Journal of Speech and HearingResearch, Volume 33, 70-83, March 1990

DEVELOPMENT OF CHILDREN WITH EARLY LANGUAGE DELAY HOLLIS S. SCARBOROUGH Brooklyn College of the City University of New York WANDA DOBRICH Rutgers the State University of New Jersey Four children with early language delays (ELD) were compared to a control group of 12 children with respect to their preschool language abilities from age 2 1/2 to 5 years and their verbal skills at the end of Grade 2. The language-delayed children each initially showed severe and broad impairments in syntactic, phonological, and lexical production. Over time, their deficits became milder and more selective, such that normal or nearly normal speech and language proficiency was exhibited by age 60 months. Nevertheless, when followed up 3 years later, three of the four cases were severely reading disabled. These findings are discussed with respect to prior findings and hypotheses about the sequelae of early language delay and the relationship of language development to reading achievement. KEY WORDS: language delay, reading disability, language development

et al., 1984). That is, even children who reach normal levels of skill by about age 5 years may remain at risk for subsequent language difficulties or reading disabilities at a later time. This situation would be especially likely to occur if normal language proficiency grows in a stepwise manner, with spurts of rapid growth separated by plateau periods of relatively little change in performance (Scarborough & Dobrich, 1985a). If ELD children undergo delayed but normal acquisition of language, they could at times appear to catch up to their normal agemates while the latter remain on an extended plateau, as illustrated in Figure 1. According to this "illusory recovery" model, differences would be expected to reemerge once the normal children underwent the next developmental spurt. The present study grew out of a much broader exarni-

What happens to children with early language delay

(ELD)? Two empirical phenomena relevant to this question have been well-documented in longitudinal studies. Short-term follow-up evaluations over the preschool period have shown that many ELD children achieve normal levels of language proficiency by about age 5 to 6 years (Bishop & Edmundson, 1987; MacKeith & Rutter, 1972; Morley, 1972; Silva, 1980). The interpretation of this evidence for recovery from ELD is unclear, however, in light of strong evidence that longer-term outcomes may often be less favorable (ASHA, 1982; Bashir, Wiig, & Abrams, 1987; Wallach & Butler, 1984). That is, it has been shown that in later childhood and beyond, from 28% to 75% of children whose preschool language was impaired exhibit residual language/speech problems, and from 52% to 95% show impairments in reading achievement (Aram, Ekelman, & Nation, 1984; Aram & Nation, 1980; Fundudis, Kolvin, & Garside, 1979; Hall & Tomblin, 1978; Levi, Capozzi, Fabrizi, & Sechi, 1982; Padgett, 1988; Stark et al., 1984; Strominger & Bashir, 1977). How can these findings-that both short-term recovery and long-term persistence are likely outcomes of ELDbe reconciled? One straightforward possibility is that ELD cases may differ in their severity or subtype, and that such diagnostic differences may determine whether recovery or persistence will ensue for particular individuals. In support of this notion, it has been observed that children with milder delays and children with a narrower range of problems (e.g., isolated phonological deficits) are less likely to exhibit persistent language or reading problems (Bishop & Edmundson, 1987; Hall & Tomblin, 1978; King, Jones, & Lasky, 1982; Levi et al., 1982). Alternatively, it is possible that both recovery from and persistence of language impairments may occur in the same child at different points in development (Bishop & Edmundson, 1987; Scarborough & Dobrich, 1985b; Stark C 1990, American Speech-Language-Hearing Association

hal yed :

'Illusory' recovery

/

c C:

2

3

4

5

6

age (years) FIGURE 1. Model of the relationship between normal and delayed language development (Scarborough & Dobrich, 1985b). 70

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

0022-4685/90/330 1-0070$01.00/0

SCARBOROUGH & DOBRICH: Early Language Delay nation of the nature and consequences of individual differences in early language proficiency, and thus was not originally designed as a study of ELD children. Having collected longitudinal data on a large sample of preschool children, however, we later discovered that several subjects met conventional criteria for the diagnosis of language delay. We could thus analyze their preschool language development in comparison to that of normal children for whom development was recorded under identical circumstances. Later, when the children were reevaluated at the end of Grade 2, their reading and language skills could be examined with respect to hypotheses about the longer-term consequences of early language delay. Four case histories of ELD children from age 2 1/2 to 8 years will be presented.

METHOD

Subjects All subjects were participants in a larger investigation of the long-term scholastic consequences of early individual differences in language abilities in a sample of 88 children residing in central New Jersey. In this report, 4 language-delayed children will be compared to 12 subjects with normal language development from the larger group. All subjects were from monolingual homes where Standard English is spoken. According to Hollingshead and Redlich's (1958) socioeconomic (SES) scale based upon parental education and occupational prestige, all of the subjects were from lower- to upper-middle class backgrounds (strata II, III, and IV). In addition, all were of normal IQ (as will be documented below), and have no gross neurological, hearing, or uncorrected visual problems, according to pediatric and audiological examinations and research observations between the ages of 2 and 8 years. Control group. There were 6 boys and 6 girls in the control group. Their selection was made without knowledge of the children's linguistic abilities; these subjects were simply the first available cases who met sex, socioeconomic, and birth order criteria for inclusion in a representative normal sample. Six of the children were firstborn, and mean SES of the group was 2.42 (SD 0.9). This group has been described previously with respect to the development of utterance length (Scarborough, Wyckoff, & Davidson, 1986). ELD subjects. The 4 focal subjects of this report were 1 girl and 3 boys who will be identified by the fictitious names Lois, Roger, Dan, and Martin. The absence of regular word combinations at age 30 months (Ingram, 1972) was used as the criterion for selecting these ELD cases from the total of 88 cases in the larger longitudinal database. When the language production of all subjects at 30 months of age was reviewed, these 4 subjects were

71

found to meet this criterion.' Dan and Martin were firstborn, and Lois and Roger were later born. Mean SES of this group was 3.0 (SD 0.8), which was not significantly lower than that of the control group (t = 1.37, p > .20). In other respects also, the backgrounds and early developmental histories of the ELD cases were comparable to those of the control subjects, except that all but Martin were from families in which one or more parent or sibling had a severe reading disability (i.e., low reading achievement in relation to IQ). This circumstance will be discussed more fully below. Roger came to our attention at age 30 months, when the examiner noticed that his language skills were poor. When informed of this, his parents dismissed our concerns on the grounds that his older brother, then 5 years old, had been "much worse," and had "turned out fine." (As might be expected, a few years later the older brother's reading disability led to his classification by the school system.) Somewhat surprisingly, in no other case did the examiners who carried out the testing and video recording (all of whom were graduate students in clinical, developmental, or educational psychology) notice any language problems at an early age, nor did parents express any concerns. Not until several years later, when their video recorded language samples were routinely transcribed and analyzed, were the other 3 children recognized as having been delayed in early language acquisition. By that time, as will be seen, their performance had improved considerably. Evaluations by the schools when the children entered kindergarten also did not identify any language impairments in these four cases. Procedures Preschool assessments were made at ages 24, 30, 36, 42, 48, and 60 months of age. Follow-up evaluations were conducted when the children had completed Grade 2, at which time their average age was 8.0 years. Grade 2

'For Dan, who did not join the larger project until age 36 months, parental reports rather than an actual language sample at 30 months were the basis of his identification as language delayed according to the criterion used. All subsequent data are consistent with these reports, justifying his inclusion in this study. One other child was also delayed in preschool language acquisition, but has not been included here because he could not be tested by us at Grade 2 due to geographic relocation of the family. His preschool data, which were similar to those of the four cases described here, were presented in a previous report (Scarborough & Dobrich, 1985b) in which he was identified as Case 3. At Grade 2, his score on a nationally standardized reading achievement test administered at his school was indicative of a probable reading disability. All of the cases identified by applying Ingram's criterion to our sample would almost certainly have been detected just as well by any other criterion at this age. Whether all other 30-montholds in the general population who meet this criterion would also appear language-delayed over subsequent assessments is a question requiring an epidemiological study.

