Narrative Speech Deficits Dyslexics* · L), Late effects of head injury, (pp. 501-524). brain syndromes: An introduction to neuro- e. Thehncet, 1. 1-5. idental head injuries: Prognosis

I. BINDER

! cerebral white matter in severe dementia. hiutry. 19. 163-185. mage due to blunt head injuries. In A. E. L), Late effects of head injury, (pp. 501-524).

brain syndromes: An introduction to neuro-

e. Thehncet, 1. 1-5. idental head injuries: Prognosis in service

,itish Medical Journal, 3, 67-7 1. srebral circulation after concussional head

S.

nt of coma and impaired consciousness:

1970). Posttraumatic dizziness. Archives of

)halograms in whiplash injury. Archives of

ew York: J. Wiley. after acute mild head injury. Acta Oto-

Long-term recovery of visual reaction time Neurosurgery. and Psychiatry, 41,452457. ,itivity to light and sound following minor

p study of head wounds in World War II. Office. 4). Experimental structural alterations in , Gynecology, and Obstetrics. 79. 56 1-572. onal and structural changes in the central concussion. Transactions of the American

T work and symptoms after minor head

.mnestic states in football trauma. Neuro-

') 270-276.

Journal of Clinical and Experimental Neuropsychology 0168-8634/86/0804-0347 S 3.00 1986, Vol. 8, NO. 4, pp. 347-361. Q Swets & Zeitlinger

Narrative Speech Deficits in Dyslexics*

Linda Davenport, Charles D. Yingling, George Fein, David Galin, and Jack Johnstone Langley Porter Psychiatric Institute

University of California, San Francisco

ABSTRACT

Narrative speech of 31 dyslexics and 33 good readers was studied. Subjects were 10- to 12-year-old right-handed boys of normal intelligence, and with normal neurological status, vision, and hearing. The dyslexics used shorter communication units (independent clauses with all their modifiers), and a higher percentage of their words were noncommunications (words which are extraneous to the speaker's intended meaning). Rate of speech was the same in both groups. These findings were replicated in an independent second cohort of 21 dyslexics and 21 controls, aged 9 to 13. Our findings support theories of a general language deficit in dyslexia, and demonstrate the involvement of the previously unexamined area of narrative speech.

While dyslexia is widely.viewed as the result of a cognitive disorder, the underlying mechanisms are still not understood. There are now thought to be several different subtypes of dyslexia, and a number of classification schemes have been proposed (see review by Malatesha & Dougan, 1982). Recent studies, some based on clinical neuropsychological models (Denckla, 1979; Mattis, 1978; Mattis, French, & Rapin, 1975) and others on cluster and/or factor-analytic techniques (Doehring & Hoshko, 1977; Satz & Morris, 1981; Watson, Goldgar, & Ryschon, 1983), report that, in a substantial number of cases, a generalized deficit in verbal language is prominent. The language deficit has been assessed with tests of auditory discrimination, sound blending, sentence repetition, aural comprehension, word meanings, and confrontation naming. None of these studies has examined the integrity of narrative speech, which we believe to be especially important to dyslexics because of their reliance on oral communication.

Two previous studies have suggested a relationship between reading ability and

* This work was supported by U.C.S.F. Academic Senate Grant 504931-09540, NICHD Contract NO 1-HD-8-2824,NINCDSGrant 1 ROl-NS- 17657,andNIMHGrant 1 R03-MH- 36871. A preliminary report of these findings was presented at the 1984 meeting of the International Neuropsychological Society, Houston, TX.

Requests for reprints should be sent to Linda Davenport, Box GD, Langley Porter Psychiatric Institute, San Francisco, CA 94143, USA.

Accepted for publication July 30, 1985.

348 LINDA DAVENPORT ET AL.

narrative speech. Fry, Johnson, and Muehl (1970) demonstrated that deficits in several aspects of oral language, including narrative speech, were characteristic of poor readers in the second grade, and Loban (1976) found ability in reading and narrative speech to be associated throughout the 12 years of his longitudinal study. In order to determine the extent to which deficits in narrative speech are characteristic of dyslexia, in this study we compared narrative speech samples of dyslexic and normal readers.

The previous studies which have identified a variety of language problems have been based on subject populations in which many variables were confounded. The inclusion of subjects with neurological or psychiatric problems, low IQ, bilin- gualism, or hyperactivity leaves open the question of the extent to which the observed deficits are associated with dyslexia per se (Vellutino, 1979). The present study employed stringent screening criteria for primary or pure dyslexia. Studying this population enables us to draw inferences about the relation of oral language deficits to reading disability itself.

METHOD

This study was part of a large-scale project examining the neuropsychology of reading disabilities (Yingling & Galin, 1982). Experiments based on both electrophysiological and behavioral measures were designed to test several of the current theories about the causes and correlates of dyslexia. The full test battery included EEG and probeevent-related potentials (ERPs) during resting conditions and complex behaviors (Fein et al., 1983,1984; Johnstone et al., 1984; Yingling, Galin, Fein, & Peltzman, 1985), ERPs to attended stimuli, tests of vestibular function (Brown et al., 1983a), eye movements (Brown et al., 1983, 1983b), sensori-motor integration, behavioral measures of lateral preference, and psycho- metric tests of academic achievement, verbal language, andcognitive abilities related to the reading process. The interrelations between the language measures examined here and these other variables will be treated elsewhere.

Subjects Screening criteria were designed to select dyslexics whose reading abilities were severely impaired and yet who were free from other conditions which might have interfered directly or indirectly with the acquisition of reading skills. All subjects were right-handed Caucasian boys, aged 9 to 13, of middle-class SES, whose only language was English, and who had attended schoolregularly. Most ofthedyslexicswerereferred fromprivateremedialschools and other special-education programs. Most of the controls were self-referred in response to classified ads in local newspapers. In order to avoid a skewed self-selection of controls, we solicited boys to participate in a “scientific research” project, without indicating the emphasis on reading.

Parent interviews were conducted to establish normal pregnancy, birth, and neonatal health, and a history free from accidents or illness which might have affected brain function. Children with a history of hyperactivity, epilepsy, emotional problems, developmental anomalies, or previously diagnosed deficits in oral language were excluded.

