Journal of Early Intervention http://jei.sagepub.com Testing Predictive Validity of the Communication Composite of the Communication and Symbolic Behavior Scales Rebecca B. McCathren, Paul J. Yoder and Steven F. Warren Journal of Early Intervention 2000; 23; 36 DOI: 10.1177/10538151000230010801 The online version of this article can be found at: http://jei.sagepub.com/cgi/content/abstract/23/1/36 Published by: http://www.sagepublications.com On behalf of: Division for Early Childhood of the Council for Exceptional Children Additional services and information for Journal of Early Intervention can be found at: Email Alerts: http://jei.sagepub.com/cgi/alerts Subscriptions: http://jei.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations (this article cites 20 articles hosted on the SAGE Journals Online and HighWire Press platforms): http://jei.sagepub.com/cgi/content/refs/23/1/36 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. JournalofEarlyIntervention,2000 Vol.23,No.1,36–46 Copyright2000bytheDivisionofEarlyChildhood,CouncilforExceptionalChildren Testing Predictive Validity of the Communication Composite of the Communication and Symbolic Behavior Scales REBECCA B. MCCATHREN University of Missouri PAUL J. YODER Vanderbilt University STEVEN F. WARREN University of Kansas This study tested the predictive validity of the Communication Composite of the Communication and Symbolic Behavior Scales as a predictor of expressive vocabulary. The Communication Composite consists of six clusters that measure specific aspects of communicative behavior. Participants were 58 children, 17 to 34 months of age, who were functioning at the prelinguistic stage of language development. All children had mild to moderate developmental delays but no sensory impairments. The Communication Composite was used to measure prelinguistic communication skills at the beginning of the study, and an unstructured play session was used to measure expressive vocabulary 1 year later. Results indicated that the Communication Composite was a significant predictor of later expressive language. In addition, all tested clusters were significant predictors of later expressive language. Early identification of language delays and cisions about intervention services are often disorders is unquestionably important. Lan- based on assessments, and using instruments guage delays and disorders can resultinlong- with low predictive validity is likely to result term negative effects on peer relationships in children receiving services they do not (Baker & Cantwell, 1982), on behavioral and need, or being denied services they do need. emotional development (Beitchman, Nair, This is particularly true in the area of com- Clegg, Ferguson, & Patel, 1986), and on municationbecauseitaffectssocial,cognitive, school achievement (Silva, Williams, & and academic achievement. McGee, 1987). Unfortunately,anhistoriclack Developing valid prelinguistic assessments of valid and reliable communication assess- poses many difficulties. For example, identi- ments for young children has resulted in lan- fying communicative behaviors that are pre- guage delays and disorders often not being dictive of later language development is chal- identifiedearly(McCathren,Warren,&Yoder, lenging. Developing measurement strategies 1996).Aninstrumentwithgoodpredictiveva- and assessment instruments that will motivate lidity, that is, an assessment instrument that young children to engage in observable com- canpredictsomefutureskillorability(ordis- munication behaviors presents another chal- ability) is particularly relevant to the identifi- lenge, and a related concern is the need for cation of children in need of early language valid and reliable measurement strategies. A and communication intervention. Clinical de- new generation of communication assessment 36 JEI, 2000, 23:1 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. instruments,however,isattemptingtoaddress CSBS assigns Likert scores to all scales, re- these problems. ducing the raw score variability betweenchil- Inthisstudy,wetestedthepredictivevalid- dren. For example, in the Communicative ity of Cluster scores and the Communication Functions cluster, a child who uses joint at- Composite score of the Communication and tention 19 times in a session is assigned a 4 Symbolic Behavior Scales (CSBS; Wetherby on the Likert scale, and the same score given & Prizant, 1993) as a predictor of later ex- to a child who uses joint attention 51 times. pressive vocabulary for a group of toddlers Although it is unclear whether the difference with developmental delays. The CSBS is not between 19 and 