Phonological processing during silent reading in aphasic patients

BRAIN

AND

LANGUAGE

19, 191-203 (1983)

Phonological

Processing during Silent Reading in Aphasic Patients PHYLLIS

Ross

In a task involving the detection of a predesignated letter during the silent reading of a series of passages, left-brain-damaged aphasic patients and rightbrain-damaged patients showed patterns of performance consistent with those of normal individuals. Both of the brain-damaged groups were more likely to detect letters when they were pronounced in a typical way (e.g., R in “ago”) than in an atypical way (e.g., R in “through”), suggesting the use of phonological recoding during silent reading. In addition, both of these groups were more likely to detect letters when they occurred in content than in function words, suggesting a differential processing of these word types. Possible differences in the strategies predominantly relied on for phonological recoding and for the differential processing of content and function words by different groups of patients are discussed.

The results of numerous studies suggest that individuals make use of phonological processing during silent reading. For example, it has been shown that in a lexical decision task, letter sequences which are homophones of real words (e.g., trate) take longer to reject as lexical items than those which are not homophones (e.g., trute) (Rubenstein, Lewis, & Rubenstein, 1971). In addition, real words which have homophones (e.g., sale) take longer to accept as lexical items than those which do not have homophones (e.g., bake; Rubenstein et al., 1971). The use of phonology for single-word reading is also suggested by the finding that subjects take longer to recognize or react to printed words which contain more syllables than other items with the same number of letters (Eriksen, Pollack, & Montague, 1970; Klapp, 1971). This research was supported by NIH Fellowship 5F32 NS05938-02. The author thanks Dr. Jason Brown and other members of the Neurology Department at New York University Medical Center who were of assistance in the planning and analysis of this study. and thanks members of the Department of Speech and Hearing at the Veterans Administration Hospital in New York who were of assistance in selecting subjects for the study. Address reprint requests to Dr. Phyllis Ross, Department of Psychology, William Patterson College. Wayne, NJ 07470. I91 0093-934X183 $3.00 CopyrIght C 1983 by Academx Press. Inc All nphts of reproduct~m in an) form reervcd.

192

PHYLLIS

ROSS

In addition to the use of phonological recoding during silent reading of single words, there is evidence for such recoding in the reading of prose. In searching through a passage for words spelled incorrectly, subjects are more likely to detect a misspelling in words spelled in a phonetically incompatible way (e.g., wetch for watch) than in a compatible way (e.g., lurn for learn; McKay, 1968). In addition, in a task involving the detection of a predesignated letter, subjects are more likely to detect the letter when it occurs in a pronounced than silent form (Corcoran, 1966). In related studies involving proofreading tasks, it has been shown that subjects are more likely to detect the omission of a pronounced than a silent letter (Corcoran, 1967; Corcoran & Weening, 1968). Although normal readers may utilize phonology during silent reading, it may be possible to read by directly accessing the meaning of a printed word from its visual properties without prior phonological analysis. Evidence for this view of reading is provided by the types of impairments shown by patients who have been categorized as “deep” (Marshal1 & Newcombe, 1973) or “phonemic” (Shallice & Warrington, 1975) dyslexics. This disorder sometimes occurs when a previously fluent reader sustains left-hemisphere brain damage. Several investigators have concluded that such patients are unable to encode a word phonologically prior to accessing its meaning, and that the phonological representation of a word and thus the ability to read the word aloud must be derived from the internal representation of the item in lexical memory (Marsha11 & Newcombe, 1973; Patterson & Marcell, 1977; Saffran & Marin, 1977; Shallice & Warrington, 1975). It has been shown that such patients cannot read aloud orthographically regular nonwords (e.g., pake; Patterson & Marcell, 1977), and cannot pronounce or comprehend the meaning of a word from the visual presentation of the homophonic nonword (e.g., kote; Saffran & Marin, 1977). When attempting to read real words aloud, such patients may produce responses which are semantically related to the stimulus word (e.g., read “sick” as “ill”) (Marshall & Newcombe, 1973). In addition, these patients have great difficulty in reading abstract words (Shallice & Warrington, 1975; Patterson, 1979; Patterson & Marcell, 1977) and even more difficulty in reading function words (Marshall & Newcombe, 1973, Patterson, 1979; Saffran & Marin, 1977). These types of reading impairments suggest that deep dyslexics are unable to encode words phonologically prior to lexical access, and that they can comprehend the meaning of printed words without the use of phonology. Thus, the meaning of a printed word may be retrieved from lexical memory in two ways. Readers may either first recode the word into a phonological form and then use this to access the semantic representation, or they may use a more direct route from the visual characteristics of the printed word to meaning. (See McCusker, Hillinger, & Bias, 1981 for a comprehensive review of such a dual model of lexical access). Use