NARRATIVE SPEEI

Candidates were given clinical hearing, v Prechtl. 1970), the Wechsler Intelligence ! oral and silent reading tests. Candidate findings on the sensory or neurological te

Reading ability was considered in rela basis of a child’s age and IQ. Oral read (Gray, 1%3), and silent-reading vocab Gates-MacGinitie Reading Tests (Gat (Myklebust, 1968) was used to determi achievement:

Reading Quotient = [(2 x Read All dyslexics accepted into the study

reading tests, which is equivalent to 2 yea IQ of 100. All control subjects’ reading above, and their oral reading scores were

These stringent screening criteria resuli Even among those who had passed thc screening battery, less than one third met I

The groups which were finally selected cl readers, with equivalent group means an( 6.0 grade level difference between the gr grade level difference in oral reading (see

It may be the case that some of the dyslexia alone. Sixty-four percent of the than one grade level below age-expectant Wide Range Achievement Test (Jastak underlying learning disability on the bas repeated one or more grades in schoc emphasized language arts, at the possi therefore, that they had not received the expected at their age level, and the interp

Age, IQ, and Reading SC Cohort 1 and

h

Age in Months 1

Performance IQ I Full Scale IQ 1

WISC-R Verbal IQ

Oral Reading Grade Reading Comprehension Grade

’ORT ET AL.

11 (1970) demonstrated that deficits in iarrative speech, were characteristic of an (1976) found ability in reading and lout the 12 years of his longitudinal which deficits in narrative speech are ompared narrative speech samples of

3 a variety of language problems have many variables were confounded. The psychiatric problems, low IQ, bilin- question of the extent to which the I per se (Vellutino, 1979). The present or primary or pure dyslexia. Studying :s about the relation of oral language

D

mining the neuropsychology of reading s based on both electrophysiological and ’ of the current theories about the causes included EEG and probe-event-related ~mplex behaviors (Fein et al., 1983, 1984; :Itzman, 1985), ERPs to attended stimuli, ), eye movements (Brown et al., 1983, isures of lateral preference, and psycho- iage, and cognitive abilities related to the language measures examined here and

:s whose reading abilities were severely ons which might have interfered directly ill subjects were right-handed Caucasian ily language was English, and who had ereferred fromprivateremedialschools :controls were self-referred in response void a skewed self-selection of controls, search” project, without indicating the

iormal pregnancy, birth, and neonatal ness which might have affected brain epilepsy, emotional problems, develop- cits in oral language were excluded.

349 NARRATIVE SPEECH DEFICITS IN DYSLEXICS

Candidates were given clinical hearing, vision, and neurological examinations (Touwen & Prechtl, 1970), the Wechsler Intelligence Scale for Children (Revised) (Wechsler, 1974), and oral and silent reading tests. Candidates with Full-Scale IQs below 88, or with positive findings on the sensory or neurological tests, were excluded.

Reading ability was considered in relation to what would normally be expected on the basis of a child’s age and IQ. Oral reading was tested with the Gray Oral Reading Test (Gray, 1%3), and silent-reading vocabulary and comprehension were tested with the Gates-MacGinitie Reading Tests (Gates & MacGinitie, 1965). A reading quotient (Myklebust, 1968) was used to determine the ratio of actual achievement to expected achievement:

Reading Quotient = [(2 x Reading Age)/(Chron. Age +Mental Age)] All dyslexics accepted into the study had reading quotients at or below .80 on both

reading tests, which is equivalent to 2 years below grafle level for a 10-year-old child with an IQ of 100. A11 control subjects’ reading quotients on the Gates-MacGinitie were .93 or above, and their oral reading scores were within 1 year of actual grade placement, or higher.

These stringent screening criteria resulted in the exclusion of several hundred candidates. Even among those who had passed the telephone interviews and who were given the screening battery, less than one third met all the criteria and were actually selected for study. The groups which were finally selected consisted of essentially “pure” dyslexics, and good readers, with equivalent group means and similar distributions of age and PIQ. There was a 6.0 grade level difference between the groups in reading comprehension scores and a 6.4 grade level difference in oral reading (see Table 1).

It may be the case that some of the dyslexics had learning disabilities broader than dyslexia alone. Sixty-four percent of the dyslexics and only 7% of the controls were more than one grade level below age-expectancy in arithmetic computation, as measured by the Wide Range Achievement Test (Jastak & Jastak, 1965). However, we cannot infer an underlying learning disability on the basis of this measure alone; many of these boys had repeated one or more grades in school, or were in remedial classes which strongly emphasized language arts, at the possible expense of other subject areas. It is likely, therefore, that they had not received the kind or amount of arithmetic instruction that is expected at their age level, and the interpretation of their performance deficit on this test is

Table 1

Age, IQ, and Reading Scores of Dyslexics and Controls, Cohort 1 and Cohort 2 Combined

Dyslexics Controls n = 52 n = 54

Mean SD Mean SD p <

Age in Months 137.9 (13.1) 141.1 (11.5) n.s. WISC-R Verbal IQ 99.8 (9.2) 112.3 (10.7) .ooO1 Performance IQ 108.3 (12.4) 111.6 (11.8j n.s. Full Scale IQ 104.0 (9.8) 113.3 (11.2) .ooO1 Oral Reading Grade 2.0 (0.5) 8.4 (2.2) .OOO1 Reading Comprehension Grade 3.1 (1.2) 9.1 (2.5) .OOO1


confounded. We are thus unable to characterize this dyslexic sample in terms of the incidence of additional cognitive deficits.

Language Samples The narrative speech samples were recorded during an experiment examining the EEG and probe-ERP during narrative speech (see Johnstone et ai., 1984, for details of the recording procedures). The EEG data collection placed certain physical constraints on the speakers, who had to sit upright with their bodies and limbs still, keep their eyes fixated on a single point in front of them, and minimize extraneous facial movements. Subjects were instructed to choose a topic on which they could talk continuously for several minutes. Length of speech samples ranged from 137 s to 330 s, with most at about 240 s. Ranges and mean times for both groups were approximately equal. Topics included favorite movies and TV shows, books, and personal experiences. The tape-recorded samples were transcribed verbatim, including all hesitation sounds, fragments of words, and errors. The transcripts were carefully checked for accuracy, and coded so that scorers were blind to the subjects’ classification as dyslexic or control.

Scoring The transcripts were scored on the following features: (1) percent of words in noncommunications, (2) length of communication unit, and (3) rate of syllable production.