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

72 Journal of Speech and Hearing Research assessments were not conducted for 2 children from the control group, 1 of whom could not be located and 1 of whom had geographically relocated. All evaluations took place in the children's homes in the presence of one or both parents. Videotaped records of all procedures were made, against which the accuracy of in-person test scoring was checked when necessary. Measures Preschool language production. Natural language samples at ages 24 through 60 months of age were drawn from mother-child play sessions with age-appropriate materials supplied by us. All maternal and child utterances during play were transcribed along with contextual notes, and every transcript was fully reviewed by a second transcriber prior to analysis. In addition, 17% of the play sessions were independently transcribed by two observers, who were found to agree in their representations of 91% to 99% of the children's vocalizations (identifiable words and unintelligible sounds). Four measures of the child's language production at each age were then computed: Index of Productive Syntax; pronunciation accuracy; lexical diversity; and mean length of utterance. For each measure, from 14% to 22% of all transcripts were independently coded by two scorers, and from 90% to 98% of coding judgments were found to be in agreement. The Index of Productive Syntax (IPSyn), a measure of grammatical complexity, was derived by crediting the occurrence of up to 2 dissimilar tokens of 56 types2 of syntactic and morphological forms (encompassing noun phrase elaborations, verb phrase constructions, inflectional morphemes, interrogative and negative forms, and simple and complex sentence structures) in a 100-utterance child language corpus in which unintelligible, imitative, and selfrepetitive utterances were not included (Scarborough, 1985). IPSyn scores have been shown to be sensitive to syntactic deficits of ELD preschoolers whose utterance

2 The 56 IPSyn items, many of which were drawn from Miller's (1981) system for "Assigning Structural Stages," are: noun, pronoun, modifier, 2-word NP, article, predicate NP, plural inflection, preverb NP, 3-word NP, NP adverb, noninflectional bound morpheme; verb, particle/preposition, prepositional phrase, copula, catenative, present auxiliary, progressive inflection, adverb, present modal, third person singular present tense inflection, past modal, past tense inflection, past auxiliary, preverb adverb, past tense copula, elliptical/emphatic auxiliary; intonational question, routine question, uninverted Wh- question, inverted Wh- question, inverted yes/no question, Whyl WhichlWhen/lWhose question, tag question; negation morpheme, simple VP negation, auxiliarized negation; 2-word utterance, S-V, V-O, S-V-O, conjunction, 2-VP sentence, conjoined phrases, simple infinitive, Let's construction, adverbial conjunction, predicate complement, conjoined sentences, Wh- clause, bitransitive predicate, 3-VP sentence, relative clause, complex infinitive, gerund, fronted/embedded subordinate clause. Additional points may be earned for producing other complex forms. A newer version of the IPSyn is now available (Scarborough, in press).

33 70-83 March 1990 lengths are more nearly age-appropriate (Rescorla & Schwartz, 1988; Scarborough & Dobrich, 1985b). The pronunciationaccuracy score was based upon the child's production of 100 successive identifiable words from the language sample, excluding determiners, pronouns, auxiliary verbs, proper nouns, and yes/no responses. The number of omissions, additions, substitutions, and transpositions of consonant phonemes in those words was tallied for each child at each age. This measure has been shown to be sensitive to age changes in the phonological production skills of normal children, and the order of mastery of individual phonemes derived from these scores corresponds well with norms for tests of elicited speech production (Dobrich, 1988). The vocabulary diversity score was the number of unique lexical types (i.e., the number of different words) produced (Nelson, 1975), excluding inflectional morphemes, in the same 100-utterance corpus from which the IPSyn score was derived. Last, each child's mean length of utterance (MLU) in morphemes was computed at each age, based on the same 100 utterances as for IPSyn and vocabulary scores. Conventional guidelines for tallying morphemes were followed (Brown, 1973; Miller, 1981) except that all selfrepetitive and imitative utterances were excluded (Bloom, 1973) and some compounds were considered multimorphemic when there was evidence of their components being combined productively elsewhere in the transcript (Scarborough, Wyckoff, & Davidson, 1986). Preschool receptive language tests. The following assessments of receptive vocabulary, syntax, phoneme discrimination, and phonological segmentation were made during the preschool period: The Peabody Picture Vocabulary Test (PPVT; Dunn, 1965; Dunn & Dunn, 1981) of receptive vocabulary was given at 30 and 42 months of age. On each item of this test, the child must indicate which of four pictures corresponds to a single word spoken by the examiner. The PPVT is widely used to assess children's comprehension of the meanings of spoken words. Poor performance on the PPVT has been associated with disabled reading in somewhat older children (e.g., Stanovich, Cunningham, & Feeman, 1984). The 40-item receptive portion of Lee's (1971) Northwestern Syntax Screening Test (NSST) was administered at 36, 48, and 60 months of age. On each item of this sentence comprehension test, the child is asked to indicate which of four pictured scenes corresponds to a sentence spoken by the examiner. NSST scores are known to increase reliably over the preschool period (Lee, 1971). Sentence comprehension deficits are also related to reading abilities in schoolchildren (e.g., Byrne, 1981). Two phoneme discriminationseries (PDS-1 and PDS2) were prepared for this study. In both versions, on each item the child indicated which of two or three pictures on a page corresponded to a word spoken by the examiner. PDS-1, used at 30 and 36 months, required the child to

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

SCARBOROUGH & DOBRICH:

discriminate 24 vowel and initial consonant contrasts. 3 PDS-2, used at 48 months, assessed 36 more difficult consonant distinctions. 4 The PDS was developed and used as a quick means of identifying children with potential difficulties hearing language spoken to them. Children who earned low PDS scores were given complete audiological examinations at a university clinic. In every such case, although their phoneme discrimination difficulties were confirmed, no audiological evidence of hearing impairment was found. In the present study, therefore, the PDS served as a measure of language-specific discrimination skill. Speech discrimination skills of older children are known to be associated with reading abilities (e.g., Catts, 1986). At age 60 months only, one aspect of metalinguistic skill was examined. A phonological segmentation (PHSEG) measure was derived by scoring the number of correct responses to a subset of 20 items from the Sounds and Letters portion of the Stanford Early School Achievement Test (SESAT), Level 1 (Madden, Gardner, & Collins, 1981). Ten of these items require the child to match spoken words on the basis of rhyme, and the other 10 require word-matching on the basis of initial phonemes. This measure was an effective predictor of later reading abilities in the larger sample from which the subjects of the present study were drawn (Scarborough, 1989), and similar assessments of metalinguistic awareness have been shown to predict the future reading achievement of kindergarteners in many other studies (e.g., Stanovich, Cunningham, & Cramer, 1984; Tunmer, Herriman, & Nesdale, 1988). IQ assessments.The GeneralCognitive Index (GCI) from the McCarthy Scales of Children's Abilities (McCarthy, 1972) was computed for each child at 36,48, and 60 months, using the shortened scale developed by Kaufman (1977). The six subtests retained in the short version are: Verbal Fluency, Word Knowledge, Conceptual Grouping, Puzzle Solving, Numerical Memory, and Counting and Sorting. At the end of Grade 2, the Revised Wechsler IntelligenceScale for Children (WISC-R; Wechsler, 1974) was administered from which Full-Scale, Performance, and Verbal IQ scores were derived. Grade 2 reading and language measures. At the end of second grade, Reading Cluster scores were obtained from the Woodcock-Johnson Psychoeducational Battery (WJPB; Woodcock & Johnson, 1977). This reading achievement measure is a weighted composite of three subscores: word recognition, pseudoword decoding, and prose comprehension. In addition, the children's parents

3In the PDS-1, the 12 vowel contrasts were: ball/bell, bow/bee, can/cane, storelstar, pen/pin, and bowl/ball. The 12 initial consonant contrasts were: pearlbear, goat/coat, nail/mail, pan/can, and tie/pie. The examiner identified all 24 pictures for the child before asking the child to make any choices between paired stimuli. 4 PDS-2 items were: lakelrakelwake, sip/ship/sick, gum/gun/ run, rice/iceleyes, car/card/cart, leave/leafileash, washlwatchl wasp, bowl/blowlbow, ghost/goatslcoats, freeze/threes/trees, schoollstool/spool,frying/flying/lying.