SILENT

READING

IN

APHASIC

PATIENTS

193

of the direct visual route would mean that an individual could read silently without the use of phonology. One purpose of the present investigation was to examine the extent to which aphasic patients make use of phonological processing during silent reading. To the extent that recoding words into a phonological form is helpful during reading, the reading impairments often seen in aphasic patients may be due to some extent to a failure to utilize phonology. The task used in the present study was based on one originally used by Corcoran (1966). As previously mentioned, Corcoran found that readers were more likely to detect pronounced than silent letters in a reading task, suggesting the use of phonological coding. Chen (1974), using a similar task with deaf college students, found that whereas hearing subjects were more likely to detect pronounced than silent letters, deaf subjects showed no such difference. Similar results for deaf and hearing children were found by Locke (1978), who compared the detection of letters which occurred in phonemically modal and nonmodal forms. Phonemically modal forms were instances where letters were pronounced in a typical way (e.g., /R/ for the letter g as in the word “go”) and nonmodal forms were instances where letters were pronounced in a nontypical way (e.g., if/ for the letter g as in the word “rough”) as well as cases in which the letter was silent (e.g., the s in “right”). Thus, both the studies by Chen (1976) and Locke (1978) suggest that deaf individuals do not phonologically recode words during silent reading. One purpose of the present study was to see whether the same was true for aphasic individuals. In addition to examining the extent to which aphasic patients utilize phonology during silent reading, the present study was also designed to examine whether aphasic individuals differentially process content and function words. Several studies have shown that normal readers are more likely to detect letters which occur in content than in function words (Locke, 1978; Hatch, Polin, & Part, 1974; Schindler, 1978). Furthermore, in comparing the effects of prose vs. scrambled prose contexts, it has been shown that the presence of a prose context increases the likelihood of letter detection in a content word and decreases the likelihood in a function word (Schindler, 1978). These findings may be due to the fact that in reading prose, normal readers can use their knowledge of language to predict grammatical relationships, and can readily supply the function words. Since function words are highly predictable and uninformative, subjects are likely to skip over them when reading. A second explanation for the fact that readers attend more to letters in content words involves the use of stress patterns in reading. Normally. content words are stressed in sentences whereas function words are unstressed. Therefore, the use of stress patterns during silent reading would make content words more prominent than function words. Locke (1978) found that deaf subjects, unlike normal subjects, showed