(1) Noncommunications Noncommunications are those sounds, words, and phrases which d o not contribute to the expression of the speaker’s intended meaning (Loban, 1%3, 1976). Loban uses the term “mazes” for all of these noncontributory speech fragments. Noncommunications (sepa- rated in the examples from the rest ofthe text by brackets) include hesitation sounds, such as “[uh]” or “[urn]”, meaningless launchers and fillers such as “[okay, well, anywayy and “[you knowy’, repeated material such as “[They all went to the, um,] they all went to the courthouse,” self-corrected fragments such as “[He’s always,] she’s always following him,” and abandoned trains of thought such as, “I hope we [didn’t get that, uh,] go into the championship.” The narrative samples below were selected for the density of illustrative examples, and although they are from the less able speakers, they are not atypical. The following sample includes several types of noncommunications:

[And um he’s, um, there’s, this, how they found out more about this animal is, the legend really began, was,] these two Indians were [urn,] canoeing down the [Missour-, the] Mississippi River [um] just [at daw-,] at daybreak.

Note that the unbracketed text, “These two Indians were canoeing down the Mississippi River just a t daybreak,” is an entirely intelligible utterance.

( 2 ) Communication unit Following the precedent set by Loban (1963), we chose the communication unit (C.U.), as the unit of segmentation. The C.U. is defined as an independent clause and its modifiers, including any dependent clauses. Two independent clauses joined by a con- junction are segmented into two C.U.S. A sample of such segmentation, marked by slashes, follows:

She was running toward the beach/ and he was straggling./ She ran into the water

NARRATIVE SPEECH I

and started swimming./ They didn’t clo! wasn’t gonna close it off./

The following passage, with noncommuni without regard to the underlying incoherenc speaker is describing a visit to the museum e Tut.

Well we went to the [um] museum abou bunch of [um] gold./ [There was, uh, at of the the] the wooden model of his hea gold chamber [about two foot, no,] ab( like sled tracks underneath./ Inside was when, like sled tracks,] then there’s like i it,/ [and] and it had [the, um,] a mode [about,] about six inches high and twc them took it because it was a lot to carry.

In scoring the samples, fragments of word such as “he’s’’ or “doesn’t” were counted i were counted separately from the words in tl of words in communication units was the dit in noncommunications.

Dependent Variables ( I ) Noncommunication Index was defined as divided by the total words. Mean numbe calculated. (2) Communication Unit Length was indf communication units (words in C.U.s divi maximum lengths of undisrupted and error- (3) Rate of Speech Production: Samples wer and out of noncommunications, were count of syllables to derive (i) the overall rate of si and (ii) the rate of effective speech, based only.

Design of Data Analysis The subject population was made up of twc dyslexics and 33 controls, aged 10 to 12. T controls, aged 9 to 13. The first cohort di showed significant group differences were s second cohort data thus served as an inde Finally, effects which had replicated in tt cohorts combined (52 dyslexics, 54 control:

Interscorer reliability was first assessed fi of C.U.s and number of noncommunicat

ET AL.

this dyslexic sample in terms of the

an experiment examining the EEG and et al., 1984, for details of the recording n physical constraints on the speakers, ;till, keep their eyes fixated on a single 11 movements. Subjects were instructed iously for several minutes. Length or at about 240 s. Ranges and mean times icluded favorite movies and TV shows, d samples were transcribed verbatim, ds, and errors. The transcripts were t scorers were blind to the subjects’

: (1) percent of words in noncommuni- te of syllable production.

ihrases which d o not contribute to the an, 1%3, 1976). Loban uses the term gments. Noncommunications (sepa- ,kets) include hesitation sounds, such rs such as “[okay, well, anyway]” and went to the, um,] they all went to the always,] she’s always following him,” we [didn’t get that, uh,] go into the

selected for the density of illustrative speakers, they are not atypical. The inications: t more about this animal is, the legend 3 canoeing down the [Missour-, the]

were canoeing down the Mississippi ‘ance.

ose the communication unit (C.U.), as an independent clause and its

idependent clauses joined by a con- of such segmentation, marked by

i straggling./ She ran into the water

35 1 NARRATIVE SPEECH DEFICITS IN DYSLEXICS

and started swimming./ They didn’t close the beach because the mayor said that he wasn’t gonna close it off./

(4 C.U.s) ‘The following passage, with noncommunications bracketed and segmented into C.U.s

without regard to the underlying incoherence, demonstrates the scoring procedures. The speaker is describing a visit to the museum exhibit of the treasures from the tomb of King Tut.

Well we went to the [um] museum about a month ago with our class/ and we saw a bunch of [um] gold./ [There was, uh, at] the first [saw] thing you saw was [the head of the the] the wooden model of his head,/ and then it came into this [uh, um,] little gold chamber [about two foot, no,] about one foot wide, solid [gold,] silver, [um,] like sled tracks underneath./ Inside was [like, um,] a little, like popsicle stick./ [Urn, when, like sled tracks,] then there’s like a popsicle stick thing, just, [um,] the shape of it,/ [and] and it had [the, um,] a model of her in full life/ and [um it, and] it was [about,] about six inches high and two inches wide/ and the tomb robbers before them took it because it was a lot to carry./

(9 C.U.s) In scoring the samples, fragments of words were counted as half-words and contractions

such as “he’s” or “doesn’t’’ were counted as two words. Words in noncommunications were counted separately from the words in the remainder of the sample, so that the number of words in communication units was the difference between the total words and the words in noncommunications.

Dependent Variables (1) Noncommunication Index was defined as the number of words in noncommunications divided by the total words. Mean number of words per noncommunication was also calculated. (2) Communication Unit Length was indexed by three measures: mean length of all communication units (words in C.U.s divided by number of C.U.s) and the mean and maximum lengths of undisrupted and error-free communication units. (3) Rare of Speech Production: Samples were timed, and the syllables of all words, both in and out of noncomrnunications, were counted. Time in seconds was divided by the number of syllables to derive (i) the overall rate of speech, based on the syllables in the total words, and (ii) the rate of effective speech, based on the syllables in communication unit words only.

Design of Data Analysis The subject population was made up of two independent cohorts. The first consisted of 31 dyslexics and 33 controls, aged 10 to 12. The second cohort had 21 each of dyslexics and controls, aged 9 to 13. The first cohort data was analyzed, and only the features which showed significant group differences were scored on the second cohort. The analysis of the second cohort data thus served as an independent cross-validation of the initial findings. Finally, effects which had replicated in the second cohort were recalculated on the two cohorts combined (52 dyslexics, 54 controls) to yield our best estimate of effect size.

