Mediciones de lenguaje en niños que aprenden inglés y español simultáneamente.pdf

Embed Size (px)

Citation preview

  • Marchman & Martnez-Sussmann: Validity of Parent Report in English and Spanish 983

    Virginia A. MarchmanThe University of Texas at Dallas

    Carmen Martnez-SussmannAmerican Institutes for Research

    Palo Alto, CA

    The validity of two analogous caregiver/parent report measures of early lan-guage development in young children who are learning both English and Spanishis examined. Caregiver/parent report indices of vocabulary production andgrammar were obtained for 26 children using the MacArthur CommunicativeDevelopment Inventory: Words & Sentences (CDI; Fenson et al., 1994) and theInventario del Desarrollo de Habilidades Comunicativas: Palabras y Enunciados(IDHC; Jackson-Maldonado, Bates, & Thal, 1992). Scores were significantlycorrelated with analogous laboratory measures in both English and Spanish,including a real-object naming task and spontaneous language use during free-play. The findings offer evidence that the CDI and IDHC provide valid assessmentsof early language milestones in young English- and Spanish-speaking children.Factors that may influence the validity of these tools for use with this populationare also discussed.

    KEY WORDS: language assessment, parent report, toddlers, bilingualism,English- and Spanish-language learners

    Concurrent Validity of Caregiver/Parent Report Measures ofLanguage for Children Who AreLearning Both English and Spanish

    Journal of Speech, Language, and Hearing Research Vol. 45 983997 October 2002 American Speech-Language-Hearing Association1092-4388/02/4505-0983

    Over the last 15 years or so, caregiver/parent report has comeinto widespread use as a method for assessing linguisticachievements in young children who are learning English(Fenson et al., 1993; Rescorla, 1989). Following the popularity of theMacArthur Communicative Development Inventories (CDI), similar in-struments have been developed in more than a dozen languages, includ-ing Italian, Hebrew, French, and German (see http://www.sci.sdsu.edu/cdi/ for a complete list). The Spanish-language Inventario del Desarrollode Habilidades Comunicativas: Palabras y Enunciados (IDHC; Jackson-Maldonado et al., 1992) represents one early adaptation of the originalCDI instruments, developed in close collaboration with the original au-thors of the CDIs. In many areas of the United States today, an increas-ing number of children live in environments in which both Spanish andEnglish are spoken. In this study, we examine the concurrent validity ofthe CDI: Words & Sentences and the IDHC: Palabras y Enunciados whenused in conjunction to assess early language milestones in this popula-tion. We compare reported word production and grammar to structuredand spontaneous behavioral measures in a sample of toddlers (23 to 34months old) who are learning both English and Spanish in a large met-ropolitan area of the United States.

  • 984 Journal of Speech, Language, and Hearing Research Vol. 45 983997 October 2002

    For toddlers with simultaneous exposure to Span-ish and English, an adequate language assessment re-quires an examination of lexical and grammatical skillsin both languages. Most typically, assessments involvestructured testing using norm-referenced or standard-ized instruments. Unfortunately, few structured instru-ments have been developed for English-speaking chil-dren in this age range, and even fewer are available forSpanish-speaking toddlers (see Jackson-Maldonado etal., 1993, 2001 for review). Most Spanish versions thatdo exist are direct translations of their English counter-parts and do not incorporate relevant linguistic and cul-tural differences or reflect adequate norming efforts (e.g.,the Sequenced Inventory of Communication Develop-mentSpanish Translation [Hedrick, Prather, & Tobin,1984]). Others are Spanish adaptations with formatsparallel to the English tests, but many of these havebeen criticized in terms of the particular lexical andmorphosyntactic items and the inappropriateness ofcomparing bilingual children to monolingual norms(Gutierrez-Clellen, 1996). Language sampling in a natu-ralistic play situation is another popular technique, butrequires a considerable investment in time and train-ing of personnel. Further, the behavior that is observedcan vary depending on context (e.g., home vs. clinic),interlocutor (e.g., clinician vs. parent), and personalityfactors (e.g., child shyness or talkativeness). Languagesamples are also highly sensitive to the fact that cul-tures differ in the types of behaviors that are viewed asappropriate during play interactions with unfamiliaradults (see Jackson-Maldonado et al., 2001).

    Caregiver/parent report is an increasingly popu-lar technique that has some advantages for use withchildren in this age range. The technique is more cost-effective than behavioral assessments in providing ageneral evaluation of early developing skills. In addi-tion, reports sample behaviors outside the laboratory orclinic and reflect skills across a broad range of contexts(Fenson et al., 1994). Caregiver/parent report may beless sensitive to context or task effects than behavioralmeasures, given that this technique does not requirethat behaviors are observed at a particular place andtime (Bates, Bretherton, & Snyder, 1988). Although limi-tations are well documented (e.g., Feldman et al., 2000;Tomasello & Mervis, 1994), there is growing confidencethat properly constructed and administered caregiver/parent report instruments can provide valid informa-tion regarding a range of early lexical and grammaticalmilestones.