194

PHYLLIS

ROSS

no difference in detecting letters in content and function words during silent reading. Hatch et al. (1974) also found no difference in detection of letters in these two word types by foreign students in an elementary English class, in contrast to native speakers of English who detected more letters in content words. Since both deaf individuals and individuals who are first learning English would be expected to be relatively poor readers, these results may reflect the fact that unskilled readers lack sufficient linguistic knowledge to predict grammatical relationships, and therefore need function words to understand the syntax of the sentence. Alternatively, these results may indicate that deaf readers and/or unskilled readers do not rely on stress patterns in sentences as an aid in reading. If the failure to differentially process content and function words is due to an impoverishment in linguistic knowledge, it might be expected that aphasic patients would also show no differences in the processing of these word types. However, if this pattern is due to an inability to use stress patterns in sentences, it might be expected that aphasic patients, like normal readers, would pay more attention to content than function words during reading. This prediction is based on evidence suggesting that aphasic patients can make use of stress patterns in sentences, and in fact may be even more sensitive to stress than normal individuals (Kellar, 1978). The present study was designed to examine these alternative possibilities. In a recent study by Locke and Deck (1982) involving a task similar to that used in the present study, aphasic and nonaphasic patients had to detect letters during the silent reading of a prose passage. Both groups detected more phonemically nonmodal than modal letters, but only the nonaphasic patients detected more letters in content than function words. These results suggest that, although aphasics appear to rely on phonological recoding during silent reading, they may not utilize syntax in the same manner as normal readers. However, since comprehension of the test passage was not examined in this study, it is possible that the aphasic patients were not reading for meaning, and that aphasics may show a syntactic effect when comprehension of the material to be read is required. The current study was designed to explore this possibility, as well as to examine whether the use of syntax and phonology during silent reading differs for different subgroups of aphasic patients. METHOD Subjects The subjects tested were 14 left-brain-damaged aphasic patients and 6 right-brain-damaged patients without aphasia. All of the subjects, with the exception of one aphasic, were male. The patients were from Goldwater Memorial Hospital and from the Veterans Administration Hospital in New York City. On the basis of clinical evaluation, seven of the aphasic patients were classified as having anterior damage and seven as having posterior damage.

SILENT

READING

IN APHASIC

195

PATIENTS

No patients were included who had a visual field defect. Two additional anterior aphasics were excluded from the study because they indicated that they were unable to understand the reading material and refused to complete the task.

Stimuli

and Procedure

Subjects were given a set of 18 simple short paragraphs to read silently. each of which was typed on a separate page. At the same time that a subject was reading each paragraph, he or she was asked to cross out all instances of a predesignated target letter. Subjects were instructed to read each paragraph only one time and not to go back and cross out any letters after the paragraph was read. Subjects were tested individually. Three target letters were used: (‘. g, and h. These were the same target letters used by Locke (1978) and by Locke and Deck (in press). Each of these served as target letters for + of the paragraphs. The order of presentation of the paragraphs was blocked by target letter. The appropriate target letter was typed in lower case in the upper right hand corner of each page. In each paragraph. the predesignated letter occurred in phonemically modal and nonmodal forms. The modal and nonmodal phonemes for the three target letters, the total number of instances in which each phoneme occurred, as well as examples of words in which each phoneme occurred are presented in Table I. After reading each paragraph. the subject was given a written multiple choice question about some information contained in the paragraph. and was instructed to check the correct response out of three. These questions were given to ensure that subjects would in fact read the paragraphs for meaning. The following is an example of a stimulus paragraph and the question associated with it. The mother went into the kitchen in order to prepare dinner. Her child helped her choose what to include in the menu for dinner, and then sat on a chair and watched her cook. The mother made some meat and corn, and baked a big rich cake for desert. When the food was ready. the mother called her husband and children to the table for dinner.

TABLE PHONEMIC

Phonemically Phoneme

CHARACTERISTICS

I OF THE TARGET LETTERS

modal letters

Phonemically nonmodal letters Phoneme

II

Examples

ICI

35

watch

behind

Id lsl .fi ldl !-I

Sh

much rush enough the night

go

lfi

29

rough ring night

II

Examples

lkl

36

because

lhl

35

fgl

37

c H

hi l-l Sum

108

120

196

PHYLLIS

ROSS

The mother went to the kitchen to have breakfast prepare dinner clean the stove. Each letter occurred in phonemically modal and nonmodal forms in nouns, verbs, adjectives, and functors (adverbs, prepositions, pronouns, conjunctions, determiners, and participles). in addition, target letters were distributed across initial, medial, and final word positions as much as possible. The number of words which contained target letters in phonemically modal and nonmodal forms in each grammatical class and in each word position is presented in Table 2. All of the words in which target letters appeared had a frequency of at least 50 times per million according to the Thorndike and Lorge word frequency count (1943). These words contained from 3 to 9 letters, with the mean length being 4.8. All subjects were additionally given the Reading Comprehension subtest from the Boston Diagnostic Aphasia Examination, and the aphasic subjects were also given the Token Test as a measure of auditory comprehension.