51 is clinically meaningful, a screening instrumentused withlargegroups such use of the Likert scale may mask differ- of children, but an assessment instrumentthat ences in performance that are necessary for provides detailed information about the com- accurate prediction. municationskillsofayoungchild.Westudied Asecondpotentialmeasurementproblemis children with developmental delay because the use of behavior frequencies instead of they are a very important clinical population rates. A frequency represents only a count of for which the CSBS is intended. We selected how many times a behavior occurs in a par- the CSBS because it measures early commu- ticular context. Frequency, however, does not nication in a complex and thorough way. consider amount of time. Rate, on the other The CSBS is an assessment instrument de- hand,considersthefrequencyofbehaviorand signedtobeadministeredbytrainedclinicians within a time unit (e.g. rate (cid:1) frequency di- to children from 8 to 30 months of age. The vided by unit of time). For example, in our CSBS measures young children’s communi- sample,theassessmentforsomechildrentook cation skills during a set of interactions with nearly30minuteswhileforothersittookonly adults and objects that allow a range of be- 11 minutes. Using only rate as the measure, haviors to be elicited and coded. This method thechildwhousedathreepointgaze10times of assessment allows a much more complex in 11 minutes received the same score as the analysis of the children’s communicationthan child who used a three point gaze 10 times in checklist type assessments (e.g. Hawaii Early 25 minutes. It is unclear, however, whether Learning Profile, Parks et al., 1984; Recep- both children demonstrated the same level of tive-Expressive Emergent Language Scale, skill. Theoretically, the CSBS protocol struc- Bzoch & League, 1978). The Communication tures the adult behavior so that each child is Composite of the CSBS is derived from a se- offered the same number of responseeliciting ries of situations designed to elicit jointatten- situations. If true, a constant number of op- tion (usually in the form of comments), be- portunities to respond may be moreimportant havior regulation (usually in the form of re- than the amount of time the child takes to re- quests), and social interaction during games. spond. The opportunities for some of the TheCommunicationCompositescoreismade scored behaviors, however, are not clear (e.g. up of six cluster scores: Communicative positive affect and gaze shift). Means-Vocal, Communicative Means-Gestur- A third potential problem is the inclusion al, Communicative Means-Verbal, Social-Af- of scales that may not actually be valid pre- fective Signaling, Reciprocity, and Commu- dictors of later language development. Al- nicative Functions. Each cluster is composed though the Communication Composite in- of three or four scales that measureaspectsof cludes scales found to predict later language the relevant behavior. (See the Appendix for (e.g., frequency of behavior regulation and a list of the clusters and scales.) joint attention, Mundy, Kasari, Sigman, & Ruskin, 1995; consonant use, Whitehurst, Measurement Issues Smith, Fischel, Arnold, & Lonigan, 1991), it On the face of it, there could be some prob- alsoincludesscalesthatlackempricalsupport lems with the metric used by the CSBS to as predictors of later language development represent communicative behaviors. First, the (e.g., sociability of communicative functions, McCathren, Yoder, & Warren 37 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. isolated gestures, and episodes of positive af- tionfor6months.Follow-uptestingwasdone fect). Perhaps including scales that are not 6 months after the end of intervention, or 12 predictive with scales that are predictive months after the initial testing. weakens the predictive validity of the Com- In the one-to-one treatment, a trainer munication Composite score. workedwithachild,utilizingmilieulanguage Noresearchhasyetestablishedwhetherthe teaching strategies to teach clear prelinguistic Communication Composite, or any of the communication skills (Warren, Yoder, Gaz- clusters, predicts later language development. dag, Kim, & Jones, 1993). The strategies in- It is possible that measuring communication cluded following the child’s lead, imitating development in various ways and creating an and expanding vocalizations and play, and aggregate score, as the CSBS does, is more modeling and prompting desired behaviors predictive of developmental outcome than fo- (i.e., conventionalgestures,coordinatedatten- cusing on one aspect of the communication