    On both the CDI and IDHC, word production is as-sessed using a parallel checklist format in which wordsare organized in semantically relevant categories (e.g.,animals [animales], vehicles [vehculos]). The IDHC isnot a direct translation of the CDI, but incorporates lexi-cal, grammatical, and cultural differences in Spanish.

    Although there is overlap in the items, each checklist isbased on research on that language and includes wordsthat are linguistically and culturally relevant for theirrespective American or Mexican/Mexican-American popu-lations (Jackson-Maldonado et al., 1993, 2001). Parallelchecklist formats are also used for reporting the use ofword combinations and grammatical morphology. How-ever, the items reflect aspects of grammar that are char-acteristic of documented age-related changes for each lan-guage. Still other sections follow formats that are uniqueto each form, focusing on constructs that are particularlyrelevant to English or Spanish acquisition. For example,the CDI asks about the production of plural and past tenseovergeneralizations (e.g., blockses, goed), in light of theirsignificance for issues relating to the acquisition of mor-phological regularities (e.g., Marchman & Bates, 1994).The IDHC includes a list of frequent and early learnedverbs from each of the three major verb classes (-ar, -er,and ir) and asks reporters to indicate if they have heardthe child produce the forms conjugated in the presentindicative, preterit, and imperative.

    Norming information for the CDI is based on ap-proximately 1200 monolingual English-speaking chil-dren from three geographic areas in the United States(Fenson et al., 1994). Validation studies compared scoresto several standardized and free-speech laboratory mea-sures (Dale, 1991; Dale, Bates, Reznick, & Morriset,1989; Tomblin, Shonrock, & Hardy, 1989). For example,Dale (1991) notes that reported vocabulary was signifi-cantly correlated (r = .74) with number of different wordsproduced in a language sample by 24-month-olds. Simi-larly high correlations were observed between reportedgrammar, Mean of the Three Longest Utterances (M3L)and Complexity, and the behavioral index of MeanLength of Utterance (MLU) (rs = .74 to .76). Normingdata for the IDHC (Jackson-Maldonado et al., 2001) arebased on monolingual Spanish-speaking children in sev-eral areas in Mexico who represent a similar, albeit some-what lower, socioeconomic distribution compared to thatin the norming of the CDI (see also Jackson-Maldonadoet al., 1993). A recent validation study with middle- andupper-middle-class toddlers (Thal, Jackson-Maldonado,& Acosta, 2000) reports significant correlations betweenreported vocabulary and number of objects labeled in anaming task and number of different words producedin a language sample (rs = .56 to .69). Similar levels ofconcordance were observed between MLU and reportedM3L (rs = .68 to .88). These results were comparable tothose reported for English-speaking children with typi-cal language development (Dale, 1991) and older lan-guage-delayed preschoolers (Thal, OHanlon, Clemmons,& Frailin, 1999).

    There is already some evidence that the techniqueof caregiver/parent report is valid with children who arelearning both English and Spanish. For example, using

  • Marchman & Martnez-Sussmann: Validity of Parent Report in English and Spanish 985

    earlier versions of the CDI and IDHC, Pearson and col-leagues (Pearson & Fernndez, 1994; Pearson,Fernndez, & Oller, 1993) showed that lexical develop-ment in Spanish- and English-speaking children gener-ally proceeds at the same rate as that observed in mono-lingual populations. In these studies, productionvocabulary was assessed individually in each language,but also in terms of total conceptual vocabulary. Thistotal score adds together those words that children werereported to produce in one language only (in English, butnot Spanish; in Spanish, but not English), plus wordsthat were reported in both languages (i.e., translationequivalents). Based on the Spanish-English VocabularyChecklist (SEVC), Patterson (1998) found that childrenlearning both English and Spanish generally demon-strated progress that is in line with age-based expecta-tions, including the timing of the 50-word point and theonset of word combinations. Adapted from the LanguageDevelopment Survey (LDS; Rescorla, 1989), the SEVCwas developed specifically for Spanish-English-speakingpopulations and contains two parallel word checklists(in English and Spanish) on which parents indicate iftheir child produces the word in English, Spanish, orboth languages. The SEVC also asks whether or not thechild is producing word combinations. In a validationstudy, Patterson (2000) compared reported vocabularywith number of different words produced during a free-play session in which parents were told to use either orboth languages, in a manner similar to how they talkedwith their child at home (p. 124). Correlations betweenobserved and reported vocabulary were similar to thosein studies with monolinguals (r = .66), and there was astrong concordance between observed and reported useof word combinations.