RESULTS It was clear from the responses to the questions following each paragraph that all of the patients tested were in fact reading the paragraphs and not merely looking for the target letters. One of the anterior aphasic patients got only 9 of the 18 questions correct, but all of the remaining patients got at least 14 correct. The mean number of correct responses for the right-damaged, anterior aphasic, and posterior aphasic patients was 17.7, 15.8, and 16.8, respectively. There were very few instances in which letters were crossed out other than the target letters. The results were analyzed in terms of detection errors, i.e., instances where a target letter occurred but was not crossed out. The mean percentage of errors for each of the groups to each of the target letters in phonemically modal and nonmodal forms can be seen

INSTANCES OF TARGET

TABLE 2 LETrEas IN WORDS IN DIFFERENT

GRAMMATICAL

CLASSES

AND WORD POSITIONS

Phonemically

modal letters

Phonemically nonmodal letters Grammatical

Nouns

Verbs

34

24

Adjectives

class

Functors

Nouns

Verbs

31

29

29

19

Adjectives 12

Functors 50

Word position Initial 85

Medial 17

Final 6

Initial 16

Medial 88

Final 16

SILENT

READING

IN APHASIC

197

PATIENTS

in Table 3. A 3 x 2 x 3 Analysis of Variance (group x phonology x letter) showed no significant difference in percentage of errors between groups, F(2, 17) = 2.32. The mean percentage of errors for the rightdamaged, anterior aphasic, and posterior aphasic patients was 14.5, 35.3, and 20.4, respectively. The relatively high error rate for the anterior aphasics was caused by two patients whose error rate was close to 70%, which was well above the error rate for any of the other patients tested. There was a significant main effect of phonology, F(l, 17) = 15.80, p < .OOl, indicating a greater percentage of errors when letters occurred in phonemically nonmodal than modal form. As can be seen in Table 3, this was true for each of the groups and for each of the target letters; thus, there were no significant interactions of phonology with group or with letter. The mean percentage of errors for letters in modal and nonmodal forms for each of the groups, collapsed across target letter, is shown in Fig. I. There was also a significant main effect of letter, F(2, 34) = 8.30, p < .005. The Newman-Keuls procedure showed that there was a significantly (r, < .Ol) greater percentage of errors to the Letter h (33.4%) than the Letters c’ (18.8%) or g (20.9%). There was no significant group x letter interaction, indicating that all three groups had a greater percentage of errors to the letter h. The mean percentage of errors for each of the groups to letters in modal and nonmodal form in words in each of the four grammatical classes is presented in Table 4. A 3 x 2 x 4 Analysis of Variance (group x phonology x grammatical class) showed a significant main effect of grammatical class, F(3, 51) = 25.33, p < .OOl. The Newman-Keuls procedure showed that there was a significantly (p < .Ol) larger percentage of errors to letters which appeared in function words (36.8%) than to letters in nouns (16.7%), verbs (20.7%), or adjectives (18.5%). The latter three categories of words did nor significantly differ from each other. As can be seen in Table 4. a larger percentage of errors occurred for TABLE

3

MEAN PERCENTAGE OF ERRORS IN DETECTING PHONEMICALLY MODAL AND NONMODAL LETTERS FOR THE THREE TARGET LETTERS BY GROUPS OF BRAIN-DAMAGED PATIENTS

H

c

Modal

Group

n

Aphasic Anterior Posterior Right

7 7 6

19.0 10.3 7.4

All subjects

19

12.5

Nonmodal

G

Modal

Nonmodal

Modal

Nonmodal

38.8 21.6 13.8

28.5 16.3 17.6

55.5 37.4 28.9

26.9 10.9 8.6

42.9 26. I 10.9

25.3

20.9

41.2

15.8

27.4

198

PHYLLIS

q

Modal

fl

Non-modal

50s 40i 6 30kz ,m 20fi e a”

