears to have shown an effect of congruency
(condition 2), but this is not especially surprising. These patients can usually get
the first letter or two of a word within the time span involved here, and having the
information r,e (or even just r ) could plausibly speed the response “red”. There
are two contrasts for which a hypothesis of comprehension would surely provide
specific predictions: condition (3) should have been both slower than ( I ) and slower
than (4). Neither of these results was obtained.
of these words.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

428

K. PATTERSON AND J. KAY

None of our techniques has yielded evidence to support the hypothesis that
comprehension can proceed normally when oral reading must proceed letter-byletter. It appears that these four patients must read a word letter-by-letter in
order to understand it, as well as to name it. This conclusion is the same as that
reached by Warrington and Shallice (1980)on the basis of tests with their case
R.A.V. There are, however, two other reports which claim to have found positive
evidence for comprehension.
Landis et al. (1980)describe a patient who, within the first few days following
removal of an occipital lobe tumour, could sometimes point to an object whose
name had been presented tachistoscopically at an exposure duration (30 ms) which
did not allow him to name even single letters. After these first few days, the
patient showed rapid “improvement” : several weeks later he could recognise some
letters when words were presented for only 20 ms, but apparently he could no
longer appreciate the meaning of the whole word without working it out letter-byletter. Thus at later testing, the patient studied by Landis et al, behaved like
our four patients. It is regrettable that we have no observations for our patients
in the first few days following their brain injuries. It is also regrettable that
Landis et al., while using an adequate technique to assess comprehension, obtained
such a small amount of data: seven words were presented which the patient could
not read, and he pointed to the correct object for 517. In view of the contrast in
extent of testing, we are inclined to claim that our conclusion is the more reliable.
Nonetheless, the interpretation suggested by Landis et al. is of interest.
Landis et al. argue that “iconic reading” (as they call the processes involved
when the patient rapidly comprehended a word which he could not name) can
only operate when “visuo-verbal reading’’ is totally impossible; even the limited
ability to identify some letters, a component of the visuo-verbal reading strategy,
prevents iconic reading. There is one other report which might be consonant
with this notion. Albert, Yamadori, Gardner and Howes (1973) described a
patient J.H. who, while very severely alexic in terms of oral reading, showed some
evidence of comprehending written words when tested a month after his CVA.
We hesitate to place too much evidence on these contrasting data because J.H.
showed a pattern of deficit rather different from our patients and from most letterby-letter readers. For example, while he had some preservation of oral spelling,
he was almost totally agraphic in written spelling. Indeed, although many of the
features of his performance evoke the picture of a severely impaired letter-by-letter
reader, it is difficult even to apply this label to J.H. with confidence, because (a)
he accomplished so little oral reading, and (b) Albert et al. provide so little description of his oral reading performance, mostly saying simply that he failed. Nonetheless, we acknowledge that J.H. may constitute a counter-example to our
conclusion. If the availability of a “visuo-verbal” reading strategy can be assessed
by letter identification, then J.H. (who could only name 10 to 20% of letters)
certainly represents a much closer approximation to the loss of this strategy than
do our cases.
The implication that one might draw, therefore, from both Landis et al. (1980)
and Albert et al. (1973)is as follows: as impaired as our four patients were at oral
reading, they would be expected to show better written-word comprehension only

429

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

if their oral reading were still worse. We find this argument intriguing but somewhat implausible on intuitive grounds. Is it sensible to have an iconic reading
process which will operate only when visuo-verbal reading is abolished and not
when it is grossly deficient? We also note that, to the extent that iconic reading
is to be equated with right-hemisphere comprehension (as suggested by Landis
et al.), the notion that this process can occur only in the absence of visuo-verbal
reading causes severe difficulty for claims of right-hemisphere word comprehension
in normal readers (e.g. Day, 1977; Marcel and Patterson, 1978).
I n concluding this section, we do not deny the possibility that letter-by-letter
oral reading might, for some cases or at some stages, co-exist with rapid word
comprehension. We found no evidence for this in our four patients, however.
We therefore challenge theorists who maintain that the normal right hemisphere
can support rapid comprehension of written words to explain why many letter-byletter readers fail to demonstrate this ability.

111. From the letters to the word
Knowing that the letter-by-letter reader begins his process of word recognition
by identifying a sequence of letters does not of course specify how he recognises
the word. As we phrased the question previously (Patterson and Kay, 1980),
how do word-form dyslexics form words? We had hoped to approach an answer
to this question by analysing the kinds of reading errors made by these patients.
This error analysis does not, regrettably, provide a definitive answer, but it does
establish a new and intriguing dimension to the problem: the procedure from
letters to word may take two different forms.
For each of our four patients, we analysed a set of reading responses from the
same word lists which produced the word length functions in Table I, namely the
Schonell graded list (1942),the Coltheart et al. list of regular and exception words
(I979), plus some others. We do not consider the exact characteristics of the
words very important because, apart from length, the dimensions of words have
little effect on the probability of correct reading for these patients. As can be
seen in Table VIII, the size of the corpus varies from 170 to almost 900 reading
responses across the four patients, and the proportion of correct responses varied
from 0.65 to 0.82. (We counted a word as correct if the patient eventually read
it correctly after some incorrect attempts ; these proportions would be noticeably

TABLE
VIII
Analysis of reading urors

Number of responses
Proportion correct
Proportions of error types :
( I ) Clear letter misidentification
(2) Letters right, word wrong
(3) Other

M.W.