    These studies indicate that caregiver/parent reportis valid when used with English- and Spanish-speakingpopulations. However, the CDI and IDHC were originallydesigned for use with monolingual learners, and it is notyet known whether these instruments provide the samelevel of validity as those specifically designed with bilin-gual families in mind. The CDI and IDHC are typicallyadministered separately and hence have the advantageof providing an independent picture of each language (asort of best look in both English and Spanish). At thesame time, following Pearson and Fernndez (1994),responses from the checklists can be merged into a com-posite view of a childs lexical progress. In addition, theCDI and IDHC allow detailed evaluations of specificmorphological or syntactic constructions, in addition tonoting whether or not the child is producing multiwordcombinations. Thus, these full-scale instruments pro-vide an integrated view of lexical skill in the languagestaken together, but also assess a range of lexical andgrammatical constructs, including those that are spe-cific to English or Spanish.

    Although caregiver/parent report is thought to pro-vide a more comprehensive view than other techniques,the quality of the report is only as good as the experi-ences of the reporters (Patterson, 2000; Pearson et al.,1993). For example, daycare providers may have a dif-ferent set of experiences upon which to base their re-ports than a mother. Together, a day care provider anda mother may provide a more accurate picture than ei-ther would alone. For children learning two languages,it is often the case that different individuals are thesources of input for the two languages. For example, if achild is exposed to English at daycare but Spanish inthe home, a parent may underestimate their childs useof English given their limited exposure to situations inwhich English is spoken. Other possible sources of biasare also pertinent. Reports may under- or overestimateability when the child is demonstrating different lev-els of skill in the two languages. Further, some report-ers may have limited proficiency in English or Span-ish and, hence, would have difficulty completing a formin that language. On the other hand, if a parent com-pletes both forms because the parent regularly use bothlanguages, that parent may report the production ofwords or combinations in one language when they wereactually produced in the other (Pearson et al., 1995).Finally, factors regarding the educational level of thefamilies or the degree to which they are acculturatedinto the Anglo or Hispanic cultures could also affect theaccuracy of the reports (Arriaga, Fenson, Cronan, &Pethick, 1998).

    The goal of the current study is to evaluate whetherthe CDI and IDHC provide valid estimates of early lexi-cal and grammatical skills in children with regular expo-sure to both English and Spanish. In order to assess thechilds best level of skill in each language, caregiverswere selected to complete the CDI and/or IDHC depend-ing on their working knowledge of English or Spanish, aswell as their familiarity with the childs use of that lan-guage. In some cases, the same individual(s) completedboth the CDI and IDHC; in other cases, differentreporter(s) filled out each form. Laboratory sessions wereconducted separately in English or Spanish with homepartners that had a history of interaction with the childin the respective language. Analogous to previous vali-dation efforts with these instruments, reported measuresof word production are compared to word use during bothfree speech and structured contexts. Lexical measuresare evaluated with respect to each language taken indi-vidually, as well as the degree to which a compositemeasure is indicative of general lexical skill (e.g.,Pearson & Fernndez, 1994). Reported grammar mea-sures are compared to observations of the use of wordcombinations, as well as a measure of their complexity,MLU-words. Finally, we examine whether the validityof these tools is affected by demographic factors and

  • 986 Journal of Speech, Language, and Hearing Research Vol. 45 983997 October 2002

    sources of reporter bias that have particular relevanceto this population.

    MethodParticipants

    Participants were 26 typically developing toddlers(n = 13 girls and 13 boys; M = 27.8 months, range = 23 to34 months) taken from a related study of language acqui-sition in children who are learning both English and Span-ish. Table 1 reports participant characteristics, includingage at test, sex, and aspects of the language-learning en-vironments. Because these children were notexpected to be near ceiling levels in either language(Fenson et al., 1993; Jackson-Maldonado et al., 2001), 8children were slightly older than the recommended agesof the CDI and IDHC for typically developing populations.In order to ensure that children would demonstrate someproduction abilities in each language, children were

    exposed to each language for a minimum of 12 hours perweek through interactions with other people (i.e., not tele-vision). Most children (85%) were exposed to both lan-guages within 3 months of birth. Four children wereintroduced to the second language at the age of 6 (n = 1),12 (n = 2), and 15 (n = 1) months. These children demon-strated vocabulary skills in both languages that were com-parable to those of the other participants. All childrenexperienced both languages for at least 14 months beforetheir participation. No child was exposed to language(s)other than English and Spanish.