Interscorer reliability was first assessed for each of the six independent variables: number of C.U.s and number of noncommunications; number of words, both in C.U.s and in


noncommunications; and, number of syllables, both in C.U.s and in noncommunications. We computed the intraclass correlation (Winer, 1971) of three scorers’ independent scoring of the same randomly selected subset of five dyslexic and five control language samples. For each variable, we calculated the reliability of the average of three scorers and the estimated reliability of a single scorer.

Dyslexic and control groups were compared using t tests. One-tailed tests were used for the cross-validation in the second cohort, two-tailed tests for the first cohort and for the combined cohorts. The Mann-Whitney U Test (Mosteller & Rourke, 1973) was used in lieu of the t test for those variables with significantly large differences in the variance between the two groups, as assessed by the F-Max test (Winer, 1971). Because the noncommunication index is a proportion, data were transformed using z = arcsin dp before analysis (Cohen, \977).

Since our first cohort was larger than the second, and since p values are affected by sample size, size of effect was calculated to facilitate comparison between the two cohorts. Effect size was calculated as the difference between the means, divided by theaverage of the standard deviations for the two groups (Cohen, 1977). This effect size measure presents group differences in standard deviation units.

NARRATIVE SPEECI

RESULTS

Interrater reliability for the average of three scorers ranged from .985 to .999 for the six independent variables. Estimates of the reliability of a single scorer ranged from .959 to .998. Because interrater reliability was so high, the remaining transcripts were scored by a single scorer only.

1. Noncommunication Index Words in noncommunicative utterances made up a larger proportion of the speech of dyslexics than of controls in each cohort, with comparable effect sizes. The data, summarized here, are presented in full in Table 2. In the combined cohorts, dyslexics’ noncommunication words made up 16.13% of the language sample,ascompared to 12.51% forthecontrols(p < .005,effectsize[e.s.]=0.60). In the first cohort, dyslexics’ mean length of noncommunicative utterance was 2.59 words, compared to 2.12 words for the controls (p < .005, e.s. = 0.76). This finding did not replicate in the second cohort, so we did not recalculate this measure on the combined cohorts.

2. Communication Unit Length Dyslexics of both cohorts produced significantly shorter communication units than did controls in all three measures of communication unit length (see Table 3). Effect sizes were larger in the first cohort than in the second in two of the measures; however, even the second cohort showed more than a medium effect size. In the combined data set, the dyslexics had 8.79 words per C.U.; the controls, 9.90 (p < .OOO1. e.s. ~ 0 . 8 3 ) . Communication units which were free of noncommunications

-I

Noncommunication Meas Cohort 1, Cohort 2

Dyslexic (n = 31) Control (n = 33)


A. NONCOMMUNICATION INDEX: Words in Noncommunicative Utterances,

co Me. 16. 12.

co Me 15. 12.

COMBIN Me 16. 12.


B. AVERAGE LENGTI



0 NONCOB cc Me 2. : 2. cc MC 2.1 2.1

COMBIb (Not calculated di

* e.s. = effect size in standard deviation

and errors were considered separatel! of the combined cohorts averaged 7 controls (p < .00025, e.s. = 0.72). communication unit averaged 16.85 \ the controls (p < .005, e.s. = 0.61) fi large-effect sizes (Cohen, 1977) reffec the narrative speech of these intelligei

)RT ET AL.

loth in C.U.s and in noncornrnunications. ter, 1971) of three scorers’ independent 3f five dyslexic and five control language iability of the average of three scorers and

ising t tests. One-tailed tests were used for :ailed tests for the first cohort and for the Mosteller & Rourke, 1973) was used in lieu I large differences in the variance between Kiner, 197 1). Because the noncommunica- n d using z = arcsin v’p before analysis

xond, and since p values are affected by tate comparison between the two cohorts. en the means, divided by the average of the I, 1977). This effect size measure presents

TS

e scorers ranged from .985 to .999 for the reliability of a single scorer ranged liability was so high, the remaining 1ly.

made up a larger proportion of the i cohort, with comparable effect sizes. d in full in Table 2. In the combined xds made up 16.13% of the language trois@ <.005,effectsize[e.s.]=0.60). of noncommunicative utterance was

le controls (p < .005, e.s. = 0.76). This ohort, so we did not recalculate this

ficantly shorter communication units )mmunication unit length (see Table 3). an in the second in two of the measures; nore than a medium effect size. In the vords per C.U.; the controls, 9.90 (p < hich were free of noncommunications

353 NARRATIVE SPEECH DEFICITS IN DYSLEXlCS

Table 2

Noncornmunication Measures in Dyslexics and Controls Cohort 1, Cohort 2, and Combined Cohorts

A. NONCOMMUNICATION INDEX: Words in Noncommunicative UtteranceslTotal Words in Sample

COHORT 1 Mean SD P <

.01 Dyslexic (n = 31) 16.74 (6.17) Control (n = 33) 12.71 (5.60)

COHORT 2

Mean SD P <


COMBINED COHORTS

Mean SD P <

.005 Dyslexic (n = 52) 16.13 (6.29)

B. AVERAGE LENGTH OF NONCOMMUNICATIVE UTTERANCE:

Control (n = 54) 12.51 (5.73)

COHORT 1 Mean SD P <


COHORT 2

Mean SD P <

(0.60) N.S. Dyslexic (n = 21) 2.08 Control (n = 21) 2.00 (0.41)

COMBINED COHORTS

(Not calculated due to failure to replicate)

e.s.*

0.685

e.s.*

0.485

e.s.*

0.602

e.s.*

0.758

e.s.*

0.016

* e.s. = effect size in standard deviation units

and errors were considered separately. In these uninterrupted C.U.s, the dyslexics of the combined cohorts averaged 7.37 words, compared to 8.27 words for the controls (p < .00025, e.s. = 0.72). The longest uninterrupted and error-free communication unit averaged 16.85 words in the dyslexic group vs. 20.53 words in the controls (p < .005, e.s. = 0.67) for the combined cohorts. These medium- to largeeffect sizes (Cohen, 1977) reflect deficits which are definitely conspicuous in the narrative speech of these intelligent, but dyslexic, children.