C.H.

T.P.

1c.c.

317

229

893

0.82

0.65

0.76

0.69
0'21

170

0.40

0.69

0.36

0.03

0'04

0.42

0'49

0'57

0.27

0'22

0.30

Downloaded by [Central Michigan University] at 06:21 22 October 2015

430

K. PATTERSON AND J. KAY

lower if one accepted first responses only.) The errors can largely be assigned to
three categories. Firstly, “clear letter misidentification” usually means that the
patient has produced the word or pronunciation exactly corresponding to an
incorrect sequence of spoken letter names (e.g. C.H. : men-+“h,e,n; hen”; K.C. :
spade+“s,h,a,d,e ; shade”). We have also accepted instances where the patient
did not say the letter names audibly but where it seems reasonably clear that a
specific letter confusion would yield the response (e.g. C.H. : made-+“wade” ;
T.P. : sort-+“soft”). The second category, “letters right, word wrong”, typically
means that the patient has named all letters of the word correctly but then has misnamed the word producing either a neologism (e.g. T.P. : slap+“s,l,a,p; slape”
/sleIp/) or a different word (T.P.: head-+“h,e,a,d; heed”). Once again, we also
accepted instances where it seemed clear to us that this process was occurring even
if all of the letter names were not spoken aloud (e.g. T.P. : knowZedge+“kno-ledge”
/knouledg/ ; K.C. : castle+iicast-lee”/kastli/). Finally, we have that famous error
category “other”. Most of these reflect instances where the patient seems to have
got near rather than to the end of the letter sequence and then produced a word
which matched the target on some number of initial letters (e.g. M.W. :$owing-+
“f,l,o,w . . . flower”; crowd+“crow”). The “other” category also includes any
responses where we did not feel certain about the underlying process. M.W.’s
response such+“suede”, for example, may have involved one or more letter misidentifications, but it does not seem sufficiently obvious.
Differences amongst the patients can be found in all of the rows of Table VIII;
but we think that only one of these differences is important. The error category
“letters right, word wrong” divides the patients into two groups of two, and indeed
we propose that M.W. and C.H. are one kind of letter-by-letter reader and T.P.
and K.C. another. M.W. and C.H. produce many incorrect reading responses
because they often (C.H. especially) misidentify letters and often (M.W. especially)
guess at a word without explicitly having identified its last few letters. If these
two patients correctly identify all of the letters in a word, however, they will
correctly produce the word. Each of them produced a few instances of “letters
right, word wrong”, but we believe that they did this no more frequently than
normal readers would do when identifying words from sequences of letter names.
Both the Schonell and the Coltheart et al. lists contain many words with difficult,
irregular spellings, like yacht and colonel and gnome. It is not the case that M.W.
and C.H. read all of the irregular words correctly (though both were correct on
approximately equal proportions of regular and irregular words) ; but if they
identified a word’s component letters correctly, then moving from letters to word
for these two patients was quite straightforward, even for words with difficult
spelling.
T.P. and K.C. behaved quite differently, making a significant proportion of reading errors in category (2). The majority of the errors are reminiscent of those
made by surface dyslexic patients (Marshall and Newcombe, 1973 ; Holmes, 1978;
Marcel, 1980); that is, the response represents a pronunciation of the graphemic
string which, although lexically incorrect, is either phonologically plausible or
nearly so. Examples of these errors appear in Table IX. Both patients (but
K.C. more often than T.P.) also have responses in this category which are not so

431

LETTER-BY-LETTER READING

TABLE
IX
Examples of “letters right, word wrong” for patients T.P. and K.C.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

Exact reylarisation
Violation of final e rule
“Naming” the vowel
Silent letters pronounced
Hard/soft consonant
Voicing problem
Vowel digraph problem

T.P.

K.C.

ache-J“aitch”/eItJ/
whiZe+“will”
back+“bake”
mord+“sword)’/sw3d/
city+“kitty”
whose+“ hooce”/hus/
group+“grope”

come+“comb”
mile+“mill”
pin+“pine”
tight+“titched”/trtJt/
angel+“angle”
this+“thiz”/eIz/
dueZ+“dull”