ROSS

46.0

20.1

23.8

2.5.6

loc-I& Right Damaged

Anterior Aphasic

Posterior Aphasic

FIG. I. Percentage of errors to letters occurring in phonemically modal and nonmodal forms by groups of brain damaged patients.

letters in function words for each of the three groups for both the modal and nonmodal forms of letters; thus, there was no significant interactions of grammatical class with group or phonology. The mean percentage of errors in detecting letters in words in each grammatical class for each group, collapsed across phonemically modal and nonmodal forms is shown in Fig. 2. Although there were no significant differences between groups with regard to error rates for letters in phonemically modal and nonmodal forms, it was possible that there would be a relationship between the magnitude of the difference between detection of letters in modal and nonmodal forms and levels of auditory or reading comprehension. In order to examine whether such relationships existed, a phonological index was computed for each subject by subtracting the percentage of errors for letters occurring in a phonemically modal form from the percentage of errors for letters occurring in a nonmodal form, and dividing by the total number of errors. Thus, a phonological index of 0 suggested no phonological coding whereas an index of 1 suggested maximum phonological coding. For the aphasic patients as a group, the phonological indices ranged from .03 to .74, with a mean of .38. For the right-braindamaged patients, these indices ranged from .lO to .78, with a mean of .41. The number of correct responses out of 36 on the Token Test was used as a measure of auditory comprehension for the aphasic patients. The scores for those patients ranged from 2 to 28, with a mean of 15.8. The correlation between the phonological indices and the Token Test scores was - .04, providing no evidence for a relation between auditory comprehension and phonological coding. The number of correct responses out of 10 on the Reading Comprehension subtest of the Boston Diagnostic Aphasia Exam was used as a measure

SILENT

READING

IN APHASIC

PATIENTS

200

PHYLLIS

0

Right Damaged

ROSS

Anterior Aphasic

FIG. 2. Percentage of errors to letters occurring classes by groups of brain damaged patients.

Posterior Aphasic

in

words

in different grammatical

of reading comprehension. The scores for the aphasic patients ranged from 3 to 10, with a mean of 6.8. The scores for the right-brain-damaged patients ranged from 7 to 10, with a mean of 9.0. The correlations between the phonological indices and the Reading Comprehension scores for both the aphasic and right-brain-damaged patients were positive but nonsignificant, r(12) = .30 and r(4) = .24, respectively. DISCUSSION

One purpose of this study was to examine whether aphasic patients encode words into a phonological form during silent reading. The fact that the anterior and posterior aphasics, as well as the right-brain-damaged patients were more likely to detect letters occurring in a phonemically modal than nonmodal form is consistent with the results of the study by Locke and Deck (1982), and suggests the use of phonological processing for all groups of patients tested. An alternative explanation for the finding that subjects had higher error rates for letters occurring in a phonemically nonmodal form is related to the fact that in almost all cases (98.3%) in which a letter occurred in nonmodal form, the letter was either preceded and/or followed by another consonant and therefore formed part of a consonant cluster (e.g., ch, ng, ght). In contrast, letters which occurred in a modal form were preceded or followed by another consonant only 15.7% of the time. Therefore, it is possible that the higher error rate for letters occurring in nonmodal form was due to subjects treating the consonant clusters as single visual units and not processing the graphemes forming these clusters as individual items. This purely visual interpretation of the results would not necessarily entail the use of phonological processing. However, the result indicating that there was a difference in the rate of detection errors among the