Early Language Delay

73

provided results of school-administered nationally standardized achievement tests of reading and mathematics at Grade 2. Most schools used recent editions of the California Achievement Test, the Comprehensive Test of Basic Skills, the Iowa Test of Basic Skills, the Metropolitan Achievement Test, or the Stanford Achievement Test. The PPVT-R was readministered at the end of Grade 2, along with a test of speech production. In the "Sesquipedalians" task, the children were asked to imitate a series of 15 unfamiliar long words, or sesquipedalians, including: sesquipedalian, cranium, accelerator, polychromatic, circumscribe, dodecagon, thermotherapy, lilliputian, semicircular, peripatetic unanimity, acinaciform, funambulist, octagenarian, and arteriosclerosis. Snowling (1981) and Catts (1986) have suggested that the pronunciation of such phonologically complex stimuli may be especially difficult for reading disabled schoolchildren.

RESULTS The performance of the control group and of the individual ELD cases on all preschool measures will first be described. Group differences will then be discussed in relation to SES, IQ, and age. Last, reading and language abilities of all subjects at Grade 2 will be presented.

Preschool Language Production In Figures 2 through 5, the longitudinal development of the 4 ELD children is compared to that of the control group for each of the four measures of natural language production. The control group has been represented in these graphs by outlining the range of scores from one standard deviation below to one standard deviation above the group mean at each age. (Control means thus fall at the midpoints of the outlined ranges.) Individual curves for the language-impaired cases are identified by the initial letters of their fictitious names. Control group. A plateau in the course of normal development is quite evident for the language production measures. That is, there was a marked deceleration in the rate of progress after an initial rapid increase in scores. If one defines the plateau onset as the age at which the mean was significantly greater (p < .05, two-tailed t test) than the preceding mean, but not significantly lower than the succeeding mean, then the plateau for vocabulary diversity began at 36 months of age, and the plateaus for the other three measures of productive language began at 42 months. Examination of individual subjects' curves revealed that plateaus began as early as age 36 months for some normal children and as late as 48 months for others (Scarborough & Dobrich, 1985a). As more of the children reached the plateau level, there consequently occurred a decrease in differences among them, which is observable in the graphs as a reduction in variance from age 36 to 48 months. By age 5, some subjects appeared to reaccelerate

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

74 Journalof Speech and Hearing Research

33 n

. to

IU

U

r

70-83 March 1990

r

20 80

Co

-o 0

x 0

a)

0 40 C _

C, ci)

60

o C o

n,

-o -0

I0 1

0

60

r. 4-

40

0

C C 0

x a0

80

0 i,

C

C 0 C-

20

Pronunciation Accuracy

100

Grammatical

Complexity

M 0

I

I

I

I

24

30

36

42

I

48

54

60

24

age (months)

30

36

42

48

54

60

age (months)

FIGURE 2. Individual longitudinal curves of the four ELD cases for the Index of Productive Syntax in relation to that of the control group, for whom Il SD of the mean at each age has been outlined. L = Lois, R = Roger, D = Dan, M = Martin.

FIGURE 3. Individual longitudinal curves of the four ELD cases

and others not, producing the increases in means and variances at age 60 months on some measures. These phenomena, it is argued, cannot be attributed merely to a measurement ceiling or final developmental asymptote (except for the pronunciation accuracy measure, on which a score better than zero cannot be obtained). The utterances of adults and older children, like those of the most advanced 5-year-olds in our control group, are known to be longer and more complex than during the plateau period. Moreover, other investigators have also obtained evidence for a similar plateau in preschool MLU growth (Wells, 1985) and for increases in language proficiency at about 5 years of age (Bowerman, 1982; Karmiloff-Smith, 1979). ELD cases. Figures 2 through 5 also show that the grammatical complexity, pronunciation accuracy, utterance length, and vocabulary diversity of the 4 ELD children were severely deficient at the outset, but approached or reached normal levels by the end of the

deficits decreased with age from an average of -3.3 at age 30 months to -1.6 at 36 and 42 months, and the average deficit was negligible at both 48 months (-0.3) and 60 months (-0.2). Of the four cases, the severity of Dan's deficits diminished most slowly, and Lois' most rapidly. The selectivity of deficits also changed with age. Each child initially showed substantial delays in all four aspects of language production. Over time, each child's range of deficits narrowed, such that eventually their productive language abilities resembled those of the normal group. These developmental changes were somewhat different for the four cases. Lois recovered (to within 1 SD of the control mean) by age 36 months for MLU and vocabulary diversity and by age 42 months for IPSyn, but not until 60 months for consonant pronunciation. Similarly but more slowly, Roger showed recovery of MLU, vocabulary diversity, and IPSyn by age 48 months, and of consonant errors by 60 months. Martin, on the other hand, was age-appropriate in consonant production by 42 months, and in other production skills 6 months later. Dan's deficits in speech and vocabulary production were overcome by 42 months, but not until 60 months did his IPSyn score indicate recovery of syntactic complexity, and even then his MLU remained low. A more qualitative picture of their morpho-syntactic production is provided in Table 1, which shows the age at

preschool period.

The severity of their deficits can be expressed as the z-score difference between their scores and correspond-

ing control means. Across all ages, average deficits were greatest for IPSyn scores (z = -1.6), least for vocabulary

diversity (-0.8), and intermediate for consonant errors (-1.3) and for MLU (-1.2). Across the four measures,

for pronunciation accuracy in relation to that of the control group, for whom 1 SD of the mean at each age has been outlined. L = Lois, R = Roger, D = Dan, M = Martin.

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

SCARBOROUGH & DOBRICH:

Early Language Delay 75

150

/

L

5.0 130 0 o a,

0 C

110

0

L

4.0

4-,

0

90

0 4-

0,-

o

70

3.0

c)

/

R

E

ai Oi E

o

50

R-

2.0

L.

Utterance Length

Lexical Diversity

l' II

IfY' 24

30

36

42

48

54

60

1.0

age (months)

24

30

36

42

48

54

60

age (months)

4. Individual longitudinal curves of the four ELD cases for vocabulary diversity in relation to that of the control group, for whom ± 1 SD of the mean at each age has been outlined. L = Lois, R = Roger, D = Dan, M = Martin.

FIGURE 5. Individual longitudinal curves of the four ELD cases

which each child first produced at least one clear instance of important grammatical morphemes and syntactic constructions from the IPSyn. Grammatical forms that were produced by age 30 months by all control subjects were not used by the ELD cases until at least 6 months later. From age 36 months on, the ELD cases also began to use some more difficult forms simultaneously with the slowest of the normal children, but 6 months later than at least 9 of the 12 controls. Finally, for forms acquired by the control subjects by 42 months or later, many fewer delays and more age-appropriate acquisitions were shown by the ELD children.

developmental changes. As expected, PPVT scores increased significantly from age 30 to 42 months (t = 11.77, p < .001), PDS-1 scores increased between ages 30 and 36 months (t = 3.58, p < .01), and NSST scores increased from age 36 to 48 months (t = 11.81, p < .001) but not from 48 to 60 months (t = 1.67, p < .07). Second, correlations between scores on language tests at successive observations were carried out to determine whether individual differences in test performance remained stable over time. Strong evidence for such continuity was indeed found between ages 30 and 42 months for the PPVT (r = .87, p < .001), between ages 30, 36, and 48 months for the two versions of the PDS (r = .83 and .80, p < .01), and between 36, 48, and 60 months for the NSST (r = .72 and .63, p < .05). Phonological segmentation ability at age 60 months was not, however, related to phoneme discrimination skill at age 48 months, suggesting that these two phonological tests (only one of which is actually "receptive") reflect rather different aspects of language skill. Third, correlations between scores on different tests given at the same or adjacent ages were carried out to examine the degree of independence or overlap among lexical, syntactic, and phonological measures. PPVT and PDS scores were strongly related to one another (r = .79 to .88, p < .01), which may in part be attributable to the similar format used in both tests. NSST scores were also

FIGURE

Preschool Receptive Language Receptive vocabulary, sentence comprehension, phoneme discrimination, and phonological segmentation scores are provided for the control group and the four individual ELD cases in Table 2. Controlgroup. Performance of the control group on the language tests was analyzed in order to examine the adequacy of these measures with respect to age differentiation, developmental continuity, and covariation. First, for each test that was administered more than once over the preschool period, t tests were carried out between successive means to verify that each test was sensitive to

for mean length of utterance in relation to that of the control group, for whom ±+1 SD of the mean at each age has been outlined. L = Lois, R = Roger, D = Dan, M = Martin.

