DEVELOPMENTAL DISABILITIES RESEARCH REVIEWS 17:141-150(2011)

LATE TALKERS: DO GOOD PREDICTORS OY OUTCOME EXIST?

Leslie Rescorla* Department of Psychology, Bryn Mawr College, Bryn Mawr, Philadelphia

Both small-scale and epidemiological longitudinal studies of early Although distinct diagnostic groups within the population language delay indicate that most late talkers attain language scores in the of young children manifesting expressive language delay are easily average range by age 5, 5, or 7. However, late talker groups typically obtain delineated, their relative percentages are less well documented. significantly lower scores than groups with typical language histories on most language measures into adolescence. These findings support a One study that did address this issue was conducted by Busch- dimensional account of language delay, whereby late talkers and typically mann et al. [2008], who studied 100 German 2-year-olds with developing peers differ quantitatively on a hypothetical language ability delayed expressive language identified in pediatric practices. Of spectrum. Variation in language ability is presumed to derive from variation these 100 children (65% boys), 78 were late talkers, 18 had lan- in skills subserving language, such as auditory perception/processing, word guage delay associated with cognitive impairment, and four had retrieval, verbal working memory, motor planning, phonological discrimi- nation, and grammatical rule learning. Expressive language screening at autism. These three groups did not differ in gender, birth posi- 18-35 months can serve an important public health function by identifying tion, or maternal education. Family history of language children whose expressive delay is secondary to autism spectrum disorder, impainnent was reported for 40% of the language-delayed group intellectual disability, hearing impairment, receptive language delay, or de- but for only 4% of a typically developing group. Of the 78 late mographic risk. Finally, the review suggests that demographic risk associ- ated with low SES may become more important as a causal factor in talkers, 61 had an expressive delay only, whereas 17 had recep- language delay as children get older. © 2013 Wiley Periodicals, inc. tive/expressive delay. The late talkers with receptive/expressive Dev Disabil Res Rev 2011; 17:141-150. delay had lower nonverbal IQs than the typically developing chil- dren, whereas the late talkers with expressive delays only did not. For children whose language delay is secondary to ASD or Key words: late talkers, outcomes, predictors ID, developmental progress tends to be slow and language may never reach nonnal levels. Outcomes also vary among late talkers. The purpose of this article is to review the literature on outcomes delay in expressive language is one of the most com- of late talkers with the aim of determining what, if any, variables mon reasons that young children are referred for are reliable predictors of outcome. We wul review two kinds of Aevaluation [Whitehurst and Fischel, 1994; Rescorla studies: (a) small-scale longitudinal studies of late talker and com- and Lee, 2000]. As noted by the US Preventive Services Task parison samples and (b) large-scale epidemiological studies of Force [2006], "Speech and language delay affects 5% to 8% of young children. The review encompasses children identified with preschool children, often persists into the school years, and language delays between 18 and 35 months of age. In the small- may be associated with lowered school performance and psy- scale studies, late talkers are typically a weU-defined group that chosocial problems" (p. 497). excludes children with hearing impairment, neurological disor- Because expressive language delay, like a fever, is a ders, ASD, and ID, and, in some studies, even children with symptom found in many conditions, children who are slow to receptive language delays. In the large-scale epidemiological stud- talk are a heterogeneous group. Expressive language delay is ies, children whose language delay is secondary to a more pdmary often secondary to another disorder [Whitehurst and Fischel, condition may not be excluded or even differentiated. 1994; Rescorla and Lee, 2000]. Late talkers may have a vocal tract malformation or a hearing loss that interferes with lan- OUTCOMES FROM SMALL-SCALE guage development. Other young children with delayed LONGITUDINAL STUDIES OF LATE TALKERS language may have a neurological disorder or general intellec- tual disability (ID). StiU other children who are slow to talk Preschool Outcome Studies of Late Talkers have autism spectrum disorder (ASD), in which case delayed In one of the first studies of late talkers, Fischel et al. [1989] language is accompanied by deficits in social relatedness, play, reported on 22 children identified with specific expressive and behavior. Children may also be slow to talk because they experience severe deprivation, neglect, or abuse. We hence- forth use the term late talkers to refer to children 18- to 35- *Correspondence to: Leslie Rescorla, Bryn Mawr College, 101 N. Merion Avenue, month-old who are slow to talk in the absence of any of these Bryn Mawr, PA 19010. E-mail: [email protected] other conditions presumed to be primary. Some late talkers Received 11 SepteEiiber 2012; Accepted 5 October 2012 have an expressive delay only, whereas others are delayed in View this article online in Wiley Online Library (wileyonlinelibrary.com) receptive language as well. DOI: 10.1002/ddrr.1108

© 2013 Wiley Periodicals, inc. language delay at 24-38 months. The late measures but not on receptive language children who scored >20th percendle. talkers scored more than two standard measures. A subsample of 34 late talkers Mean vocabulary scores were vastly dif- deviadons (SDs) below average on the (20 DR and 14 NDR) was idendfied ferent between the two groups at 18 Expressive One-Word Picture Vocabu- based on scores 1 SD below the mean on months (20 vs. 121 words, d = 1.9), lary Test [EOPVT; Gardner, 1981] and age 2 expressive language. By 42 months, and sdU very different by 30 months had mean vocabulades of <20 words. only the DR late talkers were sdU (419 vs. 569 words, d = 1.0). By 30 Five months later, 35% of the group had delayed in expressive and recepdve lan- months, 14 of the 36 late taUcers still scores in the average range on the guage. For the full sample of 200 scored <20th percentile on the CDI EOPVT (>85), with outcome predicted children, parental educadon and history (39% = delayed), whereas 22 scored by vocabulary size and proportion of con- of dyslexia, symbolic play and vocabulary above this cutoff (61% = late bloomers). sonants to vowels at intake. By age ^'^/z, comprehension at 14 months, expressive Of the 18 children delayed at 30 88% scored above 85 on the EOPVT; language at 24 months, and receptive months, 14 had been late talkers at 18 95% were in this range by 5'/2 [White- language at 30 months were all signifi- months and four had been typicaUy hurst etaL, 1992]. cant predicton of expressive language developing (9% of the typical group Thai et al. [1991] reported that outcome at age y/2 years, explaining were delayed at 30 months). These out- four of 10 late talkers idendfied at age 2 48% of the total vadance. comes yielded sensidvity of 78%, who were still delayed 1 year later (i.e., Rescoda et al. [2000] reported specificity of 66%, positive predictive the "truly delayed") had also been age 3 and age 4 outcomes for 34 late value of 39%, and negadve predictive delayed at age 2 in recepdve language talkers with normal receptive language value of 88%. and recognitory gestures. In contrast, the and nonverbal ability idendfied at 24— Femald and Marchman [2012] six "late bloomers" had been comparable 31 months. At age 3, 41% of the late reported that 18-month-old late talken to typically developing compadson chil- talkers scored above the 10th percentile had lower accuracy and speed scores than dren in receptive language and had made on MLU and 34% scored above the typicaUy developing children on looking- extensive use of gestures to compensate 10th percentile on the Index of Pro- while-listening (LWL), an online prefer- for communication. For a different ducdve Syntax [IPSyn; Scarborough, endal looking lexical processing task in sample. Thai et al. [2005] reported 1990]. At age 4, the percentages were the child listens to sentences such as age 4 outcomes of 20 children who 71 ahd 29%. For the same 34 late talk- "where is the doggy?" while a picture of scored <10th percentile in expressive ers, Rescoda et al. [1997] reported that the named referent and a foil are flashed vocabulary on the Communicative the percentage scodng in the normal on a screen. Faster reacdon dme and Development Inventory [CDI; Fenson range at age 3 (>16th percentile) was greater accuracy predicted steeper accel- et al., 1993] at 16 months. Although the 79% on the EOPVT, 58% on the Rey- eradon in vocabulary growth in the late late talker group scored in the normal nell Expressive Language Scale talker group from 18 to 30 months. range on language and cognitive tests, [ReyneU, 1977], 35% on MLU, and Adding reacdon dme as a predictor they scored significandy lower than a 24% on the IPSyn. Thus, granimadcal increased posidve predicdve value to group of 44 compadson children with skills showed the most persistent delays. 