tion, vocalizations to the adult). domain. It is also possible thatpotentialprob- The one-to-one treatment contrasted with lems with the CSBS summary measures im- the play group condition which consisted of pede the effectiveness of the clusters or the one adult and three children. In the play Communication Composite as predictors. group, the adult was responsive to children’s The purpose of this study was to determine communication and behavior, followed chil- if the Communication Composite score, or dren’sleads,andcommentedonthechildren’s any of the six cluster scores that contribute to play. The adult, however, did not imitate vo- thecompositescore,arepredictiveoflaterex- calizations or actions, or specifically prompt pressive vocabulary for young children with prelinguistic communication behaviors. developmental delay. Three specific research Because the results of the present investi- questions were addressed. First, is the CSBS gationcouldhavebeeninfluencedbydifferent Communication Composite score positively treatments the children experienced in the correlated with later expressive vocabulary in larger experiment, we tested the interactions toddlers with developmental delay? Second, between group assignment and predictor var- are any of the six clusters (Communicative iablesusedinthepresentstudy.First,wetest- Means-Gesture, Communicative Means-Vo- ed whethertherelationshipbetweentheCom- cal,CommunicativeMeans-Verbal,Social-Af- municationCompositescoresortheindividual fective, Reciprocity, or Communicative Func- cluster scores from the CSBS and later ex- tions) positively correlated with later expres- pressive language was different between sive vocabulary?And,third,doesthedeletion groups. Second, we tested to see iftherewere of non-predictive clusters significantly im- group differences on the Communication prove the correlation between the Communi- Compositescore,theindividualclusterscores, cation Composite and expressive vocabulary? or rate of expressive vocabulary. No differ- ences between the groups on the variables of METHOD interest were found and the relationships of interest were not statistically significantlydif- Participants ferent between groups. Thus, the analyses The participants in this study were part of a usedforthepresentstudytreatthechildrenas longitudinal intervention study being con- one group. ducted by the second and third authors.Inthe Participants were 58 children, 34 boys and larger study, the effects of two models ofpre- 24 girls, 17 to 34 months of age (M (cid:1) 22.6, linguistic communication intervention were SD (cid:1) 4) who were enrolled in community compared. Children were randomly assigned based early intervention programs. Of the 58 to either a one-to-one intervention or a group children, 24 were African American, 31 were intervention. Both treatments were conducted Caucasian,and3wereidentifiedbytheirfam- 4 days a week, for 20 minutes, in the child’s iliesasother.ThechildrenhadBayleyMental school. Children participated in the interven- Development Indices (MDI; Bayley, 1969, 38 JEI, 2000, 23:1 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. 1993) ranging from 35–85 (M (cid:1) 54.3, SD (cid:1) reportedbyStevensandCho.Oursamplewas 13.5). The Bayley does not provide MDIsbe- skewed, however, making the median of 23 low 50, thus for children who scored below (range 10 to 80) a better descriptor of central 50, an estimated MDI was calculated by find- tendency than the mean. Educational level of ingtheregressionequationateachageforthe participants’ head of household varied, 6 had data provided in the Bayley manual and ex- a 7th to 9th grade education, 30 had a 10th to tendingtheregressionline(seeNaglieri,1981 12th grade education, 20 attended college, and for similar application). Children with esti- 2 attended graduate school. mated MDIs below 35 were not included in the study. In addition, each study participant Procedures had to be observed demonstrating at leastone CSBS testing. Project staff were trained to instanceofintentionalcommunicationpriorto give the CSBS. Training consisted of observ- testing. ingtheadministrationoftheCSBSusingboth None of the participants showed evidence the training tapes that accompany the assess- of autism or sensory impairments,andallhad ment instrument, and previously trained pro- the motoric ability to rotate their torsos while jectstaff.Inaddition,projectstaffparticipated engaged in object play. Of the 58 children in in practice sessions conducted with children the study, 4 had Down syndrome, 4 werepre- in the targeted age range who were not study mature births with medical complications, 3 