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

76 Journalof Speech and Hearing Research

33 70-83 March 1990

TABLE 1. Age at which grammatical forms were first produced by the ELD cases, in comparison to the ages by which the forms were produced by all 12 control subjects (solid vertical lines) and by 9 of the 12 control subjects (dashed lines). L = Lois, R = Roger, D = Dan, M = Martin. Age (months) Grammaticalform

24

subject + verb + object prepositional phrase plural -s no(t) + X 2-word predicate NP catenative progressive -ing copula present tense modal NP + negative + VP Wh- (+ NP) + VP present tense do/be aux 3rd person singular -s inverted Wh- question aux + neg + V inverted yes/no question Let's + predicate infinitive propositional complement conjoined clauses det + mod + N ore-verb NP past tense -ed past tense auxiliary 3-VP sentence gerund relative clause tag question

30

36

42

LRDM LRDM LRDM LRD LRDM LRDM LDM LRDM L ALRM LRD L L I LRDM

L

~

48

60

60+

M R D M RDM RDM

LRM LR DM L D LD M L_ _ _ RDM LRD M LR L D L RM L M D R M L L I

RM LR R L DM RDM

D M

II

similarly on tests of other receptive skills. Nevertheless,

to their language production skills was not very close, as revealed by correlations between receptive and expressive measures of the same linguistic domain that were given at the same ages. With respect to lexical development, PPVT and vocabulary diversity scores at 30 and 42 months were moderately correlated (r = .52 and .55, p < .05), indicating that children who produce many different words in a natural context are also able to recognize more spoken words on a picture recognition test. In contrast, NSST scores were related to IPSyn scores only at age 48

the relationship of the normal group's receptive abilities

months (r = .78, p < .01) but not at 36 or 60 months (r =

moderately correlated with both PPVT (r = .51 to .61, p < .05) and with PDS (r = .51 to .52, p < .05) scores of control subjects. In addition, NSST scores at age 60 months were strongly related to phonological segmentation skills at that age (r = .74, p < .01). In sum, for these normal children, language test scores

increased with age, the relative abilities of the 12 children remained similar over time, and the children who did well (or less well) on one test tended to perform

TABLE 2. Preschool language test scores for the language-delayed subjects in comparison to the control group. Number correct (raw score) and number of standard deviations from the control group mean (z score) are given. Controls test PPVT

NSST

ELD Cases Lois

Roger raw z

Dan raw z

age (mos)

Mean

SD

30

24.1

(11.9)

7

-1.4

17

-0.6

-

42 36 48 60

43.4 11.9 28.3 29.7

(10.6) (6.7) (3.6) (3.2)

26 4 25 28

-1.6 -1.2 -0.9 -0.5

28 6 18 27

-1.5 -0.9 -2.9 -0.8

27 14 23 29

-1.6 0.3 -1.5 -0.2

-

-

raw

z

PDS-1

30

19.0

(4.3)

18

-0.2

refused

PDS-2 PHSEG

36 48 60

21.3 31.7 13.4

(3.1) (2.5) (4.9)

18 28 10

-1.1 -1.5 -0.7

20 -0.4 33 0.5 refused

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

23 25 6

-

0.6 -2.7 -1.5

Martin raw z 29

0.4

40 4 17 24

-0.3 -1.2 -3.1 -1.8

18

-0.2

20 31 8

-0.4 -0.3 -1.1

SCARBOROUGH & DOBRICH: Early Language Delay

.33 and .08, respectively), and were never closely related to MLU (r = .01 to .20). Furthermore, consonant error rates were not correlated with any PDS scores (r = -. 03 to -. 25), but were related to phonological segmentation abilities at age 60 months (r = -. 62, p < .05). There was only weak evidence, therefore, that tests of sentence comprehension and phoneme discrimination tapped aspects of syntactic and phonological ability that were similar to those tapped by MLU, IPSyn, and consonant production measures. Instead, these tests and natural language production measures appear to assess different components of developing language. ELD cases. In most instances, the scores of the ELD children improved over age on the receptive language tests, but in comparison to the performance of the control group, many of their scores were delayed (i.e., were more than one standard deviation below the control group mean) or were refusals. As can be seen in the table, the receptive deficits of individual children were distributed rather unsystematically across lexical, syntactic, and phonological measures, except for Martin's more selective and consistent weakness on the NSST. Across different ages, the mean z-score of the ELD group on the receptive tests was -0.95, with somewhat larger deficits shown for the NSST (- 1.2) than for the PPVT (-0.9) and phonological tests (-0.7). Across measures, receptive problems were less severe on average at ages 30 to 42 months (mean z = -0.7), when productive deficits were greatest, than at older ages (mean z = -1.3) when productive deficits had largely been overcome. In light of the rather weak relationships observed between receptive and expressive skills within the control group, it is perhaps not surprising that receptive deficits of the ELD children did not reveal the same patterns of individual difference and developmental change as their productive deficits. It must be borne in mind, however, that the receptive measures were given less frequently, and at different ages, making comparisons difficult.

77

ceeds national norms for both the McCarthy Scales and the WISC-R, such that few scores below 100 are obtained. With respect to these local standards, therefore, many of the IQ scores earned by our ELD cases were lower than those of the majority of the children in the control group. 5 Thus the mean performance of the ELD group was lower than that of the control group for the GCI at age 60 months (t = 2.30, p < .05) and for Verbal IQ at Grade 2 (t = 2.27, p < .05). Differences in GCI scores at age 36 months (t = 0.89, p > .10) and 48 months (t = 2.11, p < .06) and in Full-Scale IQ (t = 1.90, p < .10), and Performance IQ (t = 1.72, p > .10) at Grade 2 also favored the control group but were not statistically significant. Relation of IQ to language skills. Within the control group, correlations were computed between IQ scores and language measures at the same or adjacent ages. GCI scores were consistently related to both PPVT (r = .66 to .68, p < .01) and PDS (r = .50 to .51, p < .05) scores. Correlations between IQ and NSST scores, however, decreased from .85 (p < .001) at age 36 months to .41 and .47 (p < .07) at ages 48 and 60 months, respectively. The correlation of IQ with phonological segmentation skill at age 60 months was .49 (p < .06). In contrast, the only significant relationship of IQ to language production involved consonant production errors at age 60 months (r = .78, p < .01). These results indicate that IQ (or some more general aspect of testtaking ability) is more strongly related to performance on receptive language tests than to natural language production. Group Differences For PPVT, NSST, and PDS-1 scores, and for the four language production measures, separate two-way ANOVAs were carried out with age (30 to 60 months) as the repeated within-subjects variable and with group (ELD vs. normal) as the between-subjects variable. Pairwise t

IQ Scores As shown in Table 3, the four ELD cases usually obtained normal IQ scores as preschoolers and at Grade 2. For our middle-class suburban New Jersey sample, however, we have found that average performance ex-

5 Among the ELD cases, Roger's IQ scores were particularly low. It should be noted, however, that Roger was the least cooperative examinee, and that his scores thus probably underestimate his true competence somewhat. At age 76 months, for example, he was retested on the McCarthy Scales; on that occasion he was unusually cooperative, and earned a GCI of 111.