easily described as simple errors in segmenting a letter string or assembling a
phonological code for it: T.P. :fact+“f,a,c,t ;fought’’;K.C. : thirsty+“t,h,i,r,s,t,y ;
curtsy?”; quite+“q,u,i,t,e; quilled?”/kwIld/“quot?”/kwot/;
think-+“t,h,i,n,k; I
can’t think what it is”.
All four patients in this study often made a series of attempts at a word, trying
to find a satisfactory response. The interesting difference between M.W. and
C.H., on the one hand, and T.P. and K.C., on the other is that the former patients
appeared to be trying out words while the latter were trying out pronunciations.
This difference may be illustrated by the following examples. C.H. was given
the word gone to read, and he began, as so often, with some letter misidentifications :
“b,o,r,e; bore; isn’t that an r?” (We told him it was not.) “Oh, it’s an n!
That’s a bone! Wait . . . done, d,o,n,e. Is that supposed to be a c? Cone?
G,o,n . . . g,o . . . gone! That’s it, gone.” The correct pronunciation of the
vowel of course changes from done to cone to gone, but this simply was not an
issue for C.H. His pronunciation of each word was lexically correct ; the problem
consisted merely in identifying the letters. T.P. also made many letter misidentifications in reading; but superimposed upon this deficit was a problem in assigning
pronunciations to words. We cannot use the example of gone because when T.P.
was given this word, she correctly identified all of its letters (and then pronounced
it “Joan”/d3oun/). Her sequence of attempts to read bloom, however, reveals a
series of letter misidentifications plus this pronunciation problem : bloom-+
“d,r,o,o,m ; drome”/droum/; “no, that’s a b, brome”/broum/; “no, blome”/bloum/.
By saying that C.H. and M.W. were trying out words and T.P. and K.C. pronunciations, we may have inadvertently given the impression that the latter
produced neologisms but the former did not. This is not the case. If C.H. or
M.W. misidentified a letter such as to create a non-word, then they would produce
the appropriate non-word (C.H. : $ood+“f,r,o,o,d; frude”/frud/). Our point is
simply that while all four patients might well start to readpood by saying “f,r,o,o,d;
frude”, once having corrected the Y to an I, C.H. and M.W. would both correctly
pronounce “flood” while T.P. and K.C. might well say “flude”/flud/. We should
note, by the way, that although we did not do formal investigations of this, we
feel confident from their behaviour and their comments that all four patients recognised their neologisms as such. Many of the sequences of responding probably
arose because the patient was dissatisfied with a neologistic response.

K. PATTERSON AND J. KAY

432

Downloaded by [Central Michigan University] at 06:21 22 October 2015

As we see it, there are two possible interpretations of the difference between our
two “types” of letter-by-letter reader. Deciding between these interpretations
will require (a) a well specified process-model of reading and spelling and (b) a
very thorough analysis of the performance of letter-by-letter readers. Since we
do not feel that we can adequately fulfil either of these requirements, we shall
present both interpretations and shall not attempt here to decide between them.
Purely to assist our description, we present Figure I which is based almost entirely
on the model in Shallice (1981).* The two accounts are as follows.

I

Early visual
processing

Letter identification
(explicit)

(abstract, parallel)

*

.@‘
&.4

1
I

word-form syrtem

I

I
I

I

I
I

t

Writing

FIGUREI. A representation of some functional components involved in recognising, understanding and pronouncing written words, based on the model in Shallice (1981).
T h e dotted line consisting of I(a) plus I(b) corresponds to the first account of letter-by-letter
reading, while the dotted line labelled z corresponds to the second account (see text).

(I) Letter-by-letter reading arises from impairment of the visual word-form
system which (in the normal reader) recognises or categorises written letter strings.
A sequence of explicit letter names or identities permits word recognition via processes which belong to the spelling system. The reason that we have two types of
letter-by-letter reader is that in fact we have two types of speller. C.H. and M.W.
*Since the word-form system has units corresponding to individual letters as well as to letter
combinations, morphemes, etc., the reader might question the motivation for a separate, prior
system for letter-form analysis (indeed, an astute reviewer did question this). T h e function of
letter-form analysis is essentially perceptual, while the purpose of the word-form system is to
identify higher-level orthographic units with semantic and/or phonological counterparts. Individual letters are represented in the word-form system only as a default option, in case the system
fails to find larger units; this would rarely happen in a normal reader‘s normal experience. Furthermore, the operation of the word-form system is predominantly intra-word, constrained by the
presence of a space at either end. “Parallel” as a description of the letter-form system applies not
only to analysis within words but also across words (Shallice, 1981). Thus the operation of the
letter-form system accounts for the kind of segmentation errors in tachistoscopic report of multiword arrays (e.g. if mst and vent have been presented, the subject might report “vest”) which have
been documented by Allport (1977) and Shallice and McGill(1978).