participants. Practice sessions were video- were diagnosed ‘‘failure to thrive,’’ and 2 taped and feedback was given by more ex- were diagnosed with Pervasive Developmen- perienced staff members. All testing sessions tal Disorder. In addition, 6 childrenwereeach also were videotaped and the camera person diagnosed with one of the following condi- doubled as an on-line coach. In addition, all tions: macroencephaly, microencephaly, testing sessions were coded and any discrep- Duane’s syndrome, neonatal meningitis, Fetal ancies in administration were noted, dis- Alcohol Syndrome, and tuberous sclerosis. cussed, and resolved. The remaining 39 had no identifiableetiology The measures for this study were derived or diagnosis other than developmental delay. from two testing sessions, one at the begin- Atthebeginningofthestudy,childrenwere ning of the study and a second session 1 year observed to have fewer than threeproductive, later. Testing was done at each child’s school non-imitative words in their vocabulary.Each site. The Communication Composite section child’s expressive vocabulary was estimated of theCSBStakesabout30minutesorlessto using data from the initial CSBS testing ses- administer. Although the test manual encour- sion and teacher report. During the initial ages the use of parents in the assessment ses- CSBS testing, 7 out of the 58 children used a sion, we used a staff member who was famil- total of five different words (mama, bye-bye, iar with the child, rather than the child’s par- no, baby, and uh-oh). Two of the seven chil- ent. We chose to exclude the parent from the dren spoke two words and the other five chil- procedure because we wanted to reduce be- dren each spoke just one word. The number tween child differences in the extent to which of words and rate of words at the beginning the parent interacted with the child during the of the study were skewed,makingthemedian test session. There is strong evidence that a better statistic for central tendency than the adult-child interaction style influences chil- mean and standard deviation. The median dren’s immediate performance on many pre- number of words was 0 (range 0–2). linguistic and linguistic measures (Lewy & To assign socioeconomic status scores to Dawson, 1992). In addition, activities in the families,weusedtheoccupationalstatusscore CSBS were selected to elicit communication developed by Stevens and Cho (1985). For fromyoungchildrenthroughengagementwith our sample, the occupational status score, interesting objects. The child sat in a safety which is based on occupation,waslowerthan seat at the end of the table facing the video the mean national score of 34.5 (SD (cid:1) 18.0) camera. The adult sat across fromthechildor McCathren, Yoder, & Warren 39 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. on the child’s left so the video camera could at any or all of the books for a 5 minutes clearly capture the faces of both the adult and testing period. The adult is to respond to the child. child’s communications without directing or The Communicative Composite consists of questioning the child. two sections: Communicative Temptations and Sharing Books. The Communication Play Session Temptations are seven situations designed to To measure children’s expressive vocabulary, elicitcommentsandrequestsfromyoungchil- 15minute,one-to-oneplaysessionswithafa- dren.Forexample,inthefirstCommunicative miliar staff member were conducted. All play Temptation the adult winds up a small walk- sessions occurred 1 year after the pre-inter- ing toy and puts it on the table within the vention CSBS testing. Toys used during the child’sreach.Whenitstopswalking,theadult play sessions were typical of toys found in waits for the child to request the toy walk early childhood settings and included: a baby again and then rewinds the toy. If the child doll, 2 baby bottles, a baby spoon, doll hair- does not communicate, the adult implements brush, rattle, blanket, teapot with 2 cups and a prompting procedure to elicit a response. saucers, 4 colored cylindrical sticks, a large First the adult simply asks, ‘‘Need help?’’ If pink car, and a toy telephone. Duringtheplay this does not prompt a child communication, session, the adult could imitate what thechild the adult holds an open palm 12 inches from was doing and comment ontheplay,however the toy and asks again, ‘‘Need help?’’ If the they were to avoid modeling higher levels of child still does not respond, the adult moves play. For example, if the child picked up the his or her open palm closer to the toy (about sticks and started banging on the table the 3 inches) and repeats the question ‘‘Need adult would bang on the table with the other help?’’ If the child continues to not respond, sticks and say, ‘‘We’re banging on thetable.’’ the adult picks up the toy and repeats the ini- Likewise, if the child put the bottle in the tial action of winding the toy and putting it doll’s mouth the adult might say, ‘‘The baby on the table within child’s reach. The se- is hungry. She wants to eat.’’ Each play ses- quence of eliciting a communication request sion was videotaped for later coding. The is repeated up to five times. On the first, sec- child sat in a safety seat across fromtheadult ond,andfifthattempt,theadultcomplieswith or on the adult’s right so the face of both the the request and reactivates the toy, but on the child and adult was visible. third and fourth request, the adult does not immediately comply but rather comments or Outcome Variable: Expressive labels the toy and sets it down without reac- Vocabulary tivating it, and awaits a request. When the Theoutcome measure,expressivevocabulary, child communicates again, the adult activates wasquantifiedasthenumberofdifferentnon- the toy and then puts it away when it winds imitativewordsusedintheplaysession.Non- down. This procedure of complying with the imitative was defined as a word used by the first, second, and fifth request, but not the child that was not in the adult’s previous ut- third and fourth request is repeated in each of terance. Adult pronunciations or approxima- the temptation situations. Materials used for tions of an adult pronunciation were included other temptations are balloons, bubbles, blan- in this measure. Approximation was defined ket for peek-a-boo, blocks and a box, and a as vocalizations having the same number of jar with cereal. In situations designed to elicit syllablesandatleastonephonemeincommon comments, the adult pauses and waits to see withtheadultpronunciation.Inaddition,there if the child comments. If the child does not, must have been nonlinguistic support for de- the adult continues with the test but does not termining the child was saying a word (i.e. provide a specific prompt. child said ‘‘baby’’ while pointing to the doll). In the Sharing Books section, the child is Words with an -ie (horsie) or -y (doggy) and offered 4 books. The child selects and looks words that are commonly shortened byyoung 40 JEI, 2000, 23:1 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. Table 1. Table 2. Reliability Coefficients Means and Standard Deviations for CSBS Variables Variable g-Coefficient Mean SD Communicative Means—Gesture .92 Communicative Means—Vocal .96 Communicative Means—Gesture 7.1 2.6 Communicative Functions .95 Communicative Means—Vocal 11.0 2.1 Reciprocity .92 Communicative Functions 8.6 1.8 Social-AffectiveSignaling .90 Reciprocity 7.9 1.7 Communication Composite .97 Social-AffectiveSignaling 7.6 2.0 Expressive Vocabulary .96 Communication Composite 47.3 7.8 children (e.g., ‘‘sketti’’ for spaghetti and ‘‘nana’’ for banana) also were included, how- cient would be 1. Conversely if the scores of ever,ifachildsaidbothhorseandhorsiethey all the subjects were the same but the coders were credited for only one word. Finally, scored them differently the g-coefficient words had to be in the dictionary or on the would be 0. Mitchell has (1979) suggested an list of symbolic sounds included in the Mac- acceptable range for g-coefficients is .5–.7. Arthur Communication Development Inven- Thereliabilitycoefficientsforallvariablesin- tory/Infants (Fenson et al., 1991). cluded in this study are shown in Table 1. Coding RESULTS Coding of the CSBS session consisted of two separate passes through the data. On the first Preliminary analysis was doneonthedataus- pass, communication acts were identified, as- ingthestepsrecommendedbyTabachnickand signed a communicative function, and rele- Fidell (1989). According to their procedures, vant aspects of each communication act was the outcome variable needed to be trans- noted (e.g. presence of a vocalization or ges- formed. After a natural logarithmic transfor- ture). A communication act was defined as a mation, no statistical assumptions were vio- vocalization or gesture that was directed to- lated for the proposed analyses. For the sake ward the adult (Wetherby & Prizant, 1993). of clarity, means and standard deviations are The second pass through the data wastocode reported using the original scale. the behaviors related to the Social-Affective Expressivevocabularywascalculatedatthe Signaling cluster. end of the study. In the play session all but 11 children used words. The average number Reliability of words used during the 15 