354 LINDA DAVENPORT ET AL. ,

Table 3

Measures of Communication Unit Length in Dyslexics and Controls

A. C.U. LENGTH: (Total Words - Words in Noncommunicativc Utterances) Number of C.U.s

COHORT 1 Mean SD

Dyslexic (n = 31) 8.68 (1.01) Control (n = 33) 9.90 (1.34)

COHORT 2

Mean SD

Dyslexic (n = 21) 8.95 (1.32)

COMBINED COHORTS Control (n = 21) 9.90 (1.84)

Mean SD


B. LENGTH OF UNDISRUFTED C.U.s Words in undisrupted C.U.sflVumber of undisrupted C.U.S

COHORT 1 Mean SD


COHORT 2

Mean SD


COMBINED COHORTS

Mean SD


C. LONGEST UNDISRUPTED C.U. COHORT 1 Mean SD


P <

.oOo1

P <

.05

P <

.oO01

P <

.oO05

P <

.06

P <

.oOo2

P <

.02

e.s.*

1.038

e.s.*

0.601

e.s.*

0.831

e.s.*

0.949

e.s.*

0.536

e.s.*

0.723

e x *

0.657

NARRATIVE SPEEC

CC

M' Dyslexic (n = 21) 15 Control (n = 21) 19

COMBIP

MI

Dyslexic (n = 52) 16 Control (n = 54) 20

e.s. = effect size in standard deviation

To test whether these measures of n; product-moment correlation was ca index and C.U. length, for the dy! combined. The results showed little re r = +.Ol; controls r = -.15; combine that each measure contributes uniq narrative speech.

3. Rate of Speech Production Rate of speech vaned widely among producing syllables at almost two-and. first cohort, neither the mean rate, mi1 dyslexics and the controls (controls' I

1.47, max = 3.48; dyslexics' mean rate 3.25). Therefore, this measure was not should be noted that, because dyslexics noncommunicative utterances, their ri controls. That is to say, they take longt

IQ and Reading Ability in Relation to th Our control group includes some boys that of the dyslexics, but was also we1 Similarly, although the dyslexics and c VIQs were significantly higher than whether the group differences in narra ability of the controls rather than to a c

To examine whether the differenct length and the noncommunication in readers in the control group, we isola reading scores were average, Le., withii We compared the dyslexics to these

rETAL. ,

h in Dyslexics and Controls

:rances) Number of C.U.s

SD

(1.01) 0.34)

SD (1.32)

IORTS SD

(1.14) (1.53)

(1.84)

ted C.U.s

SD

(0.82) (1.14)

!

SD

(2.15)

IORTS

(1.02)

SD

(0.90) (1.59)

1 SD

(3.74) (6.92)

P <

.oO01

P <

.05

P <

.oO01

P <

.oO05

P <

.06

P <

.oO02

P <

.02

e.s.*

1.038

e.s.*

0.601

e.s.*

0.83 1

e.s.*

0.949

e.s.*

0.536

e.s.*

0.723

e.s. *

0.657


355 NARRATIVE SPEECH DEFlClTS IN DYSLEXICS

COHORT 2 Mean SD P < e.s.*

15.57 19.48

(3.57) (7.83) .025 0.686

COMBINED COHORn Mean SD P < e.s.*

.@I5 0.666 Dyslexic (n = 52) 16.85 (3.79) Control (n = 54) 20.53 (7.26)

* e.s. = effect size in standard deviation units

To test whether these measures of narrative speech were independent, a Pearson product-moment correlation was computed between the noncommunication index and C.U. length, for the dyslexic and control groups separately and combined. The results showed little relationship between the measures (dyslexics r = +.01; controls r = -. 15; combined r -. 18; all n.s.). Therefore, we conclude that each measure contributes unique information characterizing the child’s narrative speech.

3. Rate of Speech Production Rate of speech varied widely among these subjects, with the fastest speakers producing syllables at almost two-and-one-half times the rate of the slowest. In the first cohort, neither the mean rate, minimum, nor maximum differed between the dyslexics and the controls (controls’ mean rate = 2.54 syl/s, SD = 0.47; min = 1.47, max= 3.48; dyslexics’ mean ratez2.43 syl/s, SDz0.41; min = 1.79, max = 3.25). Therefore, this measure was not repeated on the second cohort. However, it should be noted that, because dyslexics have a higher proportion of their words in noncommunicative utterances, their rate of effective speech is lower than that of controls. That is to say, they take longer to say what they really mean.

IQ and Reading Ability in Relation to the Language Measures Our control group includes some boys whose reading was not only far superior to that of the dyslexics, but was also well above their own actual grade placement. Similarly, although the dyslexics and controls did not differ in PIQ, the controls’ VIQs were significantly higher than the dyslexics’. This raises the question of whether the group differences in narrative speech might be due to the exceptional ability of the controls rather than to a deficit in the dyslexics.

To examine whether the differences between dyslexics and controls in C.U. length and the noncommunication index were due to the presence of superior readers in the control group, we isolated a subset of control subjects whose oral reading scores were average, Le., within 1 year of actual grade placement (n = 21). We compared the dyslexics to these average readers on the narrative speech

356 LINDA DAVENPORT ET AL. NARRATIVE SPEECI

Table 4

Communication Unit Length and Noncommunication Index in Dyslexics Compared to Average Readers

DISC

C.U. LENGTH NONCOMMUNICATION INDEX

Mean SD p < e.s.* Mean SD p < e.s.* Dyslexics (n = 52) 8.79 1.14 16.13 6.26 Average Readers .05 0.577 .Ol 0.636 (n = 21) 9.60 1.67 12.62 4.78

measures, and found that differences in C.U. length were still significant (p < .05), with an effect size of .58, and that differences in the noncommunication index were still significant (p < .Ol), with an effect size of .64 (see Table 4). These effect sizes were comparable to those of the entire control group vs. the dyslexics. To make sure that this confirmation did not depend on the choice of oral reading as the index of superior performance, we used two alternative measures to define “average reading”: (a) comprehension scores, and (b) oral and comprehension combined. In each case, comparison of the average readers yielded the same result. Therefore, we conclude that the dyslexic/control differences did not depend on the superior readers in the control group.

To examine whether the group differences were due to the controls with higher VI@, we compared 22 dyslexics with 22 controls who were exactly matched on VIQ. VIQ scores for these pairs ranged from 91 to 117. Matched-pair t tests were used. Differences between dyslexics and VIQ-matched controls in C.U. length were at the .01 level of significance, with an effect size of .69; differences in noncommunication index were at the .05 level, with an effect size of .59 (see Table 5). These effect sizes are still in the medium-to-large range. Therefore, we conclude that dyslexic/control differences did not depend on the superior VIQs in the control group.