SILENT

READING

IN APHASIC

PATIENTS

201

three target letters supports the argument that subjects were in fact encoding the letters phonologically and makes this purely visual interpretation unlikely. The results indicated that all three groups of subjects were less likely to detect the Letter h than the Letters c or g. Since the Letter h is in no way visually less prominent than the other letters, these results suggest that subjects made use of some phonetic characteristic(s) of the letters during the reading task which caused the Letter h to be phonetically less prominent. One reason that the Letters c and g appear to be more prominent than the Letter h may be that there is articulatory contact for the former letters but not for the latter, although other explanations are certainly possible. Although the results therefore suggest the use of phonology by all patients tested, this does not necessarily indicate that the route through which the phonological representation was obtained was the same for all subjects. As discussed in the introduction, the phonological representation for a printed word may be derived prior to lexical access through the use of grapheme-phoneme correspondence rules, or alternatively, may be obtained after lexical access by verbally labeling the semantic representation which has been activated in memory. Different types of linguistic impairments in anterior and posterior aphasics may make one or the other of these routes to phonology the most accessible. Anterior aphasics are certainly more impaired than posterior aphasics in phonological production. Therefore, such patients might have difficulty in utilizing grapheme-phoneme correspondence rules and might rely mainly on a postlexical access mechanism for obtaining the phonological representation of a printed word. On the other hand, posterior aphasics are generally more impaired in semantic processing. For example, it has been found that Wernicke aphasics are often unable to match an auditorily presented word to its equivalent picture when presented in an array of phonemically similar items (Schuell, Jenkins, & Jimenez-Pabon, 1964). suggesting that they are unable to maintain a stable phonological representation for lexical items. Since such patients have difficulty in obtaining the phonological representation of a word via meaning, they might rely mainly on prelexical access mechanisms. Nonaphasic individuals probably use some combination of these reading strategies. Thus, although all three groups of patients tested showed evidence of phonological processing, it seems likely that different types of patients used different strategies in deriving phonological representations. The results of the present study suggest that the ability of aphasic patients to derive an internal phonological representation of a printed word is independent of word class; i.e., these patients showed evidence of phonological coding for both content and function words. These findings are in apparent contrast to the results of a study by Friederici and Schoenle (1980) examining the oral reading ability of two aphasic patients.

202

PHYLLIS

ROSS

In reading aloud single homophonic items in which spelling distinguished content and function words (e.g., be-bee, some-sum) the patient with Broca’s aphasia showed a greater disruption in reading function words whereas the patient with Wernicke’s aphasia showed a greater disruption with content words. Therefore, it might have been expected that in the present study, the extent of phonological coding shown by anterior and posterior aphasics would have been a function of word type. However, the present study involved the reading of prose rather than isolated words, and this may have an effect on the type of reading strategy used. A second experiment by Friederici and Schoenle (1980) in which a sentence reading task was given and the reading of homophonic content and function words was compared supports this view. In this experiment, although there were differences in the kinds of errors made for content and function words by the Wernicke’s aphasic, the level of reading performance for these two word types was identical. (No data were available for the Broca’s aphasic, since he was unable to read sentences aloud). In addition to providing evidence that aphasic patients use phonology during silent reading, the results of the present study suggest that content words are more salient than function words for both anterior and posterior aphasics as well as right-brain-damaged patients when reading for meaning. However, once again, the same pattern of performance by different groups of subjects does not necessarily reflect the same mode of processing. It seems unlikely that aphasic patients could predict and therefore skip over the function words while reading the same ability as more skilled readers. Rather, it is more probable that the relative failure to detect letters occurring in function words for aphasic patients reflected the use of stress patterns in sentences. For the right-brain-damaged patients, the differential processing of content and function words may have resulted both from their ability to predict function words as well as their use of stress patterns. In summary, the results of the present study suggest that the performance of aphasic patients is similar to that of normal individuals with regard to two aspects of reading; i.e., the use of phonology and the differential processing of content and function words. However, reading is an extremely complex task, and may be performed in different ways by individuals with varying linguistic and cognitive abilities. Thus, a similarity in performance of aphasic and nonaphasic individuals on a reading task or any other complex cognitive task may attest to the flexibility of aphasics in adopting those strategies which they can most effectively use, rather than to a uniformity in the type of processing strategy used by aphasic and nonaphasic individuals. REFERENCES Chen,

1. 1976. Acoustic Journal o.f Generul

image in visual detection Psychology:y, 94, 243-246.

for deaf

and hearing

college

students.