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

43 3

have reasonably normal lexical knowledge of spelling, and so they can correctly
identify and assign lexical pronunciations to words once they have identified the
component letters. T.P. and K.C., on the other hand, have a neurologically
acquired impairment of lexical spelling. When they try to use this deficient spelling system in reverse for word identification, not surprisingly it produces errors.
This, as we understand it, would be Warrington and Shallice’s (1980) account of
letter-by-letter reading applied to our patients.
(2) Letter-by-letter reading arises because of a disconnection between the
letter-form analysis system and the visual word-form system. Although the wordform system is primarily suited to operating on abstract letter identities in parallel,
it can also accept a sequence of letter identities. If this disconnection is the only
deficit, one will find the pattern shown by “classical” letter-by-letter readers such
as C.H. and M.W. However, in some cases there is a further impairment in the
recognition system itself; in models like those proposed by Marcel (1980) and
Shallice (1981) (and the one presented in Figure I), this impairment would correspond to whole-word or whole-morpheme specifications in the visual word-form
system. The patient with this additional deficit would be forced to assign phonology to sub-word-sized graphemic segments and then cobble together a pronunciation. The result would be many reading errors characteristic of surface
dyslexic patients (whose reading has been attributed to this kind of impairment ;
see Coltheart, 1978; Marcel, 1980; Shallice, Warrington and McCarthy, 1981),
and this is of course the pattern shown by T.P. and K.C. This second account is
the one which we offered in a preliminary presentation of results for two patients,
C.H. and T.P. (Patterson and Kay, 1980).
Our reasons for suggesting this alternative to Warrington and Shallice (1980)
were somewhat intuitive.
( I ) Watching C.H. and M.W. read, it seemed to us that enormous effort was
required for them to identify letters; but, once that had been achieved, moving
from letters to word was virtually automatic. It “felt” as if word recognition was
being handled by an efficient system designed for the purpose, which would
operate normally if only it could get proper input.
(2) The distinction between C.H. and M.W. as contrasted with T.P. and K.C.
was of course evocative of the popular contrast between two “routes” in reading
(e.g. Coltheart, 1978), and this perhaps naturally inclined us towards explanations
in terms of components of the reading system.
(3) One of us (K.P.) had been doing some work with deep dyslexic patients,
asking them to identify words presented visually, one letter at a time. Their
ability to report words under these conditions, while slightly poorer than their
performance in reading words presented normally, is rather impressive. This
might be a surprising result in view of the fact that deep dyslexic patients have a
severe deficit (a) in naming letters and (b) of short-term memory (Shallice and
Warrington, 1975). What seems certain is that the procedure by which deep
dyslexic patients recognise words presented one letter at a time must involve the
reading system and visual input logogens or word-form units. The patients
cannot be using letter names because they are so poor at naming letters; and they
cannot be using a spelling system because most deep dyslexic readers (and cer-

Downloaded by [Central Michigan University] at 06:21 22 October 2015

434

K. PATTERSON AND J. KAY

tainly the ones who participated in these experiments) are much more severely
impaired in spelling than they are in reading (Patterson, 1978). If the word
recognition system in a deep dyslexic patient can accept a sequence of letter identities, might not letter-by-letter reading be achieved in the same fashion?
(4) Although the general alignment between patterns of reading and spelling
in the four patients is a compelling argument in favour of Warrington and Shallice,
there are discrepancies which seem difficult for this interpretation. In a study of
T.P.’s writing and spelling (Patterson and Hatfield, 1980), we asked her to read a
large number of words which she herself had earlier written to dictation. She
“correctly” read many words which she had misspelled (e.g. “most”+moast+
“most”), but sometimes she was unable to make any sense of her own spelling.
Given “world” in dictation, she wrote wirlde; given wirlde to read, she said “I
cannot believe that is anything”. If reading, for a letter-by-letter reader, involves
the spelling process in reverse direction, how does this kind of event occur?
( 5 ) Finally, we felt mildly dissatisfied with the account offered by Warrington
and Shallice (1980) simply because it was not well specified. They were clear
enough about what they argued to be deficient in letter-by-letter reading (the wordform system), but they provided no description of the actual process of reading
via spelling.
At least some of points ( I ) to ( 5 ) above may require attention. We must, however, admit that there are persuasive reasons for accepting, at least in broad outline, the position taken by Warrington and Shallice.
( I ) The parallels between reading and spelling in these patients cannot be
ignored. All four patients read letter-by-letter; but T.P. and K.C. who make
“lexical” spelling errors also make “lexical” reading errors, while M.W. and C.H.
do not make these errors in either spelling or reading. Patterns of reading and
spelling can, in other patients, show major dissociations (e.g. Beauvois and
DPlrouesnC, 1979, 1981); therefore when these processes reveal strikingly similar
patterns, the notion that one is parasitic on the other must be entertained.
(2) The Warrington and Shallice account is more parsimonious than ours, not
for the classical cases (like M.W. and C.H.) but for the more complicated cases of
T.P. and K.C. The interpretation by Warrington and Shallice requires only two
deficits for T.P. and K.C. :impairments in the word-form system and in the lexical
spelling system will account for both reading and writing problems. Our interpretation requires three deficits : impairments in the transmission from letter-form
analysis to word-form system and in the word-form system itself will explain the
reading problem, but we then need an additional impairment to account for the
spelling problem.
(3) Our account is, in some respects, no better specified than theirs. I n
particular, we have not explained why the disconnection between graphemic
analysis and word-form system forces a serial, sequential use of letters. Warrington
and Shallice have not explained this feature either; but proceeding through a set of
letters in a discrete sequential manner does seem more characteristic of spelling and
writing than of reading.
Choice of one of these two accounts, or of yet a third, awaits better evidence and
better modelling.