55% versus 39% using CDI score alone. typical language (TL) histodes. The late Rescoda et al. [2000] used regres- Vocabulary plus accuracy yielded a posi- talkers also scored lower than the com- sion analyses to identify significant dve predicdve value of 52%. Femald and padson children on a nonword repetition predictors of outcome. Expressive z- Marchman [2012] also noted that March- task, particularly for longer sdmuH. score accounted for 21-34% of the var- man and Femald [2008] had reported Feldman et al. [2005] reported iance in the age 3 outcome measures, faster processing speed and larger vocabu- sensitivity of only 50% and positive pre- whereas ReyneU Recepdve z-scores lary size at 18 months as predicdve of dicdve value of only 64% when [Reynell, 1977; ReyneU et al., 1990] better working memory at age 8. language delay at age 2 (CDI vocabu- and nonverbal abiUty scores were not Although Femald and Marchman [2012] lary scores < 10th percentile) was used significant predictors. The 34 late talkers acknowledged that their study could not to predict language delay at age 3 for (age range, lA—'iX months) had very address the degree to which endogenous 113 children (over 50% firom low- similar raw scores on the ReyneU Ex- versus expedendal factors accounted for income families). Thus, many children pressive Language Scale, but because lexical processing abuity, they cited a with delayed early vocabulary caught ReyneU ^-scores are age-based, the older previous study [^Hurtado et al., 2008] up by age 3 and many children with children had lower ar-scores than the demonstradng that dchness of matemal apparently normal development at age 2 younger children [Rescoda et al., 1997]. speech to the child at 18 months pre- were delayed by age 3. Vocabulary size Therefore, intake age was correlated at dicted both vocabulary size and lexical at age 2 was a better predictor of out- —0.1(3 with ReyneU Expressive ar-score, processing skill at 24 months. come than utterance length or sentence and both measures significantly predicted complexity score. outcomes. Because expressive language skills in typically developing children Summary Lyytinen et al. [2001] traced lan- improve so rapidly between 24 and 36 Preschool outcomes from small- guage skills using the CDI at 14, 24, 30, months, the older a late talker is within scale longitudinal studies of late talkers and 42 months in 200 Finnish children, this time pedod, the more he is faUing indicate that: (a) most children scored 106 from dyslexic famihes (DR), and 94 behind on a steeply accelerating curve, in the nomial range on language tests from families with no dyslexia history as noted by Rescoda et al. [1997]. by age 4 or 5; (b) grammatical delays (NDR). By age 2, the DR group had tended to be more protracted than vo- shorter maximum sentence length but FinaUy, Femald and Marchman cabulary delays; and (c) significant not smaller vocabulades than the NDR [2012] traced lexical development fiom predictors of outcome, which vaded group. By 42 months, the DR group had 18 to 30 months in 36 late talkers who across studies and left much vadance significantly lower scores on expressive scored <20th percentile on the CDI at unexplained, included recepdve lan- vocabulary and inflectional morphology 18 months and 46 typically developing guage, gestures/play skills, degree of

142 DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OF OUTCOME EXIST? • RESCORLA delay at 2, consonant repertoire, family Grammatic Compledon, and Sentence comprehension. Regression analyses for history of reading problems, and lexical Imitation (Cohen's d values of 0.97, the combined groups indicated that age processing. 1.46, and 1.52, respectively). Ellis 2 LDS vocabulary score explained 20% Weismer [2007] reported age 5'/2 out- of the variance in age 13 vocabulary School-Age Outcome Studies of comes for 40 late talkers identified at age and verbal memory and 14% of the var- Late Talkers 2 (11% with comprehension delays) and iance in age 13 grammar. Although Girolametto et al. [2001] reported 43 typically developing peers. Only three much variance was left unexplained, age 5 outcomes for 21 late talkers iden- of the late talkers scored at least 1 SD these results indicated that slow vocabu- tified at 24—33 months with scores below the mean on TOLD-P3 Speaking lary development at age 2—21/2 was below the 5th percentile on the CDI. Quotient at age 5y2. However, even associated with a weakness in language- The children participated in an 11- with these three children excluded, the related skiUs into early adolescence rela- week parent-based intervention pro- late talkers obtained significandy lower tive to typicaUy developing peers. gram at age 2, and 13 of the 21 had scores than comparison children on both Rescorla [2009] reported language received subsequent speech-language the Listening and Speaking Quotients of and reading outcomes at 17 years of age services. At age 5 foUow-up, most of the TOLD-P3. Group differences were for 26 late talkers and 23 typicaUy devel- the late talkers scored in the normal particularly marked in sentence imita- oping children matched at intake on age, range on various language measures, tion. Performance on a fast mapping task SES, and nonverbal ability. Although late but they scored significandy lower on at 2'/2 explained 36% of the variance in talkers perfomied in the average range on most language measures at age 5 than MLU at 3V2 [EUis Weismer, 2007]; aU language and reading tasks at age 17, children with TL histories, particularly when CDI and Preschool Language they obtained significandy lower vocabu- on measures tapping more complex Scale-3 [PLS-3, Zimmerman et al., lary, grammar, and verbal memory scores skiUs, such as narrating a story. 1992] scores were added as predictors, than peers with TL histories. The largest 65% of the variance was explained. As effect size was a Cohen's d of 1.08 on Paul et al. studied about 30 late would be expected, prediction was talkers identified between 20 and 34 story recaU. With the late talker and com- weaker to age 51/2, with age 2'/2 PLS-3 parison groups combined, age 2 LDS months and a matched comparison scores, CDI vocabulary, and nonverbal group [Paul, 1993, 1996; Paul et al., vocabulary score, entered first, explained abihty accounting for 51% of the var- 17% of the variance in the age 17 vocabu- 1997]. Roughly 25% of the late talkers iance in TOLD-P3 scores. were delayed in receptive as well as ex- lary/grammar aggregate; ReyneU pressive language. At age 3 and age 4, all Rescorla [2002] reported school- Expressive and ReyneU Receptive scores, of Paul's [1993] late talkers scored in the age outcomes for 34 late talkers and 25 entered next, were not significant predic- average range for receptive and expres- comparison children matched at intake tors. Interestingly, Bayley nonverbal sive vocabulary and for receptive on age, socioeconomic status (SES), and score, added last, explained an addidonal grammar. Paul [1996] used a criterion of nonverbal cognitive ability. By age 6 13% of the variance. Results were similar >10th percentile in Developmental Sen- [Rescorla, 2002], only 6% of the late for the verbal memory aggregates, with tence Score [DSS; Lee, 1974] to indicate talkers had scores <10th percentile on the LDS accounting for 17% and Bayley recovery, which might be considered a at least two TOLD-2 [Newcomer and nonverbal score accounting for an addi- rather minimal standard for normal func- HammiU, 1988] subtests. Nevertheless, tional 11% of the variance. Thus, 28-30% tioning. By this criterion, recovery was the late talker group means were signif- of the variance in age 17 language scores achieved by 41% of the sample by age icandy lower than those for typically was explained by two age 2 measures. 