minute session Interobserver reliability for all variables was was11.3(range(cid:1)0–87;SD(cid:1)15.5).Because calculated on 22% of the data. Reliability numberof words usedwaspositivelyskewed, samples were randomly selected. Reliability the median number of words used (5.5) rep- was reported using a generalizability or g-co- resents a better descriptor of central tendency efficient (McWilliam & Ware, 1994). Unlike than mean and standard deviation. During the other ways of calculating reliability, g-coeffi- 15minuteplaysession,themediannumberof cients take into account more than one source words used was 5.5. Means and standard de- of variability (Mitchell, 1979). G-coefficients viations of all CSBS variables included in the approach 1 as the variance accounted for by analysis are listed in Table 2. the subjects is large in comparison with the The first question asked if the Communi- variance accounted for by coders (Kasari, cation Composite score was positively corre- Freeman, Mundy, & Sigman, 1995). For ex- lated with later expressive vocabulary. The ample, if there was variance among subjects Communication Composite was a significant and no variance between coders the g-coeffi- predictor of later expressive vocabulary. The McCathren, Yoder, & Warren 41 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. Table 3. Correlations Between CSBS Scores and Later Expressive Vocabulary Correlation p Value Communicative Means—Gesture .355 p (cid:1) .003 Communicative Means—Vocal .387 p (cid:1) .001 Communicative Functions .365 p (cid:1) .002 Reciprocity .340 p (cid:1) .005 Social-AffectiveSignaling .283 p (cid:1) .016 Communication Composite .426 p (cid:1) .000 correlations between CSBS variables and ex- DISCUSSION pressive vocabulary are presented in Table 3. The second question asked which of the This study tested thepredictivevalidityofthe clusters were predictive. We were unable to CSBS Communication Composite when at- test the predictive validity of the Communi- tempting to predict later expressive vocabu- cativeMeans-Verbalclusterbecausetherewas laryforchildrenwithdevelopmentaldelaysin insufficient variance, however, all tested clus- the prelinguistic stage of communication de- ters were positively correlated with later ex- velopment. The Communication Composite pressive vocabulary. Communicative Means- did predict later expressive vocabulary, ac- Vocal, Communicative Means-Gestural, Rec- counting for 18% of the variance.In addition, iprocity, Social-Affective Signaling, and five of the six clusters; Communicative Communicative Functions were all statistical- Means-Vocal, Communicative Means-Ges- ly significant predictors. ture, Reciprocity, Social-Affective Signaling, The final question asked if removing the and Communicative Functions were predic- non-predictive clusters from the CSBS Com- tive. The sixth cluster, Communicative munication Composite significantly improved Means-Verbal, was not tested. Because chil- the correlation. Because all tested clusters dren in our sample were in the prelinguistic were significantly correlated with expressive stage of development, and by definition, not vocabulary, this analysis was not needed. yet talking, we did not expect the Communi- Because MDI and mental age (MA) are of- cation Means-Verbal cluster to be predictive. ten intercorrelated with communication and Thisstudyisimportantbecauseitisthefirst other behavioral skills, we conducted corre- to report predictive validity using the CSBS. lations to ensure our results were not simply It is especially important given that MA and bi-products of age and IQ. Pearson product- MDI, two frequently used predictors of com- moment correlations were used to determine munication, were not related to later expres- the relationship between MDI and later ex- sivevocabularyforthisgroupofchildren.Be- pressive vocabulary, and MA and later ex- cause assessment results determine who re- pressive vocabulary. Neither relationship was ceives services, it is imperative that assess- statistically significant (r (cid:1) .20, p (cid:1) .12; r (cid:1) ment instruments have predictive validity for .07, p (cid:1) .59, respectively). the populations who are likely to be assessed. A post-hoc analysis also was conducted to Thisstudyalsopresentsnewfindingstothe determine if any predictive clusters added field. Before we discuss the results in detail, unique variance when tested in a multiple re- however, we note two caveats. First, before gressionwithexpressivevocabularyasthede- predictive validity can be fullyestablishedfor pendentvariableandthepredictiveclustersas the CSBS further study, which includes other