TABLE 3. IQ scores for the language-delayed subjects and the control group. Test Preschool Aptitude McCarthy GCI-36 mos McCarthy GCI-48 mos McCarthy GCI-60 mos Grade 2 Aptitude WISC-R Verbal IQ WISC-R Performance IQ WISC-R Full-Scale IQ

Controls Mean SD

Lois

ELD Cases Roger Dan

Martin

107.7 116.3 118.8

(14.3) (13.7) (9.7)

99 99 99

90 87 76

95 107 121

118 109 113

119.1 118.4 120.2

(8.2) (8.2) (7.7)

107 115 112

97 101 92

117 112 117

111 124 119

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

78 Journalof Speech and Hearing Research

33

tests were then used to examine between-group differences on these measures and on scores for tests that were not administered more than once. No significant effects of age, group, or their interaction were found for PDS-1. In every other analysis, strong main effects of age were obtained [all p < .001; F (1,13) = 79.88 for PPVT, F (1,14) = 132.67 for NSST, and all F (4,56) > 29.29 for the four language production measures]. For the language production measures, there were also significant effects of group [all F (1,14) > 5.38, p < .05] and group x age interactions [all F (4,56) > 5.44, p < .001]. Neither result was obtained, however, for PPVT scores, and for NSST scores only the main effect of group was significant [F(1,14) = 5.49, p < .05]. Results of the t tests are shown in Figure 6 (except for PDS-1 on which there were no group differences). These analyses revealed that IPSyn scores, consonant error rates, and MLUs were more advanced for the control group than for the ELD group through age 42 months but not thereafter, whereas vocabulary diversity scores differed for the two groups only through age 36 months. For the PPVT, although there was no main effect of group in the ANOVA, the group difference at 42 months reached significance. NSST scores at age 48 months and phonological segmentation scores at age 60 months (not shown

_L^

30 mos

36 mos

42 mos

48 mos

60 mos

1u 80

60o 40 20

C o

80 0 40 °

00 0

V)

*

20 120

.) V

80

V _J

40 : 0-

*

5.0 4.0 -J 3.0 2.0 1.

0

I

52 24

FIGURE 6. Mean scores of the ELD group in comparison to the

full control group (hatched bars), the low-SES control group °

(135 stripes), and the low-IQ control group (45 stripes). Aster-

isks indicate control group means that differ significantly from those of the ELD group (p < .05, t tests).

783

March 1990

in the figure) were the only other receptive measures on which the groups differed. SES control group. Although the difference between groups in SES was not statistically significant, as noted earlier, we were nevertheless concerned about the possible influence of SES on the results. Therefore, a special low-SES control group was created by selecting from the entire control group the 2 boys and 2 girls with lowermiddle-class backgrounds (SES = 3.5, SD = 0.6). ANOVA results when this group was included were the same as when the full control group was included, except that there was no main effect of group for either vocabulary diversity or NSST scores. Results of t tests, shown in Figure 6, were also substantially the same, despite the loss of statistical power due to small sample sizes. Group differences in consonant errors and PPVT scores at age 42 months, however, were no longer statistically significant. IQ control group. Because some IQ differences between groups were found, as described earlier, and because IQ was correlated with language test scores within the control group, we were also concerned about the extent to which the results were influenced by IQ differences. Therefore, a special low-IQ control group was created by selecting from the entire control group the 3 boys and 1 girl with lowest mean IQ overall. (Only one of these children was also in the low-SES control group.) Mean GCI scores of this subgroup were 93.3 (SD 4.1) at age 36 months, 103.0 (SD 7.6) at 48 months, and 112.3 (SD 5.1) at 60 months; their mean WISC-R scores at Grade 2 were 112.7 (SD 5.1) for Verbal IQ, 114.0 (SD 9.7) for Performance IQ, and 115.0 (SD 7.0) for Full-Scale IQ. None of these means differed significantly from those of the ELD group. Therefore, to the extent that language development and verbal-cognitive growth go hand-inhand during the preschool years (or, that measures of language reflect similar developing processes to those tapped by IQ tests) one would expect differences between the ELD and IQ-matched control subjects to be greatly reduced or eliminated. When this low-IQ control group, rather than the full control group, was included in the ANOVAs, the results were substantially the same except that main effects of group were found only for IPSyn and consonant error scores. Results of t tests, shown in Figure 6, indicated that early differences were again obtained between the ELD and low-IQ groups on the language production measures, although only for IPSyn did the differences remain statistically significant through age 42 months. The major difference from prior analyses, however, was that no significant differences were obtained for any test scores at any age. IQ was thus strongly related not only to differences in receptive language within the control group (as revealed by correlational results described above) but also to differences between the ELD and control groups. Consequently, the effect of controlling for IQ by sampling was virtually to eliminate group differences on receptive language measures while having little effect on results for language production measures. In other words, children who were equivalent in both verbal and nonverbal IQ did indeed perform similarly on recep-

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

SCARBOROUGH & DOBRICH: Early Language Delay

79

reading disability (Rutter & Yule, 1975), these cases were all classified as reading disabled (Scarborough, 1989). The reading abilities of Lois, Roger, and Dan were comparable to those of other poor readers in the larger sample. Low reading achievement was seen for 20 other subjects, of whom 13 also met the stricter -2 SE criterion for reading disability. Their mean scores were 477.0 (SD 7.7) for the WJPB Reading Cluster and 55.2 (SD 20.4) for school-administered tests. Math achievement was also low for Roger but not for Lois and Dan. Among the normal readers, Martin and two children from the control group (one of whom was in the low-IQ subgroup) were low math achievers. Only for Lois was pronunciation of sesquipedalians impaired at Grade 2 (in relation to both the full and the low-IQ control groups). For all four ELD cases, moreover, receptive vocabulary scores at follow-up were at least one standard deviation below the mean of the full control group. The mean PPVT score of the low-IQ control group was 112.7 (SD 12.9), and only Dan's PPVT score (z = -2.0) was low in relation to that group.

tive language tests, but nevertheless differed substantially in language production abilities.

Grade 2 Language and Reading Achievement Table 4 provides results for academic achievement and PPVT scores at the end of Grade 2, three years after the last preschool assessments, on average. According to local norms established for the WJPB Reading Cluster scores in the larger sample from which these subjects were drawn (Scarborough, 1989), all of the control cases who were followed up became normal readers. Of the ELD cases, however, all but Martin became poor readers. Lois, Roger, and Dan's WJPB Reading Cluster scores were more than 1 year behind, and their scores on schooladministered tests were at least 10 points (about 1 SD) below those of the worst control subject. Consequently, their mean reading score was significantly lower than that of the low-IQ control group (476.3 vs. 499.3, p < .01) as well as that of the full control group (p < .01). Although Martin's reading achievement was normal at Grade 2, it should be noted that he had been identified as having difficulty when he first began formal instruction in reading. At that time, his classroom teacher expressed strong concern, and suggested that he be evaluated for possible retention in kindergarten. (Among our other cases, only Roger was similarly identified before Grade 2 as having academic problems.) Some reading disabilities do not emerge or are not identified until the later elementary grades (Satz, Taylor, Friel, & Fletcher, 1978), perhaps in response to changes in the cognitive and linguistic demands of the reading curriculum beyond the primary grades (e.g., Fletcher, Satz, & Scholes, 1981). When our sample is reevaluated in coming years, we can ascertain whether Jake's academic success will be maintained. Lois, Roger, and Dan were not only poor readers on an absolute basis, but also met a more stringent definition of reading disability as underachievement in reading. When expected reading levels were determinedby a regression of WJPB scores upon IQ in the larger sample from which all subjects in the present study were drawn, and a discrepancy of 2 standard errors between actual and expected reading level was used as the criterion for

DISCUSSION What happens to children with early language delay? For the four cases examined longitudinally in this study, broad and severe early deficiencies in syntax, phonology, and lexical semantics gradually gave way to more selective impairments, and finally to normal or near-normal language proficiency by age 5 years. Nevertheless, all but 1 child subsequently developed severe reading disabilities 3 years later. Bearing in mind that generalizing from so few cases must be undertaken with caution, these developmental patterns will be discussed and some of their implications for future research and clinical applications will be considered. Preschool Development of ELD Children The four ELD cases presented here were similar in several important respects from age 2 1/2 to 5 years. First, the overall severity of impairment was roughly equivalent for these children, particularly at early ages. Second,