3, 57% by age 4, 74% by kindergarten developing comparison children on vo- and first grade, and 84% by second grade cabulary, grammar, phonology, and Summary [Paul, 1996; Paul et al., 1997]. How- verbal memory tasks, with most School-age outcomes firom smaU- ever, the late talken did more poorly Cohen's ds > 0.85. For the combined scale longitudinal studies of late talkers than the typicaUy developing comparison late talker and comparison group sam- indicate that: (a) most late talkers scored in chUdren at age 7 on the TOLD-P2 ples, vocabulary score on the Language the normal range by age 6 or 7 but con- [Newcomer and HammiU, 1988] Ex- Development Survey [LDS; Rescorla, tinued to have significandy weaker pressive language scale, even if they 1989] at age 2, grammar skiUs at age 3 language skiU than typicaUy developing were classified as recovered. The recov- as measured by MLU and IPSyn, and peers through adolescence; (b) significant ered children were not different from grammar skiUs at 5 as measured by the predictors of outcome, which varied controls on receptive language, reading, Patterned Ehcitation Syntax Test across studies, included vocabulary at age spelling, IQ, or phonological skills at age [PEST; Young and Perachio, 1983] 2, nonverbal IQ, and preschool expressive 7, whereas the children who were stül coUectively explained 35% of the var- and recepdve language, but at best only delayed (DSS < 10th percentUe) were iance in age 8 scores on the Chnical about half the variance in school-age out- worse than comparison children on Evaluation of Language Fundamentals comes was explained in these studies. everything except receptive language [CELF-R; Semel et al., 1987]. and reading/speUing [Paul et al., 1997]. Rescorla [2005] reported language OUTCOMES FROM LARGE- Moyle et al. [2007] reported age 5 and reading outcomes at age 13 for 28 SCALE EPIDEMIOLOGICAL outcomes for 30 late talkers first identi- late talkers and 25 typically developing STUDIES fied at age 2 by scores <10th percentile children from the initial intake samples. on the CDI. At age 5, late talkers had sig- As a group, late talkers performed in Outcomes to 18-24 Months for nificantly lower scores than comparison the average range on aU standardized Large-Scale Epidemiological chUdren matched on age, SES, gender, language and reading tasks at age 13, Studies and nonverbal cognitive abihty on three but they scored significantly lower than The Early Language in Victoria TOLD-P3 subtests [Newcomer and SES-matched peers on vocabulary, Study [ELVS; ReiUy et al., 2007] iden- HammiU, 1997]: Oral Vocabulary, grammar, verbal memory, and reading tified 20% of a community sample of

DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OF OUTCOME EXIST? • RESCORLA 143 1,720 2-year-olds as late talkers, based children as reported on the ASQ = 0.83, and parenting stress = 1.61, aU on CDI expressive vocabulary scores [Bricker and Squires, 1999]. A regres- quite smaU effects). <10th percentile based on U.S. norms sion model •with numerous predictors EUis and Thai [2008] summarized [Fenson et al., 1993]. When gender, (being a boy, low birth weight or gesta- age 6 outcomes for a sample of 577 preterm birth, birth weight, birth order, tional age, multiple birth, older sibhngs, children classified at 16 months into SES, maternal mental health, matemal low matemal education, maternal dis- three groups: 461 typicaUy developing vocabulary and education, maternal age, tress/depression, and non-Norwegian children; 81 "late producers" (children family history of speech-language diffi- language background) explained 4—7% with expressive delay only, 14% of the culties, and non-English family of language outcomes, leaving most of sample); and 35 "late comprehenders" background were used as predictors of the variance in language performance at (children with receptive and expressive CDI scores, the model accounted for 18 months unexplained. delay, 6% of the sample). At age 6, only 7% of the variance. When 12- Henrichs et al. [2011] tested pre- 2.2% of the 577 children met criteria month scores on the Communication dictors of 18-month scores on the 112- for specific language impairment (SLI): and Symbolic Behavior Scales [CSBS; word Dutch version of the MacArthur 1.5% of the typicaUy developing group, Wetherby and Prizant, 2002] were Short Form Vocabulary Checklist 3.7% of the late producers, and 8.5% of added to the predictive model, the par- [MCDI-N; Zink and Lejaegere, 2003]. tial R^ was 14%, but most of the the late comprehenders. Of the 13 chil- Translations were available in English variance in 24 months expressive lan- dren diagnosed with SLI at age 6, seven guage remained unexplained. and Turkish; Moroccan parents who had been typicaUy developing, three spoke only Arabic were interviewed at had been late producers, and three had Zubrick et al. [2007] assessed lan- home by Arabic-speaking research assis- been late comprehenders at age 16 guage skiUs at 24 months in a sample of tants. Matemal age, parenting stress, and months. This suggests that early recep- 1,766 Austrahan children from English- child ethnicity explained 1% of the var- tive/expressive delay confers greater risk speaking famihes. Late language emer- iance in 18 months expressive for later SLI than does expressive delay gence (LLE) was diagnosed based on vocabulary, gestational age and birth only, but that more children with SLI parental report obtained by maü on six weight explained an additional 1%, at age 6 had TL histories as toddlers receptive and expressive language items gender and age at the 18-month evalua- than language delays as toddlers. of the Communication scale of the tion explained an additional 4%, and Westerlund et al. [2006] assessed Ages and Stages Questionnaire [ASQ; MCDI-N receptive score at 18 months language development at 18 months Bricker and Squires, 1999] (e.g., point- explained an additional 16%, for 22% in and 3 years for children seen at Swedish ing to pictures on request, using two- total. child health care centers. Parents of 891 or three-word phrases). The criterion of children completed the 90-word Swed- scores >1 SD below the U.S. mean Summary ish Communication Screening at 18 identified 13% of the sample with LLE Large-scale epidemiological stud- months [SCSI8; Ericksson et al., 2002], (238 children). Children with LLE ies predicting language outcomes at 18- with an expressive delay cutoff of <8 were more hkely to also be delayed in 24 months indicate that: (a) demo- ASQ Gross Motor (6 vs. 1%), Fine words. Language delay at age 3 was graphic and birth variables accounted diagnosed if the child manifested lack of Motor (8 vs. 3%), Adaptive (21 vs. 6%), for only a modest percentage of the and Personal-Social (8 vs. 1%) skiUs three-word sentences or failed to com- variance in outcomes at 18 or 24 prehend three of five comprehension than children with TL, suggesting that months and (b) adding in earher or questions posed by health center nurses. the LLE group included children with concurrent developmental variables Only half the children delayed at 3 had ID, ASD, or other developmental dis- accounted for more variance in out- been delayed at 18 months (sensitivity abilities. Multivariate logistic regression comes, but most variance was stiU = 50%). Most of the children identified with LLE as the target outcome indi- unexplained. cated no significant prediction fi-om as delayed at 18 months were not parental education or mental health, delayed at age 3 (positive predictive matemal age, SES, parenting style, or OUTCOMES FROM 16 TO 18 value = 18%). Most children not family functioning. However, signifi- MONTHS FOR LARGE-SCALE delayed at 3 had not been delayed at 18 cant odds ratios (ORs) were obtained EPIDEMIOLOGICAL STUDIES months (specificity of 90%). The area for family history of LLE (2.1), number Horwitz et al. [2003] studied under the curve (AUC) of 77% of sibhngs (2.1), male gender (2.7), pre- 1,189 Connecticut children with ele- obtained via receiver operating charac- mature birth (1.8), <85% optimal birth vated rates of demographic risk factors teristics (ROC) analyses indicated only weight (1.9), and delays on concurrent (e.g., 35% poor, 15% more than one fair prediction. Increasing the 18-month ASQ Gross Motor (3.1), Fine Motor language spoken in the home, and 37% SCS18 cutoff to <15 words raised sen- (2.4), Adaptive (2.6), and Personal- non-white). When language delay was sitivity to 66%, but at the cost of Social (5.5) scores. Zubrick et al. [2007] defined as a CDI short form expressive positive predictive value of only 10% concluded that the results are consistent vocabulary score <10th percentile, rates and specificity of only 73%. Further- with a models of language that "posit of delay were 12.5% at 18-23 months, more, using words comprehended or strong role for neurobiological and 15% at 24-29 months, and 18% at 30- gestures produced rather than words genetic mechanisms" rather than mod- 35 months. Multivariate logistic regres- produced yielded even poorer predic- sion analysis indicated that language els attributing major influence to tion, and combining the three measures delay was associated with demographic matemal and family characteristics. yielded a lower AUC than using pro- risk, but relative risk ratios were <2.0 duction alone. except for being firom a bilingual The Norwegian Mother and Henrichs et al. [2011] reported household (2.78) (e.g., low matemal Child Cohort Study [Schjolberg et al., outcome data at 30 months for 3,759 of 2011] investigated predictors of lan- education = 1.26, poverty = 1.33, mi- nority status = 1.28, number of sibhngs the Dutch chUdren for whom expres- guage delay at 18 months for 42,107 sive and receptive vocabulary delay at

144 DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OE OUTCOME EXIST? • RESCORLA 18 months was deterimned by MCDI and children with persistent delay had Chnical Evaluation of Language Funda- scores <10th percentile. Expressive vo- lower gestational ages and lower birth mentals—Preschool [CELF-P2; Wiig cabulary delay at 30 months was weights than children with no expres- et al., 2006] to identify language delay, defined as a score <10th percentile on sive vocabulary delay and late onset 13% were delayed in expressive lan- the Dutch translation of the LDS expressive vocabulary delay. The persis- guage and 16% were delayed in [Rescorla, 1989]. The correlation tent delay group showed the lowest receptive language, with 21% scoring between MCDI-N expressive scores at nonverbal scores. below this cutpoint on either scale. Af- 18 months and LDS scores at 30 ter excluding children with low months was only 0.34, much lower Summary nonverbal IQs, ASD, hearing impair- than the concurrent correlation of 0.95 Large-scale epidemiological stud- ment, and non-English-speaking between the CDI and the LDS in 2- ies predicting language outcomes from background (NESB), 17% were diag- year-olds [Rescorla et al., 2005]. 16 to 18 months indicate that: (a) lan- nosed with SLI based on receptive and/ Of the 3,759 children in the guage status at 16-18 months had low or expressive delay. A multivariate Henrichs et al. [2011] study, 85% had positive predictive value for later lan- regression model with 12 predictors no expressive vocabulary delay at either guage delay; (b) most children delayed (male, twin, preterm, low birth weight, age, 6% were delayed at 18 months and at later ages had not been delayed at older siblings, NESB, low SES, family not delayed at 30 months (late bloom- 16—18 months (low sensitivity); (c) de- history of language problems, matemal ers), 6% had late onset expressive mographic and birth variables education, maternal mental health prob- vocabulary delay (not delayed at 18 accounted for only a modest percentage lem, matemal age, and matemal months but delayed at 30 months), and of the variance in outcomes at 30 vocabulary) explained 19% of the var- 3% had persistent expressive vocabulary months; and (d) adding in earher lan- iance in CELF-P2 receptive language delay (delayed at both ages). Most chil- guage status accounted for more and 21% in CELF-P2 expressive lan- dren (71%) delayed at 18 months on variance in outcomes, but most variance guage. Adding in language delay at age the MCDI-N scored in the normal was StiU unexplained. 2 increased the variance accounted for range at 30 months on the LDS (posi- to 24% (receptive) and 30% (expressive), indicating stronger prediction from age tive predictive value = 29%), and most OUTCOMES FROM 24 MONTHS children (70%) delayed at 30 months 2 to age 4 than firom 12 months to age IN LARGE-SCALE 2 but StiU much unexplained variance. had not scored below the 10th percen- EPIDEMIOLOGICAL STUDIES tile at 18 months (sensitivity = 30%). The strongest predictor of delayed ex- Armstrong et al. [2007] reported pressive language at 4 was a foreign Most children who scored in the nor- findings through fifth grade for 689 mal range at 18 months continued to language home background, with an children from the NICHD Early Child OR of 6.96. Being a boy (OR = 1.90), score in the normal range at 30 months Care Research Network data set. Three (negative predictive value = 93%), and being a fourth child (OR = 2.36), and groups were defined based on CDI fan-lily history of language problems most children with normal skills at 30 scores at 24 months and Reynell Ex- months also had normal skills at 18 (OR = 1.82) were the next best predic- pressive Language Scale scores [Reynell tors of delayed expressive language. months (specificity = 93%). The ROC and Gruber, 1990] at 36 and 54 curve using MCDI-N expressive vo- months: 131 late talkers, who scored Dale et al. [2003] identified late cabulary scores at 18 months to predict <10th percentile on the CDI and <85 talkers in a sample of UK twins at age LDS delay status at 30 months had an on the Reynell at 36 and 54 months; 2 based on expressive vocabulary scores AUC of 0.74 (95% CI: 0.71, 0.77; p = 39 late bloomers, who scored <10th <10th percentile (<15 words) on the < 0.001), indicating only fair predictive percentile on the CDI and >85 on the 100-word MacArthur Communicative accuracy. Reynell at 54 months; and 558 typically Development Inventory UK Short Henrichs et al. [2011] reported developing children, who scored >10th Form [MCDI: UKSF; Dionne et al., that hierarchical hnear regression analy- percentile on the CDI at 24 month and 2003]. All data were collected via par- sis examining prediction of 30 months >85 on the Reynell at 54 months. Dif- ent reports sent by mail. After vocabulary score yielded significant but ferences between the three groups excluding children with uncertain zygo- small effects (complete model = 18%). persisted through fifth grade on the sity, genetic syndromes, ASD, and non- Matemal age and education, marital sta- Woodcock-Johnson-Revised [WJ-R; Enghsh-speaking famihes, 802 of the tus, fan-lily income, child ethnicity, and Woodcock and Johnson, 1992] Picture 8,386 children (9.6%) were identified parenting stress explained 5% of the Vocabulary, Letter Word Identification, with early language delay (ELD). Com- variance in LDS vocabulary at 30 and Memory for Sentences subtests, pared with the 7,584 children with TL, months; gestational age and birth although the late talker group scored in the age 2 ELD group had more boys weight explained an additional 0.2%; the average range on the first two subt- (65 vs. 47%) and lower nonverbal abil- gender and age at the 18- and 30- ests at all time points. For all measures, ity, grammar, and "displaced reference" month assessments explained an addi- the late talker group performed worst, scores. Their mothers also had lower tional 1%; 18 months MCDI-N the typically developing group per- educational attainment. However, on expressive 2r-scores explained an addi- formed best, and the late bloomer aU these measures, there was consider- tional 11%; and 18 months MCDI-N group perfonned in between the other able variation within the ELD group. receptive 2--scores explained an addi- two groups, with httle change in the The largest effect size was for nonverbal tional 0.5%. Children in the late onset gaps between groups over time. abihty (5.7%). group and the persistent delay group Outcomes fi-om age 2 to age 4 When a score <15th percentile tended to have lower fan-iily income, were reported for 1,596 of the children on two of three language measures (vo- less educated mothers, and more paren- in the Austrahan ELVS cohort [Reilly cabulary, grammar, and abstract tal stress than children in the no delay et al., 2010]. Based on a cutpoint of language) was used to identify ELD at and late bloomer groups. Late bloomers >1.25 SDs below the mean on the age 3, 11% of the sample was delayed

DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OF OUTCOME EXIST? • RESCORLA 145 (835 children), 61% of whom had not given. However, they had significantly robust finding across many studies is been delayed at 2; 44% of the 2-year- lower scores than the NLE group on that most late talkers attained language olds with ELD were still delayed at 3. receptive vocabulary, articulation, and scores in the average range by age 5, 6, When the same criterion was used at numerous grammatical scales, with the or 7 [Fischel et al, 1989; Paul et al., age 4, 12% of the sample was delayed largest Cohen's ds on the morphosyntax 1997; Girolametto et al., 2001; in language (746 children), only 34% of measures. The percentages of LLE chil- Rescorla, 2002, 2005, 2009; EUis whom had been delayed at 2; 40% of dren who were impaired (—1 SD Weismer, 2007; EDis and Thal, 2008; the 2-year-olds with ELD were stül below the mean) ranged fi-om 4 to 23% Rice et al., 2008]. Furthermore, most delayed at 4. across the 17 outcome measures, with late talkers scored in the average range Dale et al. [2003] used logistic significant group differences for seven by elementary school even on the tasks regression analyses to determine deci- of the measures. that appear to be the most chaUenging sion statistics for prediction from age 2 for them, namely grammatical and to ages 3 and 4. Age 2 vocabulary, dis- Summary verbal memory measures [Rescorla, placed reference, and nonverbal abüity Large-scale epidemiological stud- 2002; Rice et al., 2008]. scores as well as gender and maternal ies predicting language outcomes firom Another very robust finding from education were used in a prediction 24 months indicate that: (a) most late outcome studies is that late talker model for age 3 and age 4 delay for the talkers scored in the nomial range by groups consistently obtained signifi- age 2 children with ELD. Sensitivity age 6 or 7 but continued to have signif- cantly lower scores than groups with was 42% at age 3 and 52% at age 4, icantly weaker language skuls than TL histories on most language meas- meaning that half or fewer of the chil- typicaDy developing peers; (b) few sig- ures, even when the late talkers dren delayed at foUow-up were nificant predictors of outcome were performed in the average range. Paul predicted to be delayed based on the found, and (c) positive predictive value [1996] and Rescorla [2002] were model. Similarly, only 57% of the chil- firom age 2 was generaUy low. dren predicted to be delayed by the among the first to report this finding, model at age 3 (and 64% at age 4) were which was rephcated in Moyle et al. CONCLUSIONS actually delayed at those ages (positive [2007], EUis Weismer [2007], Rescorla predictive value). When the full age 2 [2005, 2009], Thai et al. [2005], Arm- sample was used, sensitivity was very Outcomes of Late Talkers strong et al. [2007], and Rice et al. poor (19% at both ages), as was positive Although several studies have [2008]. This finding provides support predictive value (53% at 3 and 58% at reported outcomes for late talkers iden- for the dimensional account of language 4). In sum, language delay at age 2 and tified between 18 and 35 months of delay, according to which late talkers a number of additional factors poorly age, findings vary as a flinction of age and children with TL development dif- predicted language delay at a later age, at intake, age at foUow-up, composition fer quantitatively on a hypothetical more than half of children with ELD at of the sample, and outcome measures language abihty spectrum [Rescorla, 2 were in the normal range at ages 3 or used. For example, late talkers identified Fischel, 2009]. 4, and most children with language at 18 months tend to have smaller per- As articulated by Rescorla [2002, delay at ages 3 and 4 had normal lan- centages of persistent delay by age 3 2005, 2009], the language abihty spec- guage at age 2. than late talkers identified at 24 months, trum, like inteUigenee, can be as seen in rates of 18% in Westerlund conceptualized as deriving from varia- Because the Dale et al. [2003] et al. [2006] and 29% in Henrichs et al. tion in many discrete skiUs. The distinct sample was comprised of twins, the rel- [2011] versus 44% in Dale et al. [2003]. yet interrelated abihties hypothesized to ative contributions of genetic and Furthermore, the older late talkers are subserve language include auditory per- environmental factors associated with at foUow-up, the larger the percentage ception/processing, word retrieval, language delay could be tested. Bishop who wul be recovered. For example, verbal working memory, motor plan- et al. [2003] concluded that shared the percentage of Paul's late talkers ning, phonological discrimination, and who scored in the normal range for environmental factors accounted for grammatical rule leaming. This notion syntax increased from 41% at age 3 to more variance in age 2 ELD (71-72%) of a speetmm of language abihty is con- than did genes (22-25%), regardless of 84% at age 7 [Paul, 1996]. With respect sistent with Leonard [1991] and Bishop whether the children had a persistent or to sample composition, EUis and Thai and Edmundson [1987]. EUis Weismer a transient delay as measured by parent [2008] reported that 8.5% of late talkers [2007] has related the notion of a lan- report at ages 3 and 4. with slow comprehension had persistent delay firom 16 months to age 6 versus guage endowment speetmm to the Rice et al. [2008] reported age 7 broader theoretical debate regarding outcome data for children with LLE only 3.7% of those with Fischel normal comprehension. FinaUy, vocabulary continuous versus dichotomous charac- and children with normal language terizations of language impairment. emergence (NLE) fi-om the Zubrick measures tend to show better outcomes et al. [2007] study. At age 2, 128 chil- than grammar measures, and grammar measures vary in their degree of strin- Success in Predicting Late Talker dren had expressive language delay Outcomes (19% of the sample), 88 of whom also gency. For example, the percentage of Rescorla's 34 late talkers scoring >16th When predicting language delay had receptive delays at age 2. Exclusion as a dichotomous outcome, both false criteria included ID, ASD, deafness. percentile was 79% on the EOPVT, 58% on the ReyneU Expressive Lan- positive and false negatives have been Down syndrome, and cerebral palsy. At common. The false positive rate for late age 7, the children with LLE did not guage Scale, 35% on MLU, and 24% on the IPSyn [Rescorla et al., 1997]. talkers is often very high (i.e., late talker differ firom the NLE group on SES/de- scoring in the nomial range at out- mographic variables or on nonverbal Despite the variations in out- come), as indicated by generaUy low inteUigenee, and they scored in the nor- comes reported across both smaU-scale positive predictive values, such as 18% mal range on all language measures and epidemiological studies of ELD, a in Westerlund et al. [2006]; 29% in