the independent variables.Althoughtheequa- populations of children (e.g., typically devel- tion was statistically significant,F(cid:1)2.9,p(cid:1) oping and children with autism), needs to be .02,noneoftheclustersaddedsignificantlyto conducted.ItmaybethattheCSBShasstron- the prediction. ger predictive validity for some groups of 42 JEI, 2000, 23:1 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. children than for others. Second, although risk’’ are relatively rare. However, using the there were no group differences on the rela- Early Social-Communication Scales (ESCS; tionships of interest, all children included in Seibert & Hogan, 1982), researchers tested this study were part of an intervention study. prelinguistic communication skills to predict Replications should be done with children language and IQ for young childrenwithlow participating in different types of intervention birthweight (Ulvund & Smith, 1996). Both programs. joint attention and behavior regulation were In the introduction we raised three possible found to be predictive of both language and problems related to the measurement strate- IQmeasured1,2,and4yearsaftertheESCS gies used in the CSBS. First, the CSBS as- testing. Unlike our study, the mothers were signs Likert scores to all scales, reducing the included in the testing session for the ESCS. raw score variability between children. In our The magnitude of the significant correlations study, all tested correlations were statistically in the Ulvund and Smith study ranged from significant, thus use of Likert scores did not .19 to .37, while ours ranged from .27 to .43. prevent us from finding the relationships in Thus,itappearsthatourexclusionofthepar- the data. This is an important finding because ent in the testing session may not have had becoming reliable on Likert scores is much a significant impact on the child’s commu- easier,lesstimeconsuming,andthereforeless nication or on the amount of variance ac- expensivethantryingtodevelopandmaintain counted for when predicting expressive vo- reliability on the occurrence of eachbehavior. cabulary. Having the Likert scores makes the CSBS a Many of the constructs included in the more user friendly, cost-efficient instrument CSBS have support in the empirical literature than it might appear, given the complexity of as predictors of expressive language for typi- the behaviors that are being assessed. cally developing children and children with A second potential problem, use of fre- disabilities. Both amount of vocalization and quencies rather than rates in an assessment theuseofconsonantsintheprelinguisticcom- that takes different amounts of time for dif- munication period are predictive of spoken ferent children, was raised. Again, given that language (Camp, Burgess, Morgan, & Zerbe, all the tested correlations were statistically 1987; Kagan, 1971; Menyuk, Liebergott, & significant, this did not turn out to be a prob- Shultz,1986;Murphy,Menyuk,Liebergott,& lem.Becausetheadult’sbehaviorisstructured Shultz, 1983; Roe, 1977; Stoel-Gammon, (i.e., toys were offered a specific number of 1989; Whitehurst, Smith, Fischel, Arnold, & times, adult responses to child requests were Lonigan, 1991). Use of pragmatic functions prescribed), the amount of time the assess- hasbeenshowntopredictlaterexpressivelan- ment takes may be of less importance. Some guagefortypicallydevelopingchildren(Mun- children in our sample took twice as long as dy et al., 1995), children with Down syn- others, however, we still found predictive re- drome (Mundy, Sigman, Kasari, & Yirmiya, lationships. 1988; Smith & vonTetzchner, 1986) and chil- The third potential problem we raised was dren with autism (Mundy, Sigman, & Kasari, the inclusion of scales that may not be valid 1990). Children’s overall rate of communica- predictors of later language development. Al- tion has also been shown to predict later ex- though we did not test the predictive validity pressive vocabulary (McCathren, Yoder, & of each scale, the clusters composed of those Warren, 1999). For some of the other clusters scales did predict expressive vocabulary. If there is not empirical support, but the impor- there were particular scales not predictive by tance of the behaviors is intuitively obvious. themselves,theydidnotimpedethepredictive For example, scales that seem tobeimportant validity of the clusters or of the Communi- in early communication, look at (a) how re- cation Composite as a whole. sponsive the child is to adult communication, Predictive validity studies using groups of (b) positive engagement with a communica- children with disabilities or who may be ‘‘at- tive partner, and (c) how determined thechild McCathren, Yoder, & Warren 43 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution. istogetacrossamessage.Research,however, J. S. (1991). Technical manual for the Mac- Arthur developmental inventories. San Diego: is needed to determine if these behaviors are Singular Publishing Group. predictive of some aspects of later communi- Kagan,J.