SILENT

READING

IN APHASIC

PATIENTS

203

Corcoran, D. W. J. 1966. An acoustic factor in letter cancellation. Ntrtlrrr (Lo/?don). 210. 658. Corcoran. D. W. J. 1967. Acoustic factor in proofreading. Ntrtrrrc~ (Lor~clorr). 214, 851852. Corcoran, D. W. J.. & Weening. D. L. 1968. Acoustic factors in visual search. Qrrtrrter!\ Jo7trnu/ of E.\-prritnenfu/ Ps.vf~ho/o,!g, 20, 83-85. Ericksen. C. W., Pollack. M. D.. & Montague. W. E. 1970. Implicit speech: Mechanism in perceptual encoding. Jolr,ntr/ c!f‘Erpcrirn~nl~I P.s.vc~/~o/o~~. 84, 502-507. Friederici, A. D.. & Schoenle. P. W. 1980. Computational dissociation of two vocabulary types: Evidence from aphasia. Nerrr-op.c?‘c,ho/c,Xicr. 18, I I-20. Hatch. E.. Polin, P.. & Part, S. 1974. Acoustic scanning and syntactic processing in three reading experiments-First and second language learners. Jownd of Reuding Behtr~~ior. 3, 275-285. Kellar. L. 1978. Srres\ rrnd svntrr.~ in clphusitr. Paper presented at the Academy of Aphasia Meeting. Chicago. Klapp. S. T. 1971. Implicit speech inferred from response latencies in same-different decisions. Paper presented at Western Psychological Association. (Cited in: Posner. M. I.. Lewis. J. L., & Conrad, C. Component processes in reading: A performance analysis. In J. F. Kavangh & 1. G. Mattingly (Eds.), Langrrcc~r h.v enr clnd CHIC: Tile ~.~/~(~iomhip~ /,et+~ec~n speech cind retrdinfi, Cambridge, MA: MIT Pres5.j Locke, J. L. 1978. Phonemic effects in the silent reading of hearing and deaf children. Cognition. 6, I75- 187. Locke. J. L., & Deck, J. W. 1982. The processing of printed language by aphasic adults: Some phonological and syntactic effects. Jmrnul of Speech und Hearing Resrarch. 25, 314-319. Marshall, J. C.. & Newcombe, F. 1973. Patterns of paralexia: A psycholinguistic approach. Jolrrnul of Psycholinglcistic Re.seurc,h, 2, 175-199. McKay, D. G. 1968. Phonetic factors in the perception and recall of spelling errors. Nrlrrops?‘c,llo/oRio, 6, 32 I-325. McCusker. L. X.. Hillinger. M. L.. & Bias, R. G. 1981. Phonological recoding and reading. Psychological Bulletin, 2, 217-245. Patterson. K. E. 1979. What is right with “deep” dyslexic patients‘? Bruin und Langttug~, 8, 11 l-129. Patterson. K. E.. & Marcell, A. J. 1977. Aphasia. dyslexia and the phonological coding of written words. Quarter/y Jmrnrrl of Exp-perirnentul Psychology, 29, 307-318. Rubenstein. H., Lewis. S. S., & Rubenstein. M. A. 1971. Evidence for phonemic recoding in visual word recognition. Jorrmul of‘ Verbal Leur-ning und Verhnl Beha\~ior. 10, 64% 657. Saffran. E. M., & Marin, 0. S. M. 1977. Reading without phonology: Evidence from aphasia. Qrrnrrrrly Jorrrnul of E.~perirnenrul Psy~holo,q~. 29, 5 15-575. Schindler. R. M. 1978. The effect of prose context on visual search for letters. Memo,:\ und Cognirior7. 6, l24- 130. Schuell, H.. Jenkins, J. J.. & Jimenez-Pabon. E. 1964. Aphu.\iu in trdrtlt.~. New York: Harper & Row. Shallice, T.. & Warrington, E. K. 1975. Word recognition in a phonemic dyslexia patient, Qlrarterly Joltrnul of‘ Experitnenful Psychology, 27, 187-199. Thorndike. E. L., & Lorge, I. 1944. The teuchrr’s \tvrd hook c$‘30.000 lords. New York: Teachers College Press.