146 DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OF OUTCOME EXIST? • RESCORLA Hendchs et al. [2011]; 44% (age 3) and for fijrther research, which can best be 15 years for the Rescorla age 17 results) 40% (age 4) in Dale et al. [2003]; 64% explored in diverse epideniiological than for epideniiological studies span- in Feldman et al. [2005]; and 39% in samples that are being foUowed from 18 ning short dme pedods is that the Fernald and Marchman [2012]. Because or 24 months, such as the Hendchs smaU-scale studies, by design, had most late talker outcome studies do not et al. [2011] Dutch sample or the Reilly sharply contrasdng groups. The late report intervention infomiadon, it is et al. [2007] Australian sample. talkers were selected to have severe difficult to determine the extent to As Hendchs et al. [2011] noted, delays (generaUy 1.5 SDs below age which low positive predictive values are poor decision statistics may dedve in expectations) and the compadson chil- due to treatment success versus sponta- part from imposing a fixed cutpoint on dren were selected to have solidly neous remission. an underlying continuum, whereby average or even above average language High rates of false negative errors children just missing the cutpoint (i.e., skiUs. Therefore, even though the late have also been widely reported in at the 11th percentile) are classified as talkers eventually performed in the av- large-scale epideniiological late talker normal. However, Hendchs et al. erage range, they continued to have studies, in that the majodty of children [2011] also noted that the poor predic- weaker skills than the compadson chil- with language delays at foUow-up had don in their sample was not only dren and the vadables that inidaUy TL development at intake. Although attdbutable to dichotoniization, because differentiated the two groups continued most children with typical development the correlation between 18-month to differentiate them over time. By at intake continue to have normal lan- MCDI-N and 30-nionth LDS was only contrast, the epideniiological samples guage skiUs at foUow-up [e.g., negative 0.34, and ROC analysis, which tests all represented the fuU spectrum of lan- predictive value of 93% in Hendchs cutpoints in a continuous fashion, guage skiUs, with the late talkers at the et al., 2011], children who were not yielded an AUC of only 74%. tail of a continuous distribution (e.g., the bottom 10%) and the rest of the late talkers at intake generally compdse Prediction of outcomes can also sample considered "normal." Further- the majodty of children delayed at out- utihze continuous scores. Multiple more, the epidemiological samples come, e.g., 70% in Hendchs et al. regression analyses with smaU-scale tended to be much more diverse in [2011], 54% in EUis and Thai [2008], studies involving late talker and typi- terms of SES than the smaU-sample 50% in Westedund et al. [2006], and cally developing groups have generally studies, and the follow-ups to date have 61% (age 3) and 66% (age 4) in Dale yielded moderately strong prediction. been quite short-term. It may be the et al. [2003]. However, this pattern has For example, EUis Weismer found that case that when diverse samples such as not generally been found in the smaU- 30-month fast mapping performance, the Dutch, UK, and Australian samples scale late talker studies [e.g., Paul, 1996; CDI vocabulary, and PLS scores are foUowed up into the elementary EUis Weismer, 2007; Rescorla, 2009], explained 65% of the vadance in age 5 school years, family SES factors will where very few children in the typically language outcome. Rescorla [2009] play an increasingly important role in developing compadson groups mani- reported that 30% of the vadance in accoundng for vadation in language fested language delays at any foUow-up age 17 vocabulary/grammar scores was skiUs. age. The reason for this difference predicted by LDS score and nonverbal between epideniiological and smaU-scale ability at age 2. Fernald and Marchman late talker studies is that the small-scale [2012] reported that faster reaction time Predictors of Outcome studies selected typicaUy developing and greater accuracy in a lexical proc- Although several studies have compadson groups to have solidly aver- essing task at 18 months predicted tested predictors of outcome for late age or even above average language steeper acceleration in vocabulary talkers, results have been inconsistent. skiUs, rather than just being above a growth among late talkers fiom 18 to For example, although Bishop et al. 10% percentile cutoff. The only small- 30 months. In contrast, regression anal- [2003] reported that shared environ- scale study in which many children yses with epideniiological samples have ment impacted language delay and with delays at 3 had been typically been able to account for rather smaU Horwitz et al. [2003] indicated strong developing at age 2 was Feldman et al. percentages of vadance in outcome demographic effects on language delay, [2005], which used a CDI score >10th measures. For example, Hendchs et al. two large Australian studies found that percentile to define typical development [2011] found that only 22% of the var- a host of environi-nental factors at 2 and which contained many chil- iance in 18 months expressive explained very little vadance in lan- dren from low SES families, known to vocabulary score was explained (6% by guage skiUs [ReiUy et al., 2007; be at dsk for language delays with demographic/perinatal factors and 16% Zubdck et al., 2007]. It would seem increasing age [Hart and Risley, 1995]. by concurrent receptive vocabulary that family income, marital status, eth- scores). Reilly et al.'s [2007] model pre- EUis Weismer [2007] has high- nicity, parendng stress, and matemal dicted only 14% of the vadance in CDI lighted the puzzMng mismatch between educational level should be among the scores at 24 months (7% by demo- the smaU percentage of late talkers who most important factors for explaining graphic factors and 7% by 12-month manifest SLI at age 5 and the 7% of individual differences in language devel- communication scores on the CSBS). kindergarten children identified with opment. However, across most studies Zubdck et al. [2007], using logisdc SLI [TombHn et al., 1997]. As Ellis reviewed, their predicdve abuity has regression to predict LLE versus NLE at Weismer so cogendy states, "Given the been quite modest. As noted above, 24 months, reported that demographic reladvely low proportion of late talkers longer temi foUow-ups in these large factors were not significant predictors who display clinical language impair- epidemiological samples are likely to and that the highest ORs were for con- ment at school entry, we must continue indicate increasing effects of SES with current ASQ scores. to ask where those 7% of kindergarten age. children with SLI come fiom if not In summary, the likely reason that Reproductive factors such as ges- from the ranks of late talkers" (p. 95). regression results are stronger for smaU- tadonal age, birth weight, and pednatal This apparent paradox should be a focus scale studies spanning many years (i.e.. complications have also explained very

DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OF OUTCOME EXIST? • RESCORLA 147 little variance in later language skiUs, of the variance. Clearly, however, cates. Most late talkers catch up to consistent with many earher studies of much variance was left unexplained. normative expectations, although some perinatal risk. This is probably because, An excidng possibility in recent so not do so untU age 5 or later and as proposed by Sameroff and Chandler late talker research is that some of the their language skiUs continue to be [1975], the "continuum of caretaker unexplained variance in late talker out- weaker than those of their peers, on av- casualty" is more potent than the comes may be accounted for by erage. Furthermore, the existing "continuum of reproductive casualty" language processing skiUs. As Femald hterature strongly suggests that most in determining developmental out- and Marchman [2012] and Marchman children with language delays at age 5 comes. As Sameroff and Chandler and Femald [2008] have shown, reac- were not late talkers. Therefore, early noted, transacdonal effects between tion time and accuracy in a lexical identification and intervention with late caregivers and children serve to magnify processing task at 18 months signifi- talkers wiU not have much impact on the effects of reproductive risk in more candy predicted later vocabulary preventing SLI in school-age children. disadvantaged famihes but to minimize growth in both typicaUy developing these effects in higher SES families. An implication of this review is children and late talken, as weU as pre- that expressive language screening at 18- FaniUy history of language delay dicting age 8 