LETTER-BY -LETTER READING

435

Downloaded by [Central Michigan University] at 06:21 22 October 2015

IV. Impaired letter identiJication
Neither of the attempts (in section 111) to explain why and how these patients
read letter-by-letter says anything about why they misidentify letters. Warrington
and Shallice (19So)may not have considered this issue because the majority of
their observations come from a letter-by-letter reader (R.A.V.) who apparently had
no deficit in identifying letters.* A survey of some of the literature on letter-byletter reading suggests, however, that R.A.V. is atypical; most patients of this type
do have a deficit in identifying letters (e.g. Stachowiak and Poeck, 1976;Staller et
al., 1978). Since there are patients like R.A.V., letter-by-letter reading cannot be
a result of difficulty in identifying letters. The nature of this difficulty and its
effect on reading does, however, seem to warrant comment.
It was our vague impression that all of our patients misidentified letters more
often in reading words than in naming individual letters. We therefore did some
tests to assure ourselves that these patients were not suffering from attentional
dyslexia (Shallice and Warrington, 1977),a syndrome where symbols like letters
are harder to identify when other similar elements are present in the visual field.
we were able to rule out attentional dyslexia
Like Warrington and Shallice (I~SO),
as an important component of letter-by-letter reading, by showing that the patients’
naming of random letters in an array was comparable in accuracy to their naming
of letters in isolation. I n any case, however, it was not so easy to validate our
initial impression about higher rates of letter misidentifications in words. Firstly,
while something like 85% correct in naming individual letters may seem not too
serious an impairment, even that rate of errors will be very noticeable and disruptive in reading words. Secondly, the tasks are of course different. If one is
attempting simply to name letters, whether in isolation or in an array or in a word,
the identification of a letter is an end in itself; identifying letters in order to recognise the word which they compose may be procedurally rather different, involving,
for example, a memory load. Finally, the process for these patients of arriving at
a word from a letter sequence undoubtedly involves use of knowledge about constraints of English orthography. Initial hypotheses about a letter’s identity may
be rejected because the patient knows that this creates an impossible letter sequence.
Sometimes this happens aloud (e.g. K.C. : both--+“b,o,f,h, no, it can’t be . . .”), but
the rejection must often occur before the letter is spoken, making it harder to
observe. Thefinal response to a word in fact probably reflects a lower rate of letter
misidentifications than occurs in naming random letters.
An analysis of the letter misidentifications which occurred in word reading ought
to provide information germane to two questions. Firstly, are the misidentifications consistent, such that a predictable subset of letters causes difficulty, or is the
process of identification just unstable, resulting in a changing subset which are
vulnerable? T h e answer to this, with the usual clarity of research answers, appears
to be :some of each. There is a specific subset of misidentifications which recurred,
in the sense that all four (or at least three of the four) patients produced the same
error and also that each patient produced multiple instances of it. There is also
*Warrington and Shallice (1980)do mention that R.A.V. made “very occasional errors” (p. 101)
in reading short words and non-words, but there is no further description of these errors.

K. PATTERSON A N D J. KAY

Downloaded by [Central Michigan University] at 06:21 22 October 2015

436

a larger set of confusions where we observed perhaps only one or a few instances.
Nonetheless, the number of different types of observed confusions is substantially
less than the total possible set. Collapsing the confusion matrix for direction
(that is, ignoring whether b-tp or p-tb, for example), 325 different confusion types
are possible ((26*-26)/2). Observed instances of clear letter misidentifications
included only 44 of these 325 types.
Secondly, can one discern any coherent basis for the confusions, in particular
visual similarity? As a first approximation to answering this question, we produced an ordering of the lower-case alphabet in rough accordance with degree of
visual similarity.* This ordering (which appears in Figure 2) is very close to the
one used by Bouma (1971) for confusion matrices in letter recognition by normal
subjects. The ordering is primarily meant to reflect clusters within which the
s a o c e u v w m n r i l t f k h b d p g q y j

+ mnriItfkhb-

d.

*+*

1

P‘
9qY’

FIGURE
2. A confusion matrix for lower-case letters, with letters sharing visual features placed
close together in the ordering. Clear instances of letter misidentifications by the four patients are
shown in the matrix, which is folded for direction (thus, confusion between a and o may mean
a+o or o+a but often both). Confusion types with a number of tokens are represented by the
symbol
in a circle; confusion types with only one or a few tokens are represented by .

+

+

similarity is rather clear, such as (I) the ascenders, (2) the descenders, (3) the
letters composed of a “loop on a stick” (b,d,p,g,q), (4) the “little roundish letters”
(a,o,c,e), etc. We have largely ignored both the problem of transitions between
clusters, and the fact that some letters could fit into more than one cluster (e.g. y
is a descender but is also somewhat similar to -a and u). We have also eliminated
x and x which do not occur very often and neither of which entered into any
*Almost all of the words which our patients read in these tests were printed in lower-case (on
the rationale that this is what they are most likely to encounter in real life). It is difficult to be
more explicit about specific features of the letters, since sometimes the words were typed, sometimes
they were printed in Letraset, and sometimes they were printed by hand.

437

LETTER-BY-LETTER READING

observed confusion. Figure 2 presents the collapsed 24 x 24 matrix based on this
ordering. Each of the 44 types of observed confusion provides an entry in the
matrix; types with many tokens (i.e. the small subset, referred to above, which
recurred both across and within patients) are indicated by the symbol
in a
circle; types with few tokens are indicated by a simple
The implication from Figure 2 is that visual similarity does play a prominent
role in the patients’ letter misidentifications. Although there are exceptions, the
majority of the confusions lie near the diagonal, suggesting that visual features of
letters may be a major factor in these misidentifications. Also consonant with this
suggestion is the fact that the patients’ letter misidentifications (in words and under
conditions of adequate visual stimuli) show reasonable overlap with errors of normal
subjects asked to identify single letters either at a distance or in eccentric vision
(Bouma, 1971). For example, the eight confusions for which we observed many
instances with our patients are all represented in the errors of normal subjects.
Seven of these eight types occurred in both of Bouma’s conditions (distance and
eccentricity), while one type (confusion between b and p ) was shown by normal
subjects only with eccentric presentation.
It may be of interest that only six of the patients’ ++observed types of letter confusions crossed the consonant/vowel boundary ; all of the others involved substitution of one consonant for another or of one vowel for another. Thus there
appears to be some mild degree of category specificity in identifying letters in
words. Whether this has anything to do with physical features of letters, or
whether it merely arises from the patients’ knowledge of constraints in English
orthography, is an open question.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

+.