4. Grade 2 academic achievement and language scores for the ELD children and the control group. TABLE

ELD cases Dan Roger

Controls Lois

Martin

Tests

Mean

SD

raw

z

raw

z

raw

z

raw

z

WJPB Reading Cluster School tests Reading Math PPVT Sesquipedalians

502.4

(6.6)

471

-4.8

477

-3.8

481

-3.2

499

-0.5

87.5 87.8 119.1 9.8

(9.3) (15.8) (9.1) (2.4)

46 93 102 7

46 -4.5 50 0.3 -1.9 · 108 11 -1.2

-4.5 -2.4 -1.2 0.5

51 98 87 10

-3.9 0.6 -3.5 0.1

87 66 108 11

-0.1 -1.4 -1.2 0.1

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

80 Journal of Speech and Hearing Research each child showed dramatic decreases in severity over the preschool period. Third, their productive language skills were very broadly impaired at the outset, despite the fact that these cases were selected solely on the basis of their lack of combinatorial language at age 30 months. Fourth, their weaknesses became increasingly more selective over time. In brief, over the preschool period these children underwent concurrent developmental changes in the severity and breadth of their productive language impairments. These observations are consistent with results from larger-scale studies by other investigators regarding the severity and scope of childhood language impairments. Aram and Nation (1975), for example, found that among 3to 7-year-olds, the younger cases more often exhibited broad deficits and the older children tended to show more selective deficits, an age difference that was seen within each of our longitudinal cases. Wolfus, Moscovitch, and Kinsbourne (1980) reported that 4- to 8-year-olds with the weakest semantic and syntactic abilities also tended to be the cases with the greatest breadth of impairment. This finding that greater severity is associated with greater generality (less selectivity) of language impairments was also shown by Bishop and Edmundson (1987). In that longitudinal study of ELD children from age 4 to 5 1/2 years, language deficit profiles were seen to change in the direction of reduced severity and greater selectivity, as was the case for our four ELD cases. Taken together, the available results suggest that severity and selectivity of language delays may often covary inversely. In particular, the longitudinal data have helped to clarify the developmental progression of this relationship. Accordingly, it now appears that differences among children in the selectivity of their impairments may not always be indicative of subtypes, as is sometimes presumed. For instance, while some children may indeed exhibit only selective articulation disorders throughout the preschool years, others who appear to have isolated speech problems, like our subject Lois at 42 and 48 months of age, may actually be children who were previously more globally delayed. More longitudinal studies of the development of children with different profiles of impairment will be required to explore these issues fully. As the selectivity of their deficits increased, differences among the children were observed with respect to the aspect(s) of language that remained weak. No child ever showed a purely lexical deficit. Instead, residual phonological and syntactic problems, in combination and in isolation, were seen in most cases. These observations are consistent with previous findings that selective deficits occur only rarely for semantic abilities, but are most likely to be observed for syntactic and especially phonological skills (Aram & Nation, 1975, 1980; Bishop & Edmundson, 1987). These investigators have interpreted this as suggesting a hierarchy of vulnerability of language components. Aram and Nation (1975) also speculated that age differences in selectivity might result from therapeutic intervention, but the present results for untreated cases argue otherwise. An alternative explanation is that plateaus may begin earlier in normal vocabulary growth

33 70-83 March 1990 than in the development of formal (phonological and syntactic) systems. If so, "illusory recovery" could occur at different ages for different language measures, producing apparent changes in the selectivity of deficits. More research on these phenomena is needed before a full account of the observed progression can be provided. In contrast, the ELD children's performance on tests of receptive language was more erratic, and did not improve with age. Moreover, their receptive abilities generally resembled those of other children of low-normal IQ. Only Martin showed a clear pattern of consistent isolated deficits in sentence comprehension despite his relatively high IQ scores and normal receptive vocabulary and speech discrimination abilities. Implications for assessment and research sampling. The longitudinal data of the present study revealed progressive reductions in each child's apparent degree of impairment relative to normal variation on language production measures. This phenomenon poses problems for diagnosing such cases clinically and for selecting wellmatched cross-sectional research samples, when evaluations must be made on the basis of language assessments taken at only one age per child. Bishop and Edmundson (1987) have also pointed to the need for conventional diagnostic and sampling procedures to adopt more agedependent criteria. Paradoxically, using a uniform criterion (such as N standard deviations below normal) will evidently result in the identification of more severely delayed children at older ages. Many of Bishop and Edmundson's subjects would not have been included in their sample, they judged, if assessments at age 5 1/2 years rather than at age 4 years had been used. Likewise, our subjects as 3year-olds would almost certainly have been identified as language-delayed, but as 5-year-olds would have been seen as marginally behind, if at all. Accordingly, older preschoolers who do meet stiff criteria would be more severely delayed than the cases we have described, and thus would not necessarily be comparable in severity to 3-year-olds selected according to the same criteria. The changes in selectivity over time also make it difficult to equate diagnoses for children of different ages. When research samples include language-impaired subjects from a wide range of ages, as has usually been the case in prognostic studies, age differences may produce misleading results with respect to the heterogeneity of disorders or the relationship of prognosis to current functioning. Although it is undeniably more difficult to locate subjects within a narrow age band, by doing so "one may find far more regularity and order in data from languageimpaired samples than might have seemed possible" (Bishop & Edmundson, 1987, p. 52). Although the present sample was very small, the results underscore that message. Finally, with regard to the interpretation of language test results, the question must be raised as to whether severity of language impairments should be judged in relation to chronological age alone, or in relation to mental age also. In the present study, many of the ELD children's test scores were low in comparison to those of

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

SCARBOROUGH & DOBRICH:

the full control group but were not low in comparison to a group with IQs more similar to their own. Just as reading disabilities are usually defined as underachievement relative to IQ (e.g., Rutter & Yule, 1975), so perhaps should IQ be taken into consideration when diagnoses of ELD are made on the basis of test results. This is less of an issue with respect to natural language assessments, which appear to be less closely related to IQ. Recovery From and Persistence of Early Language Impairments Our longitudinal findings, along with those of Bishop and Edmundson (1987), are consistent with earlier evidence that apparent recovery from ELD may be seen quite often in the later preschool years, particularly among children with less severe delays (MacKeith & Rutter, 1972; Morley, 1972; Silva, 1980). By age 60 months, our four cases exhibited few remaining productive or receptive language problems, if any. Likewise, recovery to normal or nearly normal levels of language skill was seen by age 5 1/2 years for 44% of Bishop and Edmundson's subjects who had been delayed as 4-yearolds. Nevertheless, when our four cases were followed up at the end of Grade 2, the results also proved to be consistent with prior findings that negative outcomes are quite frequent during the school years and beyond (Aram et al., 1984; Fundudis et al., 1979; Hall & Tomblin, 1978; Padgett, 1988; Stark et al., 1984). Only one of our cases became a normal reader and all four exhibited poor receptive vocabulary skills. Thus two developmental patterns shown in separate studies in the research literature on ELD-namely, shortterm recovery and long-term persistence-were seen to occur successively within individual children in the present study. This cannot be easily accounted for by the hypothesis that differences among cases in severity or subtype are predictive of differences in outcome. All four cases showed mild and selective delays toward the end of the preschool period, and thus should have been less likely to exhibit subsequent problems than would children with more severe and generalized language deficits. The "illusory recovery" hypothesis (Figure 1) provides a better account of the findings. According to that model, normal language development proceeds in a stepwise manner, with spurts in the third and sixth years of life separated by an extended plateau, as seen for the present control group. For ELD children, recovery during the preschool years may thus often be illusory, resulting from the temporary convergence of stepwise growth functions of normal and delayed children. In actuality, we would argue, they may be as seriously delayed as ever in terms of the timing of their stepwise language development. It should be emphasized, furthermore, that the occurrence of plateaus in longitudinal age curves does not necessarily imply that language ceases to develop at those times, but only that our measurements may not adequately tap changes in language proficiency that may