Table 5

Communication Unit Length and Noncornmunication Index in Dyslexics and Controls Who Are Matched on Verbal IQ

C.U. LENGTH NONCOMMUNICATION INDEX

Mean SD p < e.s.* Mean SD p < e.s.*

Dyslexics (n = 22) 9.04 1.23 15.74 6.92 Controls .OI 0.691 .OS 0.590 (n = 22) 10.16 2.01 12.01 5.73

This group of carefully selected dyslexi normal readers in two measures of narr words in noncommunicative utteranc independent sample of equally well-sc shown to be independent of each 0th control differences in narrative speech VIQ or reading ability of some membei

Even though other investigations h language deficits in dyslexics, our finc particular theoretical importance due tc ensuring the normality of clinical neur emotional status of the dyslexics, we relate to dyslexia itself, rather than conditions commonly found in reading As for the nature of the relationshir

deficits, the two simplest and most like secondary to dyslexia, and (b) that language deficit which also produce: possible argument in favor of model deprived of a major source of modeli therefore less effective speakers. Howe in narrative speech in the less able I

against this model. At this grade lev material is still substantially below th; readers could not yet have missed very ability in reading and narrative speeck levels. We therefore reject the idea tha dyslexia, and favor the model that b deficits reflect a more generalized imp;

The practical importance of our finc in the dyslexics which needs to be rei above, from the account of a visit to t boy who had been previously identifi tion had been limited to training in rea his persistent difficulty in learning communication needs expert attent noncommunication index were not cc tion should be planned according dysfunctions.

The deficits reported here are defin these intelligent, but dyslexic, childn

TAL.

tmrnunication Index in Ige Readers

NONCOMMUNICATION INDEX

Mean SD p < e.s.*

16.13 6.26

12.62 4.78 .01 0.636

gh were still significant (p < .05), ie noncommunication index were 14 (see Table 4). These effect sizes group vs. the dyslexics. To make :he choice of oral reading as the alternative measures to define

tnd (b) oral and comprehension ze readers yielded the same result. trol differences did not depend

re due to the controls with higher 11s who were exactly matched on to 117. Matched-pair t tests were natched controls in C.U. length effect size of .69; differences in ith an effect size of 3 9 (see Table ge range. Therefore, we conclude id on the superior VIQs in the

mmunication Index in .ched on Verbal IQ

NONCOMMUNICATION INDEX

Mean SD p < e.s.*

15.74 6.92

12.01 5.73 .os 0.590

NARRATIVE SPEECH DEFICITS IN DYSLEXICS 357

DISCUSSION

This group of carefully selected dyslexic boys was found to differ from a group of normal readers in two measures of narrative speech: length of C.U. and percent of words in noncommunicative utterances. These findings were replicated in an independent sample of equally well-screened subjects. The two measures were shown to be independent of each other. We further showed that the dyslexic/ control differences in narrative speech are not an artifact of the high to superior VIQ or reading ability of some members of this particular control group.

Even though other investigations have already documented several kinds of language deficits in dyslexics, our finding of narrative speech impairment is of particular theoretical importance due to the stringency of our screening criteria. By ensuring the normality of clinical neurological, sensory, general intellectual, and emotional status of the dyslexics, we increased the likelihood that our findings relate to dyslexia itself, rather than to low IQ or to any of the overlapping conditions commonly found in reading-disabled populations. As for the nature of the relationship between dyslexia and the narrative speech

deficits, the two simplest and most likely models are (a) that the speech deficits are secondary to dyslexia, and (b) that both are symptomatic of a generalized language deficit which also produces the impairments reported elsewhere. A possible argument in favor of model (a) is that children who cannot read are deprived of a major source of modeling in the use of formal language, and are therefore less effective speakers. However, the Fry et al. findings (1970) of deficits in narrative speech in the less able readers of the second grade, would argue against this model. At this grade level, the linguistic sophistication of reading material is still substantially below that of oral language, and therefore the poor readers could not yet have missed very much, if anything. Loban (1976) also found ability in reading and narrative speech to be associated, even at the earliest grade levels. We therefore reject the idea that narrative speech deficits are secondary to dyslexia, and favor the model that both the dyslexia and the narrative speech deficits reflect a more generalized impairment in language.

The practical importance of our findings is that they reveal an area of weakness in the dyslexics which needs to be remediated. For instance, the exerpt, quoted above, from the account of a visit to the museum, was produced by a 12-year-old boy who had been previously identified only as dyslexic. His extensive remedia- tion had been limited to training in reading and spelling skills. Especially in light of his persistent difficulty in learning to read and write, his disability in oral communication needs expert attention. Our finding that C.U length and noncommunication index were not correlated suggests that educational interven- tion should be planned according to the specific nature of oral language dysfunctions.

The deficits reported here are definitely conspicuous in the narrative speech of these intelligent, but dyslexic, children. In addition to considering significance

.

I

358 LINDA DAVENPORT ET AL. NARRATIVE SPEECI

levels, we find the quantitative measure of effect size to be useful in putting these results in perspective (Cohen, 1977). Statistical significance depends on sample size, so a small and clinically meaningless effect can be highly “significant” if then is large enough. The level of significance by itself gives no information about the magnitude of the differences encountered. On the other hand, effect size is independent of sample size, assuming it is large enough to not misrepresent the variance in the population. Effect sizes of .5 SD are considered medium, whereas effect sizes of .8 SD are large and obvious, for example the difference in mean IQ between “holders of the Ph.D. degree and typical college freshmen” (Cohen, 1977). Our finding of an effect size of .83 1 in C.U. length is therefore a large and clinically relevant effect, not just a “significant” one.

Significance of Length of Communication Units C.U. length is expected to increase throughout the years of schooling, with mean increments of about one-half word per C.U. per year in children in the age range of our sample (Loban, 1976). The dyslexics’ shortfall of more than one word per C.U. therefore indicates about a 2-year lag behind the controls. Loban found at all age levels an association between length of C.U.s and effectiveness in the use of phrases and clauses, with longer C.U.s characterized by an increased use of “other forms of elaboration contributing to clear and meaningful communication.” This is not to imply that longer is always better; the effectiveness of a communication depends on a number of factors, including the pragmatic aspect of appropriate use of language in the social context (Bates, 1976). However, neither our study nor Loban’s allows this dimension to be addressed, since both studies analyzed narrative monologues rather than dialogues. Within the context of our study, we belie-re that the shorter C.U. length of these dyslexics reflects reduced effectiveness of their efforts to communicate clearly and meaningfully.