+

Concluding comments
By no means all of the interesting features of letter-by-letter reading have been
addressed here, and some features (like the process of letter identification itself)
have only received cursory attention. Our major purpose has been three-fold.
Firstly, while the preponderance of reports on letter-by-letter reading have
emphasised neuro-anatomical disconnection between brain centres, our interest
was in a psychological level of description. A neurological level of description
might seem satisfactory as an account of a patient’s failure to read (i.e. this or that
brain centre, which is essential for reading, has been damaged) but is considerably
less appropriate as an account of how the patient accomplishes what reading he
does manage. Procedures of reading, whether normal or impaired, require a
procedural description.
Secondly, while most of the available data regarding this syndrome reflect oral
reading, one of our primary interests concerned reading comprehension. There
are at least two motivations for a careful assessment of word comprehension in
letter-by-letter reading. (a) Conclusions about impaired and intact functional
components in this syndrome require comprehension data. Two potential alternative outcomes-that letter-by-letter analysis must precede both comprehension and
oral reading for these patients, or that it precedes oral reading but is not necessary
for comprehension-entail different interpretations. The latter outcome, had it

Downloaded by [Central Michigan University] at 06:21 22 October 2015

438

K. PATTERSON AND J. KAY

occurred, would have indicated that both graphemic access to the visual word-form
system and the word-form system itself were intact. T h e actual outcome, that
the patient’s comprehension of a word also requires letter-by-letter reading, indicates that either the word-form system or normal access to it must be disrupted.
(b) Comprehension data for letter-by-letter readers also address the more general
hypothesis of word comprehension in the right hemisphere of a person with normal
(i.e. left-dominant) language lateralisation. While our results cannot invalidate
claims for right-hemisphere comprehension in deep dyslexic patients (Coltheart,
1980; Saffran et al., 1980) or in normal readers (Day, 1977; Marcel and Patterson,
1978) or in some letter-by-letter readers (Landis et al., 1980), the results do make
such claims more difficult and/or require more tortuous versions of the hypothesis.
Finally, while sequential letter identification characterises all letter-by-letter
readers (by definition !), this procedure begins but certainly does not complete the
process of word recognition by these patients. The most interesting question, we
think, concerns how the information in a sequence of letter names or identities is
used for word recognition. The process of recognising a printed word for these
patients appears to be rather like the process of identifying an orally spelled word
for a normal reader ; but this analogy scarcely helps because little is known about
the procedures involved in recognising orally spelled words. T h e main contribution of our data has been the demonstration that, depending on other characteristics of the patient, a sequence of correctly identified letters may or may not be
sufficient for correct lexical recognition of the word. I n “classical” or “pure”
letter-by-letter readers like our patients M.W. and C.H., word recognition only
requires a correct sequence of letters. Patients like T.P. and K.C., however, have
a further lexical deficit such that even a correct sequence of letters may fail to
provide access to the target word. Different interpretations place this lexical
deficit in (a) the system which deals with words in reading or (b) the system which
deals with words in spelling. Future data (for example, the finding of either continued association or of a new dissociation between patterns of reading and spelling
errors in letter-by-letter readers) will determine which interpretation is more
appropriate, and thus whether the syndrome of letter-by-letter reading is more
informative about reading or spelling.
For permission to study these patients and to publish our observations, we are grateful
to Mr J. W. Gleave and Mr A. E. Holmes (Department of Neurological Surgery and
Neurology, Addenbrooke’s Hospital, Cambridge), Dr J. R. Jenner (Rehabilitation Unit,
Addenbrooke’s Hospital, Cambridge) and Dr M. Small (Department of Neurology, Queen
Elizabeth Medical Centre, Birmingham).
For helpful discussions and comments upon the manuscript, we thank Max Coltheart,
Andrew Ellis and Tim Shallice.
This paper was written while the first author was on study leave at the Institute for
Perception Research, Eindhoven, The Netherlands.

References
AJAX,E. T., SCHENKENBERG,
T. and KOSTELJANETZ,
M. (1977). Alexia without agraphia
and the inferior splenium. Neurology, 27, 685-8.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