working memory. These 35 months can serve an important pubhc significandy increased the odds of being findings are noteworthy, because verbal health function because it can identify a late talker at 2 in the Zubrick et al. memory deficits are among the most chUdren whose expressive delay is sec- [2007] sample (OR = 2.1) and at 4 in robust and enduring weaknesses mani- ondary to another disabihty. This is the Reilly et al. [2010] sample (OR = fested by late talkers, even when they particularly important for children with 1.8), and history of language delay in perform in the normal range at foUow- receptive language delay and weak non- the family has also been commonly up. Femald and Marchman [2012] verbal skiUs but who do not have ID or reported in smaU-scale late talker stud- acknowledged that variation in lexical ASD and so might not be identified by ies. Lyytinen et al. [2001] found that processing could be due to endogenous other screening procedures. ChUdren genetic risk for dyslexia was strongly factors, but they also highlighted that with receptive delays are fairly hkely to associated with persistent delay in late variation in matemal speech to children have continuing language delay, as weU talkers identified at 2. Although these has been demonstrated to predict later as being at risk for later leaming and be- findings suggest that ELD may have a lexical processing skiU at 24 months. It havioral/emotional problems. Therefore, major genetic component. Bishop et al. is hkely that children vary in their they are the late talken most in need of [2003] noted that the magnitude of the inborn auditory processing abihty, but early intervention. genetic effect was very dependent on that this abihty is also shaped by subse- the particular outcome measure used to Although not providing definitive quent language environment, such that evidence, the current review suggests determine language delay as weU as on chUdren with richer language environ- the cohort analyzed. that demographic risk associated with ments develop in their abihty to low SES may become more important Being a boy is one of the best efficiendy process lexical information as a causal factor in language delay as predictors of being a late talker in epi- more than children with less rich lan- chUdren get older. In contrast, individ- demiological studies. SmaU-scale late guage environments. ual differences in basic language talker samples typicaUy have many endowment seem more important as a more boys than girls, and late talkers in Implications primary causal factor in language delay epidemiological studies also are more When expressive language delay from 18 to 35 months. To be sure, often boys than girls. Because girls typi- is secondary to a more primary disabil- chUdren from very disadvantaged fami- caUy have larger reported vocabularies ity, such as autism or ID, then the hes, or from famihes where more than than boys, using the same delay crite- likelihood of continuing delay is high one language is spoken, may show rion for both genders (i.e., <50 words and the benefits of early intervention delays in this 18- to 35-month period, or <10th percentUe on the CDI) wiU are substantial. AdditionaUy, when ex- and the more such famihes there are in almost always yield more boys than pressive language delay is the result of a large, diverse sample, the more hkely girls. If a gender-specific cutpoint is significant environmental neglect or demographic factors are to emerge as used (i.e., 10th percentile on the CDI abuse, it is important to intervene in significant predictors. by gender), then genders wiU be equal order to protect the child and provide a In summary, language is a com- in the late talker group, but the mean more supportive and stimulating envi- plex set of skiUs. A large number of vocabulary score of the girl late talkers ronment. To the extent that children biopsychosocial factors are most likely wiU be higher than the score of the boy with ID, ASD, or environmental depri- responsible for individual differences in late talkers. vation would not be otherwise the skiUs that subserve language abihty. The best predictors of expressive identified, picking them up in an ex- Early in development, biological factors language appear to be earher expressive pressive language screening would have may have the most important influence, language skiUs. This was found in both some pubhc health benefit. One might as indicated by the famUial aggregation smaU-scale studies, such as those by also make this argument for late talkers of language problems. However, inborn Rescoda 2002, 2005, 2009] and EUis who are delayed in receptive language, characteristics are influenced by envi- Weismer [2007], as yveU as in epidemi- given that receptive/expressive delay is ronmental factors. Therefore, by the ological studies [Lyytinen et al., 2001; often associated with nonverbal deficits time children start talking, their capacity ReiUy et al., 2007, 2010; Henrichs and generaUy presages a poorer out- for language learning reflects both their et al., 2011]. For example, Henrichs come than expressive delay alone. biological endowment and their linguis- et al. [2011] reported that the strongest On the other hand, children tic environment. As chUdren enter the predictor of 30 months expressive whose only developmental disability at preschool period, psychosocial factors vocabulary was 18 months expressive 18-35 months is an expressive language such as how much (and how weU) vocabulary, which accounted for 11% delay are at less risk, as this review indi- parents talk to them probably become

148 DEV DISABIL RES REV. LATE TALKERS: DO GOOD PREDICTORS OE OUTCOME EXIST? • RESCORLA increasingly important. Thus, children FischelJ, Whitehurst G, Caulfield M, et al. 1989. Rescorla L. 2009. Age 17 language and reading who manifest SLI at age 5 most likely Language growth in children with expressive outcomes in late-talking toddlers: support language delay. Pediatrics 82:218-227. for a dimensional perspective on language represent a mixture of children, some Gardner MF. 1981. Expressive one-word picture delay. J Speech Lang Hear Res 52:16-30. who have been delayed since they were vocabulary test. Novato, CA. Academic Rescorla L, Achenbach TM. 2002. Use of the toddlers, as weU as some who are at Therapy Pubbcadons. language development survey (LDS) in a psychosocial risk. Children with SLI at Girolametto L, Wiigs M, Smyth R, et al. 2001. nadonal probability sample of children 18 to 5 typicaUy have poorly developed lis- Children with a history of expressive lan- 35 months old. J Speech Lang Hear Res tening skiUs, limited vocabularies, weak guage delay: outcomes at 5 years of age. Am 45:733-743. J Speech-Lang Pathol 10:358-369. Rescorla L, Alley A. 2001. Validadon of the lan- grammatical skiUs, and poorly devel- Hart N, Risley T. 1995. Meaningful differences guage development survey (LDS): a parent oped higher level language abilities such in everyday experiences of young American report tool for identifying language delay in as defining, describing, and narrating. children. Baldmore: Paul H. Brookes Pub- toddlers. J Speech Lang Hear Res 44:434— lishing Co. 445. REFERENCES Henrichs J, Rescorla L, Schenck J, et al. 2011. Rescorla L, Dahlsgaard K, Roberts J. 2000. Late- Examining continuity of early expressive vo- talking toddlers: MLU and IPSyn outcomes Amistrong ES, Marchman VA, Owen MT. 2007. cabulary development: the Generation R at 3;0 and 4;0. J Child Lang 27:643-664. School age cognitive and achievement out- study. J Speech Lang Hear Res 54:854-869. Rescorla L, Lee EC. 2000. Language impaimients comes for late talkers and late bloomers: do Horwitz SM, Irwin JR, Briggs-Gowan MJ, et al. in young children. In: Layton T, Watson L, late bloomers really bloom? Poster presented 2003. Language delay in a community editors. Handbook of early language impair- at American Speech-Hearing Associadon cohort of young children. J Am Acad Child ment in children: Volume I: Nature. New Annual Conference, Boston, November Adolesc Psychiatry 42:932-940. York: Delniar Publishing Company, pp. 1— 2007. Hurtado N, MArchman VA, Femald A. 2008. 