We were unable to test this hypothesized relation between length of C.U. and effectiveness of communication, because effectiveness is a function of the nature and complexity of the topic as well as of the ability of the speaker. The topics which the boys had selected varied widely in their complexity, with some evidently much more difficult to describe clearly than others. An ideal evaluation of narrative speech would include a direct and quantitative measure of communication effectiveness, using a scale derived from a set of uniform language samples. In the absence of such a measure, we propose that length of C.U. is a valuable, although indirect, indicator of effectiveness. Accordingly, our findings suggest that dyslexics are less skillful at expressing their ideas with clarity and precision than are their normal-reading peers.

Significance of Noncommunication Index One striking difference between spoken and written language is that the written form represents the writer’s final effort, with all of the false starts and errors eliminated, whereas in unrehearsed spoken language we have the chance to

” u-serve the details of the process by wt The hesitations, false starts, and error which we summarize as noncommun takes to translate ideas into language.

It is important to distinguish betwec cation, both of which interfere with a 1 By “coherence” we mean the precisior of organization, within and among sen speaker’s line of thought. A child who have a low noncommunication inde another, more alert to his own languag of noncommunicative utterances on tk clarity. There is at present no adequate

Noncommunicative utterances may stalling, either to organize his though syntactic structure, (b) when he is thin initial fragments of sentences until he c direction, (c) when he self-corrects ar usage, and (d) when he wishes to ir enhance the precision, accuracy, or COI

It is possible that particular functior might reflect dysfunctions in specific relationship would depend on a greate speech deficits than we yet have even f for classifying noncommunications b! uniform content in language samplir tiveness, and of classifying noncomrr this kind would substantially enhar speech deficit in dyslexia, and would lying mechanisms.

REF;

Bates, E. (1976). Language and context. P Brown, B., Haegerstrom-Portnoy, G., Ad;

& Marcus, M. (1983). Predictive eye m and control children. Neuropsychologia

Brown, B., Haegerstrom-Portnoy, G., Yin (1983a). Dyslexic children have norm Neurology. 40, 370-373.

Brown, B., Haegerstrom-Portnoy, G., Yin (1983b). Tracking eye movements are I

Optometry and Physiological Optics, 61

’ ET AL. NARRATIVE SPEECH DEFICITS IN DYSLEXICS 359

:ct size to be useful in putting these :a1 significance depends on sample t can be highly “significant” if then self gives no information about the On the other hand, effect size is -ge enough to not misrepresent the ;D are considered medium, whereas . example the difference in mean IQ :a1 college freshmen” (Cohen, 1977). gth is therefore a large and clinically

b

it the years of schooling, with mean :r year in children in the age range of ortfall of more than one word per iind the controls. Loban found at all :.Us and effectiveness in the use of terized by an increased use of “other I meaningful communication.” This le effectiveness of a communication pragmatic aspect of appropriate use 5). However, neither our study nor s e d , since both studies analyzed Within the context of our study, we rslexics reflects reduced effectiveness :aningfully . elation between length of C.U. and ctiveness is a function of the nature e ability of the speaker. The topics qeir complexity, with some evidently an others. An ideal evaluation of pantitative measure of communica- set of uniform language samples. In that length of C.U. is a valuable, Accordingly, our findings suggest

heir ideas with clarity and precision

written language is that the written th all of the false starts and errors

language we have the chance to

observe the details of the process by which the speaker achieves the final utterance. The hesitations, false starts, and errors made on the way to the final utterance, which we summarize as noncommunications, serve as evidence of the effort it takes to translate ideas into language.

It is important to distinguish between coherence and amount of noncommunication, both of which interfere with a listener’s ability to understand what is said. By “coherence” we mean the precision of expression, and particularly the clarity of organization, within and among sentences which enable a listener to follow the speaker’s line of thought. A child who fails to monitor and correct his output may have a low noncommunication index but be quite difficult to follow, while another, more alert to his own language production, may falter through a number of noncommunicative utterances on the way to a final formulation of substantial clarity. There is at present no adequate measure of coherence per se.

Noncommunicative utterances may arise in several ways: (a) when the speaker is stalling, either to organize his thoughts, or to retrieve a word or an appropriate syntactic structure, (b) when he is thinking aloud in an exploratory way, uttering initial fragments of sentences until he chooses one which will take him in the right direction, (c) when he self-corrects an error in pronunciation, word choice, or usage, and (d) when he wishes to improve on what he has said correctly, to enhance the precision, accuracy, or completeness of the final utterance.

It is possible that particular functional classes of noncommunicative utterances might reflect dysfunctions in specific neural subsystems. To establish such a relationship would depend on a greater consensus on the localizing significance of speech deficits than we yet have even for adult aphasics, and on a reliable method for classifying noncommunications by functional type. Future studies which use uniform content in language sampling could develop methods of rating effectiveness, and of classifying noncommunications by functional types. Studies of this kind would substantially enhance the characterization of the narrative speech deficit in dyslexia, and would advance our understanding of its underlying mechanisms.

REFERENCES

Bates, E. (1976). Language and conrext. New York: Academic Press. Brown, B., Haegerstrom-Portnoy, G., Adams, A. J., Yingling, C. D., Galin, D., Herron, J.,

& Marcus, M. (1983). Predictive eye movements do not discriminate between dyslexic and control children. Neuropsychologia, 21, 121-128.

Brown, B., Haegerstrom-Portnoy, G., Yingling, C. D., Herron, J., Galin, D., & Marcus, M. (1983a). Dyslexic children have normal vestibular responses to rotation. Archives of Neurology. 40, 310-373.

Brown, B., Haegerstrom-Portnoy, G., Yingling, C. D., Herron, J., Galin, D., &Marcus, M. (1983b). Tracking eye movements are normal in dyslexic children. American Journal of Optometry and Physiological Optics, 60. 376-383.

360 LINDA DAVENPORT ETAL.

Cohen, J. (1977). Statisticalpower analysis for the behavioral sciences (rev. ed.). New York: Academic Press.

Denckla, M. B. (1979). Childhood learning disabilities. In K. M. Heilman & E. Valenstein (Eds.), Clinical neuropsychology (pp. 535-573). New York: Oxford University Press.

Doehring, D. G., & Hoshko, I. M. (1977). Classification of reading problems by the Q-technique of factor analysis. Cortex, 13. 281-294.