439

ALBERT,
M. L., YAMADORI,
A., GARDNER,
H. and HOWES,D. (1973). Comprehension in
alexia. Brain, 96, 317-28.
ALLPORT,
D. A. (1977). On knowing the meaning of words we are unable to report: the
effects of visual masking. In DORNIC,S. (Ed.), Attention and Performance V I .
Hillside, N. J. : Lawrence Erlbaum.
BATTIG,W. F. and MONTAGUE,
W. E. (1969). Category norms for verbal items in 56
categories : a replication and extension of the Connecticut category norms. Journal
of Experimental Psychology Monograph, 80, No. 3, Part 2.
BEAUVOIS,
M. F. and DEROUESNB,
J. (1979). Phonological alexia: Three dissociations.
Journal of Neurology, Neurosurgery, and Psychiatry, 42, I I 15-24.
BEAUVOIS,
M. F. and DEROUESNE,
J. (1981). Lexical or orthographic agraphia. Brain,
104,21-49.
BENSON,
D. F. and GESCHWIND,
N. (1969). The alexias. In VINKEN,P. J. and BRUYN,
G. W. (Eds), Handbook of Clinical Neurology, Volume 4. Amsterdam :North-Holland.
BISHOP,D. V. M. (1980). Comprehension of complex sentences by language disordered
children. Paper presented to the Experimental Psychology Society, Cambridge, July,
1980.
BOUMA,
H. (1971). Visual recognition of isolated lower-case letters. Vision Research, 11,
459-74CAPLAN,
L. and HEDLEY-WHITE,
T. (1974). Cueing and memory dysfunction in alexia
without agraphia. Brain, 97, 251-62.
G. (Ed.),
COLTHEART,
M. (1978). Lexical access in simple reading tasks. In UNDERWOOD,
Strategies of Information Processing. London : Academic Press.
COLTHEART,
M. (1980). Deep dyslexia : A right-hemisphere hypothesis. In COLTHEART,
M., PATTERSON,
K. and MARSHALL,
J. C. (Eds), Deep Dyslexia. London: Routledge
and Kegan Paul.
COLTHEART,
M. (1981). Disorders of reading and their implications for models of normal
reading. Visible Language, 15, 245-86.
COLTHEART,
M., BESNER,
D., JONASSON, J. T. and DAVELAAR,
E. (1979). Phonological
encoding in the lexical decision task. Quarterly Journal of Experimental Psychology,
319489-507.
COUGHLAN,
A. K. and WARRINGTON,
E. K. (1978). Word-comprehension and wordretrieval in patients with localised cerebral lesions. Bruin, 101,163-85.
DALRYMPLE-ALFORD,
E. C. (1972). Associative facilitation and interference in the Stroop
colour-word task. Perception and Psychophysics, 11, 274-6.
DAY,J. (1977). Right-hemisphere language processing in normal right-handers. Journal
of Experimental Psychology: Human Perception and Performance, 3, 5 18-28.
DBJERINE,J. (1892). Contribution A 1’Ctude anatomo-pathologique et clinique des
diffkrentes variCtCs de cCcitC verbale. Compte Rendu Hebdomadaire des SCances
et Memoires de la SociCtC de Biologie, 4,61-90.
FREDERIKSEN,
J. R. and KROLL,J. F. (1976). Spelling and sound: Approaches to the
internal lexicon. Journal of Experimental Psychology: Human Perception and Performance, 2, 361-79.
GARDNER,
H. and ZURIF,E. (1975). Bee but not be: Oral reading of single words in aphasia
and alexia. Neuropsychologia, 13, 181-90.
GAZZANIGA,
M. S. and HILLYARD,
S. A. (1971). Language and speech capacity of the right
hermisphere. Neuropsychologia, 9, 273-80.
GESCHWIND,
N. (1965). Disconnection syndromes in animals and man, Part 11. Brain,
889585444.
GREENBLATT,
S. H. (1976). Subangular alexia without agraphia or hemi-anopsia. Brain
and Language, 3,229-45.
HECAEN,
H. and KREMIN,
H. ( I 976). Neurolinguistic research on reading‘disorders resulting
H. and
from left hemisphere lesions: Aphasic and “pure” alexias. In WHITAKER,
H. A. (Eds), Studies in Newolinguistics, Volume 2. New York: Academic
WHITAKER,
Press.