(1971).Changeandcontinuityininfancy. cation. New York: Wiley and Sons. The CSBS manual provides scaled scores Kasari, C., Freeman, S., Mundy, P., & Sigman, M. and normed scores that can be used to meet D. (1995). Attention regulation by children state criteria for determining eligibility for with Down syndrome: Coordinatedjointatten- services. In addition, because the CSBS en- tion and social referencing looks. American gages young children in interaction, theinfor- Journal on Mental retardation, 100(2), 128– mationgatheredisusefulforidentifyingcom- 136. munication goals and strategies for interven- Lewy, A. L., & Dawson, G. (1992). Social stimu- lationandjointattentioninyoungautisticchil- tion. This is important because many stan- dren. Journal of Abnormal Child Psychology, dardized tests do not provide information that 20(6), 555–566. is useful for program planning. McCathren, R. B., Yoder, P. J., & Warren, S. F. In conclusion, the CSBS is one of a new (1999). Prelinguistic pragmatic functions as generation of prelinguistic assessment instru- predictorsoflaterexpressivevocabulary.Jour- ments that is much better than assessments nal of Early Intervention, 22(3), 205–216. available even 10 years ago. Overall, the McCathren, R. B., Warren, S. F., & Yoder, P. J. CommunicationCompositeandthefivetested (1996).Prelinguisticassessment.InK.Cole,P. clusters of the CSBS were shown to be mod- Dale,&D.Thal(Eds.)Advancesinassessment of communication and language. Baltimore: erately powerful and valid predictors of later Brookes. expressive vocabulary. The CSBS Communi- McWilliam, R. A. & Ware, W. B. (1994). The re- cation Composite has great potential in iden- liabilityofobservationsofyoungchildren’sen- tifying,ininfancy,childrenwhowillbeatrisk gagement: An application of generalizability forlanguageandcommunicationdisorderslat- theory. Journal of Early Intervention, 18(1), er in childhood. 34–47. Menyuk, P., Liebergott, J., & Schultz, M. (1986). REFERENCES Predicting phonological development. In B. Lindblom & R. Zetterstrom (Eds.) Precursors Baker,L.&Cantwell,D.P.(1982).Developmental, of early speech. New York: Stockton Press. social, and behavioral characteristicsofspeech Mitchell,S.K.(1979).Interobserveragreement,re- and language disordered children. Child Psy- liability, and generalizability of data collected chiatryandHumanDevelopment,12,195–206. in observational studies. Psychological Bulle- Bayley, N. (1969). Bayley scales of infant devel- tin, 86(2), 376–390. opment. New York: The Psychological Corp. Mundy, P., Kasari, C., Sigman, M., & Ruskin, E. Bayley, N. (1993). Bayley scales of infant devel- (1995). Nonverbal communication and early opment:Secondedition.SanAntonio:ThePsy- language acquisition in children with Down chological Corp. syndromeandinnormallydevelopingchildren. Beitchman,J.H.,Nair,R.,Clegg,M.,Ferguson,B. Journal of Speech and Hearing Research, 38, &Patel,P.G.(1986).Prevalenceofpsychiatric 157–167. disordersinchildrenwithspeechandlanguage Mundy,P.,Sigman,M.,&Kasari,C.(1990).Alon- disorders.JournaloftheAmericanAcademyof gitudinal study of joint attention and language Child Psychiatry,25, 528–535. development in autistic children. Journal of Bzoch, K. R., & League, R.(1978).Assessinglan- Autism and Developmental Disorders, 20(1), guage skills in infancy: A handbook for the 115–128. multidimensional analysis of emergent lan- Mundy, P., Sigman, M., Kasari, C., & Yirmiya, N. guage. Baltimore: University Park Press. (1988). Nonverbal communication skills in Camp, B., Burgess, D., Morgan, L., & Zerbe, G. Down syndrome children. Child Development, (1987). A longitudinal study of infant vocali- 59, 235–249. zations in the first year. Journal of Pediatric Murphy, R., Menyuk, P., Liebergott, J., & Schultz, Psychology, 12, 321–331. M.(1983)Predictingrateoflexicalacquisition. Fenson, L., Dale, P. S., Reznick, J. S., Thal, D., PaperpresentedatBiennialMeetingofSociety Bates, E., Hartung, J. P., Pethick, S., & Reilly, for Researchin Child Development, Detroit. 44 JEI, 2000, 23:1 Downloaded from http://jei.sagepub.com by M Peterson on May 28, 2008 © 2000 Division for Early Childhood of the Council for Exceptional Children. All rights reserved. Not for commercial use or unauthorized distribution.