Fein, G., Galin, D., Johnstone, J., Yingling, C. D., Marcus, M., & Kiersch, M. E. (1983). EEG power spectra in normal and dyslexic children: I. Reliability during passive conditions. Electroencephalography and Clinical Neurophysiology, 55, 399-405.

Fein, G., Galin, D., Yingling, C. D., Johnstone, J., & Herron, J. (1986). EEG spectra in dyslexic and control boys during resting conditions. Electroencephdography and Clinical Neurophysiology, 63, 81-97.

Fein, G., Galin, D., Yingling, C. D., Johnstone, J., &Nelson, M. A. (1984). EEG spectra in 9 to 13 year old boys are stable over one to three years. Electroencephalography and Clinical Neurophysiology, 58, 158- 160.

Fry, M. A., Johnson, C. S., & Muehl, S. (1970). Oral language production in relation to reading achievement among select second graders. In D. J. Bakker & P. Satz (Eds.), Specific reading disability: Advances in theory andmethod@p. 123-146). Rotterdam: Rot- terdam University Press.

Gates, A. I., & MacGinitie, W. H. (1%5). Gates-MacGinitie Reading Tests. New York: Teachers College Press.

Gray, W. F. (1963). Gray Oral Reading Test. Indianapolis: Bobbs-Memill. Jastak, J. R., & Jastak, S. R. (1965). 7he Wide Range Achievement Test. Arlington, DE:

Guidance Associates of Delaware. Johnstone, J., Galin, D., Fein, G., Yingling, C. D., Herron, J., & Marcus, M. (1984).

Regional brain activity in dyslexic and control children during reading tasks: Visual probe event-related potentials. Brain and Language, 21, 233-254.

Loban, W. D. (1963). The language of elementary school children. Urbana, IL: National Council of Teachers of English.

Loban, W. D. (1976). Language development: Kindergarten through grade 12. Urbana, IL: National Council of Teachers of English.

Malatesha, R. N., & Dougan, D. R. (1982). Clinical subtypes of developmental dyslexia: Resolution of a n irresolute problem. In R. N. Malatesha & P. G. Aaron (Eds.), Reading disorders: Varieties and treatments @p. 69-92). New York: Academic Press.

Mattis, S. (1978). Dyslexia syndromes: A working hypothesis that works. In A. L. Benton & D. Pearl (Eds.), Dyslexia: An appraisal of current knowledge (pp. 43-58). New York: Oxford University Press.

Mattis, S., French, J. H., & Rapin, I. (1975). Dyslexia in children and adults: Three independent neuropsychological syndromes. Developmental Medicine and Child Neurology. 17, 150-163.

Mosteller, F., & Rourke, R. E. (1973). Sturdy statistics: Nonparametrics andorder statistics. Reading, MA: Addison-Wesley.

Myklebust, H. R. (1968). Progress in learning disabilities. New York: Grune and Stratton. Satz, P., & Morris, R. (1981). Learning disability subtypes: A review. In F. Pirozzolo & M. Wittrock (Eds.), Neuropsychological and cognitiveprocesses in readtng @p. 109-141). New York: Academic Press.

Touwen, B. C. L., & Prechtl, H. F. R. (1970). The neurological examination of the child

NARRATIVE SPEEC

with minor nervous dysfunction. Topics Vellutino, F. (1979). Dyslexia: Theory and Watson, B. U., Goldgar, D. E., & Ryscho

Journal of Clinical Neuropsychology, 5. 1 Wechsler, D. (1974). Wechsler Intelligent

York: The Psychological Corporation. Winer, B. J. (1971). Statistical principres

McGraw-Hill. Yingling, C. D., Galin, D., Fein, G., & Pel

event-related potentials in dyslexic and nor ed at the meeting of the International h

Yingling, C. D., & Galin, D. (1982). Neur NICHD Contract NOI-HD-8-2824, Un

' AL.

wioralsciences (rev. ed.). New York:

s. In K. M. Heilman &E. Valenstein N York: Oxford University Press. ication of reading problems by the 1. narcus, M., & Kiersch, M. E. (1982). n: I. Reliability during passive condi-

k Herron, J. (1986). EEG spectra in itions. Electroencephalography and

I Nelson, M. A. (1984). EEG spectra years. Electroencephalography and

bysiology, 55, 399-405.

11 language production in relation to 5. In D. J. Bakker & P. Satz (Eds.), ethodbp. 123-146). Rotterdam: Rot-

(accinitie Reading Tests. New York:

iapolis: Bobbs-Merrill. ge Achievement Test. Arlington, DE:

., Herron, J., & Marcus, M. (1984). hildren during reading tasks: Visual le. 21. 233-254. chool children. Urbana, IL: National

pr ten through grade 12. Urbana, IL:

subtypes of developmental dyslexia: Malatesha & P. G. Aaron (Eds.),

19-92). New York: Academic Press. ypothesis that works. In A. L. Benton mt knowledge @p. 43-58). New York:

:xia in children and adults: Three in- vmental Medicine and Child Neurology,

cs: Nonparame trics and order statistics.

kries. New York: Grune and Stratton. ubtypes: A review. In F. Pirouolo & ;rive processes in reading @p. 109- 141).

neurological examination of the child

NARRATIVE SPEECH DEFICITS IN DYSLEXICS 36 1

with minor nervous dysfunction. Topics in Developmental Medicine, 30, 1- 105. Vellutino, F. (1979). Dyslexia: Theory and research. Cambridge: MIT. Watson, B. U., Goldgar, D. E., & Ryschon, K. L. (1983). Subtypes of reading disability.

Journal of Clinical Neuropsychology, 5, 31 7-399. Wechsler, D. (1974). Wechsler Intelligence Scale for Children-Revised (WISC-R). New

York: The Psychological Corporation. Winer, B. J. (1971). Statistical principles in experimental design (2nd ed.). New York:

McGraw-Hill. Yingling, C. D., Galin, D., Fein, G., & Peltzman, D. (1985, February). Multiple modality

event-related potentials in dyslexic and normal children: A replication study. Paper presented at the meeting of the International Neuropsychological Society, San Diego, CA.

Yingling, C. D., & Galin, D. (1982). Neuropsychology of reading disorders. Final report: NICHD Contract NO 1-HD-8-2824, University of California, San Francisco.

Documents

Narrative Speech Deficits Dyslexics* · L), Late effects of head injury, (pp. 501-524). brain syndromes: An introduction to neuro- e. Thehncet, 1. 1-5. idental head injuries: Prognosis