440

K. PATTERSON AND J. KAY

Downloaded by [Central Michigan University] at 06:21 22 October 2015

A. and
HOLMES,
J. M. (1978). ‘‘Regression” and reading breakdown. In CARAMAZZA,
ZURIF,E. (Eds), Language Acquisition and Language Breakdown: Parallels and Divergences. Baltimore :Johns Hopkins University Press.
KINSBOURNE,
M. and WARRINGTON,
E. K. (1962). A disorder of simultaneous form perception. Brain, 85,461-86.
KURACHI,
M., YAMAGUCHI,
N. INASAKA,
T. and TORII,
H. (1979). Recovery from alexia
without agraphia: Report of an autopsy. Cortex, 15,297-312.
LANDIS,
T., REGARD,
M. and SERRAT,
A. (1980). Iconic reading in a case of alexia without
agraphia caused by a brain tumour: A tachistoscopic study. Brain and Language,
11, 45-53.
LEVY,
J. and TREVARTHEN,
C. (1977). Perceptual, semantic and phonetic aspects of elementary language processes in split-brain patients. Brain, 100, 105-18.
MARCEL,
T. (1980). Surface dyslexia and beginning reading: A revised hypothesis of the
K. and
pronunctiation of print and its impairments. In COLTHEART,
M., PATTERSON,
MARSHALL,
J. C. (Eds), De@ Dyspexia. London: Routledge and Kegan Paul.
A. J. and PATTERSON,
K. E. (1978). Word recognition and production: reciprocity
MARCEL,
in clinical and normal research. In REQUIN,
J. (Ed.), Attention and Performance V I I .
Hillside, N. J. : Erlbaum.
MARSHALL,
J. C. and NEWCOMBE,
F. (1973). Patterns of paralexia: A psycholinguistic
approach. Journal of Psycholinguistic Research, 2, 175-99.
MEHLER,
J. and MORTON,
J. (1981). On reducing language to biology: playing with your
language organ. Behavioural and Brain Sciences, submitted for publication.
J. (1979). Facilitation in word recognition: experiments causing change in the
MORTON,
P. A., WROLSTAD,
M. E. and BOUMA,
H. (Eds), Processing
logogen model. KOLERS,
of Visible Language, Volume I. New York: Plenum Press.
MORTON,
J. and PATTERSON,
K. (1980). A new attempt at an interpretation, or, an attempt
J. C.
at a new interpretation. In COLTHEART,
M., PATTERSON,
K. and MARSHALL,
(Eds), Deep Dyslexia. London: Routledge and Kegan Paul.
K. E. (1978). Phonemic dyslexia: errors of meaning and the meaning of
PATTERSON,
errors. Quarterly Journal of Experimental Psychology, 30, 587-607.
PATTERSON,
K. E. (1981). Neuropsychological approaches to the study of reading. British
Journal of Psychology, 72, I 5 1-74.
PATTERSON,
K. E. and HATFIELD,
F. M. (1980). Surface dysgraphia: Parallels with and
divergences from surface dyslexia. Paper presented to the Experimental Psychology
Society, Cambridge, July 1980.
PATTERSON,
K. E. and KAY,J. A. (1980). How word-form dyslexics form words. Paper
presented to the British Psychological Society, Exeter, March 1980.
PATTERSON,
K. E., PURELL,
C. and MORTON,
J. (in press). Facilitation of word retrieval in
D. J. (Eds), Aphasia Therapy. London: Edward
aphasia. In CODE,C. and MULLER,
Arnold.
SAFFRAN,
E. M., BOGYO,L. C., SCHWARTZ,
M. F. and MARIN,0. S. M. (1980). Does deep
dyslexia reflect right-hemisphere reading? In COLTHEART,
M., PATTERSON,
K. and
MARSHALL,
J. C. (Eds), Deep Dyslexia. London: Routledge and Kegan Paul.
SCHIEPERS,
C. (1980). Response latency and accuracy in visual word recognition. Perception and Psychophysics, 27, 71-8 I.
SCHONELL,
F. J. (1942). Backwardness in the Basic Subjects. Edinburgh and London:
Oliver and Boyd.
SHALLICE,
T. (1979). Case study approach in neurosychological research. Journal of
Clinical Neuropsycliology, I, 183-21 I.
SHALLICE,
T. (1981). Neurological impairment of cognitive processes. British Medical
Bulletin, 37, 187-92.
SHALLICE,
T.and MCGILL,J. (1978). The origin of mixed errors. In REQUIN,J. (Ed.),
Attention and Performance VII. Hillsdale, N.J. : Lawrence Erlbaum.
SHALLICE,
T. and WARRINGTON,
E. K. (1975). Word recognition in a phonemic dyslexic
patient. Quarterly Journal of Experimental Psychology, 27, I 87-99.

Downloaded by [Central Michigan University] at 06:21 22 October 2015

LETTER-BY-LETTER READING

44'

SHALLICE,
T . and WARRINGTON,
E. K. (1977). The possible role of selective attention in
acquired dyslexia. Neuropsychologia, 15, 3 1-41.
SHALLICE,
T. and WARRINGTON,
E. K. (1980). Single and multiple component central
dyslexic syndromes. In COLTHEART,
M., PATTERSON,
K. and MARSHALL,
J. C. (Eds),
Deep Dyslexia. London: Routledge and Kegan Paul.
SHALLICE,
T., WARRINGTON,
E. K. and MCCARTHY,
R. (1981). Reading without semantics.
Paper presented to the Experimental Psychology Society, London, January 1981.
STACHOWIAK,
F. J. and POECK,K. (1976). Functional disconnection in pure alexia and
colour naming deficit demonstrated by facilitation methods. Brain and Language, 3,
135-43STALLER,
J., BUCHANON,
D., SINGER,
M., LAPPIN,
J. and WEBB,W. (1978). Alexia without
agraphia: An experimental case study. Brain and Language, 5, 378-57.
STROOP,
J. R. (1938). Factors affecting speed in serial verbal reactions. Psychological
Monographs, 50, 38-48.
VINCENT,
I;. M., SADOWSKY,
C. H., SAUNDERS,
R. L. and REEVES,
A. G. (1977). Alexia
without agraphia, hemianopia, or colour-naming defect : A disconnection syndrome.
Neurology, 27,689-91.
WARRINGTON,
E. K. and SHALLICE,
T. (1979). Semantic access dyslexia. Brain, 102,
43-63.
WARRINGTON,
E. I<. and SHALLICE,
T. (1980). Word-form dyslexia. Brain, 103, 99-112.
WOLPERT,
I. (1924). Die Simultanagnosie-Storung der Gesamtauffassung. Za'tschriftfiir
die gesumte Neurologie und Psychiatrie, 93, 397-415.
ZAIDEL,E. (1978). Lexical organisation in the right hemisphere. I n BUSER,P. and
ROUCEUL-BUSER,
A . (Eds), Cerebral Correlates of Conscious Experience. Amsterdam :
Elsevier.
Manuscript received 3 September 1981