Early Language Delay

81

be occurring. In this regard, however, the hypothesis that measures of "higher level" metalinguistic skills would be especially sensitive to residual deficits of ELD children was not supported by the results for the phonological segmentation test at age 60 months. Moreover, quantitative spurts and plateaus in development may not occur for all aspects of language performance. In particular, the "illusory recovery" model does not account as well for receptive abilities, the development of which seems instead to be more closely tied to general cognitive growth than to productive language level. While the consistent relations we observed between IQ and language test scores suggest that there is considerable overlap among such measures, it is not clear whether this stems from the similar verbal-cognitive content of the tests or from their similar attentional-behavioral demands. Clearly, there is a need for more study of latepreschool language development in normal children, and for the development of measurement tools that might be more independent of IQ and more sensitive to residual deficits in language-delayed cases like ours. The Relationship of ELD to Reading Disabilities Although there is considerable evidence that language delays are often antecedent to reading disabilities, the etiological nature of this relationship is not entirely clear. It has generally been presumed that children with a weaker mastery of language are at a disadvantage for the task of learning to read (e.g., Catts & Kamhi, 1986), and this assumption receives support from many recent demonstrations that individual differences in verbal skills among kindergarteners are moderately predictive of differences in reading achievement several years later (e.g., Horn & Packard, 1985; Scarborough, 1989; Share, Jorm, Maclean & Matthews, 1984; Stanovich, Cunningham, & Cramer, 1984). With respect to prognoses for individual children, however, this explanation may not be completely satisfactory. There is no evidence, for instance, that Martin's language skills were less weak than those of the other cases upon school entry, so his disadvantage in early reading acquisition would presumably have been equivalent to theirs. His greater success in reading achievement is thus not easily accounted for. One factor of potential relevance to this issue is the family background of a language-delayed child. As mentioned earlier, the four cases described here were drawn from a larger sample in which the relationships between early language skills and later academic achievement are being investigated. In that sample, by design, about half the children were selected because they had at least one parent or older sibling who was reading disabled (RD). Naturally, subjects from such families were excluded from consideration when the present control group was selected. When we reviewed 30-month language samples to identify ELD cases for the present study, however, all children in the larger sample were considered. Except for Martin, our ELD cases turned out to have family incidence of RD. (This was the case not only for Lois, Roger,

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

33 70-83 March 1990

82 Journal of Speech and Hearing Research

and Dan, but also for a fifth ELD preschooler who could not be followed up directly at Grade 2. See footnote 1.) As reported elsewhere (Scarborough, 1989), Grade 2 reading abilities xere strongly associated with familial incidence of reading problems (but not with socioeconomic background) in the larger study. Of the 66 children who were followed through second grade, 23 became poor readers (including Lois, Roger, and Dan), of whom all but two of had disabled readers in their immediate families. Likewise, only 32% of the children with a family incidence of RD became normal readers. Thus on the basis of family background alone, Roger, Lois, and Dan were at high risk for reading disabilities regardless of their early language abilities, and Martin's risk was slight. What, then, is the etiological relationship of ELD to reading disabilities? Is it similar for children with familial backgrounds of RD and for children from families of normal readers? Because familial risk has rarely been considered in prognostic studies of ELD, the answers to such questions remain elusive. In addressing them, it would first of all be helpful to know more about the reading abilities of the parents and older siblings of preschoolers seen clinically for language delays; at present, it is not known what proportion of the clinical population--or the general populationresembles Martin versus Roger, Lois, and Dan in this regard. We have found that assessing familial RD through self-report methods (interviews and questionnaires) provides a less accurate indication of risk than does direct testing of family members (Scarborough, 1989). In our ongoing research, we are also comparing the language development of children from each kind of family background, with respect to early individual differences and subsequent long-term outcomes. We already know, of course, that among the children from RD families who themselves became disabled readers, only three (14%) met our criterion for language delay at age 30 months. We next will investigate whether the other children who became RD showed language impairments at later ages or were consistently at the bottom end of the normal distribution. If so, this would be consistent with the hypothesis that language weaknesses directly contribute to reading difficulties. Preliminary results (Scarborough, 1988) indicate that this may indeed be the case. It will also be important to ask what other characteristics distinguish them from the children (from each type of family background) with relatively weak early language who became more successful readers. In addition, the nature of verbal communication and literacy habits within the families will be examined. Although there is growing evidence that reading disabilities may have a genetic basis (e.g., DeFries, Fulker, & LaBuda, 1987), it may nevertheless be the case that a young child's communicative environment and exposure to books and reading may affect his or her early language skills and future reading achievement. We hope that clearer answers to some of these questions about the relationships among familial reading disability, preschool language development, and reading achievement will soon be available. In closing, we must point out that the limitations of this study are very clear, and that the results perhaps pose as many questions as'they answer. We have described only

four cases, and it is simply not known how representative they are of ELD children generally, in light of their family backgrounds of reading disability. In hindsight, furthermore, we wish that we had assessed receptive abilities more thoroughly and more often during the preschool period, and that we had obtained more information about language skills at Grade 2. Nevertheless, these four case histories have shown quite clearly that ELD children can undergo changes in the severity and scope of their deficits over time, and that dramatic improvements in preschool abilities do not necessarily support a positive prognosis for future reading achievement and language skills. We hope that future research will take into account some of the questions we have raised, in order to further our understanding of differences among language delayed children and of the etiological relationship between early language impairments and reading disabilities. ACKNOWLEDGMENTS The research was supported in part by grants HD18409, HD18571, and HD22351 from the National Institute of Child Health and Human Development. We are grateful to Maria Hager, Karin Lifter, Pam Johnson, and Janet Wyckoff for their contributions to the collection, analysis, and interpretation of the data. We are also deeply indebted to the subjects and their families for their participation in the project. Some of the findings were reported previously to the Wisconsin Symposium on Research in Child Language Disorders.

REFERENCES ASHA AD Hoc COMMITTEE ON LANGUAGE/LEARNING DISABILITIES. (1982). Position statement on language learning disorders. Asha, 24, 937-944. ARAM, D. M., EKELMAN, B. L., & NATION, J. E. (1984). Pre-

schoolers with language disorders: 10 years later. Journal of Speech and Hearing Research, 27, 232-244. ARAM, D. M., & NATION, J. E. (1975). Patterns of language

behavior in children with developmental language disorders. Journal of Speech and Hearing Research, 18, 229-241. ARAM, D. M., & NATION, J. E. (1980). Preschool language disorders and subsequent language and academic difficulties. Journal of Communication Disorders,13, 159-170. BASHIR, A. S., WIIG, E. H., & ABRAMS, J. C. (1987). Language

disorders in childhood and adolescence: Implications for learning and socialization. PediatricAnnals, 16, 145-156. BISHOP, D. V. M., & EDMUNDSON, A. (1987). Language-impaired

4-year-olds: Distinguishing transient from persistent impairment. Journalof Speech and HearingDisorders, 52, 156-173. BLOOM, L. (1973). One word at a time: The use of single word

utterances before syntax. The Hague, The Netherlands: Mouton.

BOWERMAN, M. (1982). Reorganizational processes in lexical and semantic development. In E. Wanner & L. R. Gleitman (Eds.), Language acquisition: The state of the art (pp. 319-346). Cambridge, MA: Cambridge University Press.

BROWN, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press. BYRNE, B. (1981). Deficient syntactic control in poor readers: Is a weak phonetic memory code responsible? Applied Psycholinguistics,2, 201-212. CATTs, H. W. (1986). Speech production/phonological deficits in reading disordered children. Journalof LearningDisabilities, 19, 504-508.

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

SCARBOROUGH & DOBRICH: Early Language Delay H. W., & KAMHI, A. G. (1986). The linguistic basis of reading disorders: Implications for the speech-language pathologist. Language, Speech, and Hearing Services in Schools, 17, 329-341.