38. Bishop DVM, Edmundson A. 1987. Language Does input influence uptake? Links between Rescorla L, Ratner NB, Jusczyk P, et al. 2005. impaired 4-year-olds: Disdnguishing tran- matemal talk, processing speed, and vocabu- Concurrent validity of the language devel- sient from persistent impairment. J Speech lary size in Spanish-leaming children. Dev opment survey: associadons with the Mac- Hear Disord 52:156-173. Sei 11:F31-F39. Arthur-Bates communicadve development Bishop DV, Price TS, Dale PS, et al. 2003. Out- Lee L. 1974. Developmental sentence analysis: a inventories: words and sentences. Am J comes of early language delay. II. Etiology gramniadcal assessment procedure for speech Speech-Lang Pathol 14:156-163. of transient and persistent language difFicul- and language clinicians. Evanston, IL: Rescorla L, Roberts J, Dahlsgaard K. 1997. Late des. J Speech Lang Hear Res 46:561—575. Northwestern University Press. talkers at 2: outcome at age 3. J Speech Bricker D, Squires J. 1999. Ages and Stages Leonard LB. 1991. Specific language impairment Hear Res 40:556-566. Questionnaires: a parent-completed, child- as a clinical category. Lang Speech Hear ReyneU JK. 1977. ReyneU Developmental Lan- monitoring system, 2nd ed. Baldmore: Paul Serv Schools 22:66-68. guage Scales. Windsor, UK: NFER. H. Brookes. Lyydnen P, Poikkeus A, Laakso M, et al. 2001. Lan- ReyneU JK, Gruber CP. 1990. ReyneU Develop- Buschmann A, Jooss B, Rupp A, et al. 2008. guage development and symbolic play in chil- mental Language Scales, U.S. edidon. Los Children with developmental language delay dren with and without familial risk for dyslexia. Angeles: Westem Psychological Services. at 24 months of age: results of a diagnosdc J Speech Lang Hear Res 44:873-885. Rice ML, Taylor CL, Zubrick SR. 2008. Lan- work-up. Dev Med Child Neurol 50:223- Marchman VA, Femald A. 2008. Speed of word guage outcomes of 7-year-old children with 229. recognidon and vocabulary knowledge in or without a history of late language emer- Dale PS, Price TS, Bishop DV, et al. 2003. Out- infancy predict cognitiven and language gence at 24 months. J Speech Lang Hear comes of early language delay. 1. Predicdng outcomes in later childhood. Dev Sei Res 51:394-407. persistent and transient language difFiculdes 11:P9-F16. Sameroff AJ, Chandler MJ. 1975. Reproducdve at 3 and 4 years. J Speech Lang Hear Res Moyle MJ, Ellis Weismer S, Lindstrom M, et al. risk and the condnuum of caretaker casualty. 46:544-560. 2007. Longitudinal reladonships between In: F.D. Horowitz, M. Hetherington, S. Dionne G, Dale PS, Boivin M, et al. 2003. lexical and grammatical development in typ- Scarr-Salapatek, G. Siegel (Eds). Review of Genedc evidence for bidirecdonal effects of ical and late talking children. J Speech Lang child development research. Vol. 4. Chi- early lexical and grammatical development. Hear Res 50:508-528. cago: Univenity of Chicago Press. Child Dev 74:391-412. Newcomer P, HanimiU D. 1988. Test of language Scarborough HS. 1990. Index of productive syn- EUis EM, Thai DJ. 2008. Early language delay developiiient-2 (primary). Ausdn, TX: tax. Appl Psycholinguist 11:1—12. and risk for language impainnent. Perspec- Pro-Ed. Schjolberg S, Eadie P, Zachrisson H, et al. 2011. dves 15:93-100. Newcomer P, HamniiU D. 1997. Test of language Predicdng language development at 18 EUis Weismer S. 2007. Typical talken, late talkers, development-3 (primary). Ausdn, TX: months: data fi-om the Norwegian Mother and children with specific language impair- Pro-Ed. and Child Cohort Study. J Dev Behav ment: a language endowment spectrum? In: Paul R. 1993. Outcomes of early expressive lan- Pediatrics 32:375-383. Paul R, editor. The influence of develop- guage delay. J Childhood Commun Disord Semel E, Wiig E, Secord W. 1995. CUnical evalu- mental perspectives on research and pracdce 15:7-14. adon of language fundamentals, 3rd ed. San in communication disorders: a festschrift for Paul R. 1996. Clinical implications of the natural Antonio, TX: Psychological Corporadon. Robin S. Chapman. Mahwah, NJ: Lawrence history of slow expressive language develop- Semel E, Wiig E, Secord W, et al. 1987. Clinical Erlbaum Associates, pp. 83-101. ment. AmJ Speech-Lang Pathol 5:5—21. evaluadon of language fundamentals-revised. Ericksson M, Westerkund M, Berglumd E. 2002. Paul R, Murray C, Clancy K, et al. 1997. Read- San Antonio, TX: Psychological Corporadon. A screening version of the Swedish Com- ing and metaphonological outcomes in late Thai DJ, MiUer S, Carlson J, et al. 2005. Non- municadve Development Inventories talkers. J Speech Lang Hear Res 40: word repedtion and language development designed for use with 18-month-old chil- 1037-1047. in 4-year-old children with and \vithout a dren. J Speech Lang Hear Res 45:948-960. ReiUy S, Wake M, Bavin EL, et al. 2007. Predicdng history of early language delay. J Speech Feldman HM, Dale PS, CampbeU TF, et al. language at 2 years of age: a prospecdve com- Lang Hear Res 48:1481-1495. 2005. Concurrent and predicdve validity of munity study. Pediatrics 120:el441-1449. Thai DJ, Tobias S, Morrison D. 1991. Language parent reports of child language at ages 2 ReiUy S, Wake M, Ukoumunne OC, et al. 2010. and gesture in late talkers: a 1-year foUow- and 3 years. Child Dev 76:856-868. Predicting language outcomes at 4 yean of up. J Speech Hear Res 34:604-612. Fenson L, Dale PS, Reznick JS, et al. 1993. The age: findings from Early Language in Victo- Tomblin JB, Records NL, Buckwalter P, MacArthur Communicadve Development ria Study. Pediatrics 126:1530-1537. Zhang X, Smith E, O'Brien M. 1997. The Inventories: user's guide and technical man- Rescorla L. 1989. The Language Development prevalence of specific language impaimient ual. Baltimore: Paul H. Brookes Publishing. Survey: a screening tool for delayed lan- in kindergarten children. J Speech Lang Femald A, Marchman VA. 2012. Individual dif- guage in toddlen. J Speech Hear Disord Hear Res 40:1245-1260. ferences in lexical processing at 18 months 54:587-599. US Prevendve Services Task Force. 2006. Screen- predict vocabulary growth in typicaUy Rescorla L. 2002. Language and reading out- ing for speech and language delay in pre- developing and late-talking toddlers. Child comes to age 9 in late-talking toddlers. J school chUdren: recommendadon statement. Dev 83:203-222. Speech Lang Hear Res 45:360-371. Pediatrics 117:497-301.

DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OE OUTCOME EXIST? • RESCORLA 149 Westerlund M, Berglund E, Eriksson M. 2006. Whitehurst GJ, Fischel JE, Arnold DS, et al. 1992. Zimmerman 1, Steiner V, Pond R. 1992. Pre- Can severely language delayed 3-year-olds Evaluating outcomes with children with ex- school Language Scale-3. Chicago: Psycho- be idendfied at 18 months? Evaluadon of pressive language delay. In: Warren SF, Reichle logical Corporation. a screening version of the MacArthur- J, editors. Causes and effects in communication Zink I, Lejaegere M. 2003. N-CDIs: Korte vor- Bates Communicative Development and language intervention. Vol. 1. Baltimore, men, aanpassing en hemonnering van de Inventories. J Speech Lang Hear Res MD: Brookes Publishing p 277-313. MacArthur Short Fonn Vocabulary Checklists 49:237-247. Wiig E, Secord W, Semel E. 2006. Clinical evalua- van Fenson et al. [N-CDIs: short forms, adap- Wetherby AM, Prizant BM. 2002. Communica- tion of language iundamentals-preschool-2. tation and nonns of the Dutch version of the tion and symbolic behavior scales. Balti- San Antonio, TX: Psychological Corporation. MacArthur Short Fomi Vocabulary Checklists more, MD: Paul H. Brookes Publishing. Woodcock RW, Johnson MB. 1989. Woodcock- of Fenson et al.]. Leuuven, Belgium: Acco. Johnson psychoeducational battery-revised. Zubrick SR, Taylor CL, Rice ML, et al. 2007. Whitehurst G, Fischel J. 1994. Early developmen- Chicago: Riverside Publishing. Late language emergence at 24 months: an tal language delay: what, if anything, should Young EC, Perachio JJ. 1983. Patterned eUcita- epidemiological study of prevalence, predic- the clinician do about it? J Child Psychol tion syntax test, revised. Tucson, AZ: Com- tors, and covariates. J Speech Lang Hear Psychiatry 35:613-648. munication Skill Builders. Res 50:1562-1592.

150 DEV DISABIL RES REV- LATE TALKERS: DO GOOD PREDICTORS OF OUTCOME EXIST? • RESCORLA Copyright of Developmental Disabilities Research Reviews is the property of John Wiley & Sons, Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.