CATTS,

DEFRIES, J. C., FULKER, D. W., & LABUDA, M. C. (1987). Evi-

dence for a genetic etiology in reading disability in twins. Nature, 329, 537-539. DOBRICH, W. (1988). Phonologicaldevelopment in preschoolers: Learning to speak English by speaking English. Unpublished doctoral dissertation, Rutgers University. DUNN, L. M. (1965). PPVT. Circle Pines MN: American Guidance Service. DUNN, L. M., &DUNN, L. M. (1981). PPVT-Revised. Circle Pines MN: American Guidance Service. FLETCHER, J. M., SATZ, P., & SCHOLES, R. J. (1981). Develop-

mental changes in the linguistic correlates of reading achievement. Brain and Language, 13, 78-90. FUNDUDIS, T., KOLVIN, I., & GARSIDE, R. (1979). Speech re-

tarded and deaf children: Their psychological development. London: Academic Press. HALL, P. K., & TOMBLIN, J. B. (1978). A follow-up study of children with articulation and language disorders. Journal of Speech and HearingDisorders, 43, 227-241. HOLLINGSHEAD, A. B., & REDLICH, F. C. (1958). Social class and

mental illness. New York: Wiley. HORN, W. F., & PACKARD, T. (1985). Early identification of learning problems: A meta-analysis. Journal of Educational Psychology, 77, 597-607. INGRAM, T. T. S. (1972). The classification of speech and language disorders in young children. In M. Rutter & J. A. M. Martin (Eds.), The child with delayed speech. Clinics in developmental medicine (No. 43). London: Wm. Heineman Medical Book, Ltd. KARMILOFF-SMITH, A. (1979). Micro- and macrodevelopmental changes in language acquisition and other representational systems. Cognitive Science, 3, 91-118. KAUFMAN, A. S. (1977). A McCarthy short form for rapid screening of preschool, kindergarten, and first grade children. Contemporary EducationalPsychology, 2, 149-157. KING, R. R., JONES, C., & LASKY, E. (1982). In retrospect: A fifteen-year follow-up report of speech-language-disordered children. Language, Speech, and HearingServices in Schools, 13, 24-32. LEE, L. L. (1971). Northwestern Syntax Screening Test (NSST). Evanston, IL: Northwestern University Press. LEVI, G., CAPOZZI, F., FABRIZI, A., & SECHI, E. (1982). Language disorders and prognosis for reading disabilities in developmental age. Perceptualand Motor Skills, 54, 1119-1122. MACKEITH, R. C., & RUTTER, M. (1972). A note on the prevalence of language disorders in young children. In M. Rutter & J. A. M. Martin (Eds.), The child with delayed speech. Clinics in developmental medicine (No. 43, pp. 48-51). London: Heinemann Medical Book, Ltd. MADDEN, R., GARDNER, E. F., &COLLINS, C. S. (1981). Stanford

Early School Achievement Test. New York: Psychological Corporation. MCCARTHY, D. (1972). McCarthy Scales of Children'sAbilities. New York: Psychological Corporation. MILLER, J. F. (1981). Assessing languageproduction in children. Baltimore, MD: University Park Press. MORLEY, M. E. (1972). The development and disorders of speech in childhood. Edinburgh & London: Churchill Livingstone. NELSON, K. (1975). The nominal shift in semantic-syntactic development. Cognitive Psychology, 7, 461-479. PADGETT, S. Y.(1988). Speech- and language-impaired three and four year olds: A five year follow-up study. In R. L. Masland & M. W. Masland (Eds.), Preschool prevention of readingfailure (pp. 52-77). Parkton, MD: York Press.

83

retardation. Journal of Child Psychology and Psychiatry, 16, 181-197. SATZ, P., TAYLOR, H., FRIEL, J., & FLETCHER, J. (1978). Some

developmental and predictive precursors of reading disabilities: A six year follow-up. In A. L. Benton & D. Pearl (Eds.), Dyslexia: An appraisal of current knowledge (pp. 313-347). New York: Oxford University Press. SCARBOROUGH, H. S. (1985, April). Measuring syntactic devel-

opment: The Index of Productive Syntax. Paper presented to the Society for Research in Child Development, Toronto. SCARBOROUGH, H. S. (1988, December). Early language devel-

opment of children who became dyslexic. Paper presented to the Child Language Group, New York. SCARBOROUGH, H. S. (1989). Prediction of reading disability

from familial and individual differences. Journal of Educational Psychology, 81, 101-108. SCARBOROUGH,

H. S. (in press). Index of productive syntax.

Applied Psycholinguistics. SCARBOROUGH, H., & DOBRICH, W. (1985a, December) Does

stage V last until age five? Evidence from normal and language-delayed children. Paper presented to the New York Child Language Group, New York. SCARBOROUGH, H., & DOBRICH, W. (1985b) Illusory recovery

from language delay. Proceedings of the Symposium on Research in Child Language Disorders, 6, 90-99. SCARBOROUGH, H., WYCKOFF, J., & DAVIDSON, R. (1986). A

reconsideration of the relation between age and mean utterance length. Journal of Speech and Hearing Research, 29, 394-399. SHARE, D. L., JORM, A. F., MACLEAN, R., & MATTHEWS, R.

(1984). Sources of individual differences in reading acquisition. Journal of EducationalPsychology, 76, 1309-1324. SILVA, P. A. (1980). The prevalence, stability and significance of language delay in preschool children. Developmental Medicine and Child Neurology, 22, 768-777. SNOWLING, M. J. (1981). Phonemic deficits in developmental dyslexia. PsychologicalResearch, 43, 219-234. STANOVICH, K. E., CUNNINGHAM, A. E., &CRAMER, B. B. (1984).

Assessing phonological awareness in kindergarten children: Issues of task comparability. Journal of Experimental Child Psychology, 38, 175-190. STANOVICH, K. E., CUNNINGHAM, A. E., &FEEMAN, D. J. (1984).

Intelligence, cognitive skills, and early reading progress. Reading Research Quarterly, 19, 278-303. STARK, R. E., BERNSTEIN, L. E., CONDINO, R., BENDER, M., TALLAL, P., & CATTS, H. (1984). Four-year follow-up study of

language impaired children. Annals of Dyslexia, 34, 49-68. STROMINGER, A. Z., & BASHIR, A. S. (1977, November). A nine-

yearfollow-up of language-delayedchildren. Paper presented to the American Speech and Hearing Association, Chicago. TUNMER, W. E., HERRIMAN, M. L., & NESDALE, A. R. (1988).

Metalinguistic abilities and beginning reading. Reading Research Quarterly, 23, 134-158. WALLACH, G. P., & BUTLER, K. G. (1984). Language learning

disabilities in school-age children. Baltimore: Williams & Wilkins. WECHSLER, D. (1974). Wechsler Intelligence Scale for Children-Revised. New York: Psychological Corporation. WELLS, G. (1985). Language development in the pre-school years. Cambridge: Cambridge University Press. WOLFUS, B., MOSCOVITCH, M., & KINSBOURNE, M. (1980) Sub-

groups of developmental language impairment. Brain and Language, 10, 152-171. WOODCOCK, R. W., &JOHNSON, M. B. (1977). Woodcock-Johnson

PsychoeducationalBattery. Boston: Teaching Resources. Received November 7, 1988 Accepted July 21, 1989

RESCORLA, L., & SCHWARTZ, E. (1988, August). Outcome of

specific expressive language delay. Paper presented to the International Conference on Infant Studies. Washington DC. RUTTER, M., & YULE, W. (1975). The concept of specific reading

Requests for reprints should be sent to Hollis Scarborough, Psychology Department, Brooklyn College of CUNY, Brooklyn, NY 11210.

Downloaded From: http://jslhr.pubs.asha.org/ by a Western Michigan University User on 03/28/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

Development of children with early language delay.

Four children with early language delays (ELD) were compared to a control group of 12 children with respect to their preschool language abilities from...
2MB Sizes 0 Downloads 0 Views