    Approximately one quarter (n = 7, 27%) of the chil-dren were first-born, and half (n = 13, 50%) were sec-ond. Children with major birth complications, extendedhospitalizations, developmental disabilities, and/or hear-ing loss were excluded. Three children were reported tohave chronic otitis media (more than 5 episodes), andtwo received intervention (P.E. tubes) at least one yearbefore participation. After participating, one child wasdiagnosed with a speech-language delay and began

    Table 1. Overview of participant characteristics, language learning environments, and reported vocabulary.

    Language used byAge in Home non-parental

    Child Sex months Mother Father language a,c caregiver(s) a English Spanish English e Spanish e Composite f

    1 F 23 Mixed English English Mixed 61.5 38.5 424 116 4432 M 23 Spanish Spanish Mixed Mixed 50.6 49.4 28 22 393 F 24 Mixed Spanish Mixed Spanish 5.0 95.0 116 268 3244 F 24 Spanish English Spanish Spanish 14.8 85.2 98 103 1825 M 24 Spanish English Mixed na 43.1 56.9 105 103 1706 M 25 English English English Mixed 82.9 17.1 192 19 1957 F 25 English Spanish Mixed English 59.9 40.1 164 84 2118 F 25 Spanish Spanish Spanish na 46.0 54.0 46 47 789 M 26 Spanish Spanish Spanish na 21.9 78.1 37 192 211

    10 M 26 English English English Mixed 85.1 14.9 236 76 26211 F 26 English Spanish English English 90.3 9.7 324 6 32312 M 26 Spanish Spanish Mixed English 42.8 57.2 24 30 4713 F 27 Spanish Spanish Spanish na 24.3 75.7 33 120 14614 M 29 Spanish Spanish Mixed na 65.7 34.3 242 81 26015 F 29 Spanish Spanish Spanish English 11.8 88.2 77 52 11316 F 30 Mixed Mixed Spanish English 11.3 88.7 255 629 66317 M 30 Spanish English English na 90.0 10.0 427 4 41618 F 30 Mixed Spanish English Mixed 69.2 30.8 368 119 38519 M 31 English Spanish Mixed Spanish 49.3 50.7 205 31 21920 M 31 English English Mixed Spanish 68.1 31.9 629 242 65621 F 31 Spanish Spanish Spanish English 41.7 58.3 176 262 35522 M 31 Spanish Spanish Mixed na 33.1 66.9 291 442 52623 F 31 Spanish Spanish Spanish na 35.8 64.2 129 232 29324 F 32 Spanish Spanish Mixed English 48.0 52.0 328 404 49225 M 32 Spanish Spanish Mixed Spanish 37.6 62.4 356 165 40726 M 34 English English English Spanish 80.5 19.5 422 97 444

    Note. na = not applicable.a Based on bilingual background interview. b What is your/your partners native language? c What is the primary language spoken in your home?d Proportion of total talk to child. e Total words reported on CDI and IDHC. f Total concepts reported in English Only + Spanish Only + Both English andSpanish (see text).

    Native language a,b Reported input to child a,d Reported vocabulary

  • Marchman & Martnez-Sussmann: Validity of Parent Report in English and Spanish 987

    intervention services at a local clinic. All of the childrenlived with both parents, and eight families had addi-tional adults living in the home (e.g., grandmother).Eighteen of the children experienced some regularchildcare either by a nonparent caregiver (n = 11) or ata daycare center (n = 7).

    The participants represented a range of socioeco-nomic backgrounds, as indexed by maternal years ofeducation (e.g., graduation from high school = 12 years).On average, mothers reported at least some college at-tendance (M = 14.4, range = 6 to 18). All of the motherswho were employed outside the home (n = 12, 46%) hadcompleted at least a high school education, and 4 hadcompleted a higher degree. Of the mothers who reportedtheir occupation as homemaker (n = 12, 46%), 5 had highschool degrees or less.

    The majority of the children were born in the UnitedStates (n = 25, 92%), although 65% (n = 17) of the fathersand 54% (n = 14) of the mothers were born in other coun-triesprimarily Mexico. Other families were from Cen-tral America (n = 4), South America (n = 1), or PuertoRico (n = 1). The majority of mothers (n = 15, 57.7%) andfathers (n = 17, 65.4%) were native Spanish speakers,but 42% of the families reported that the home languagewas a mix of both English and Spanish. The ethnicity ofmost parents was reported to be Hispanic (n = 15, 58%),with 27% of the sample reporting mixed ethnicities(Anglo-Hispanic: n = 6, African AmericanHispanic: n =1). In four cases, both parents reported Anglo-Americanheritage, with the child experiencing Spanish-languageinput from a nonparental caregiver. Whenever possible,mothers completed a modified version of the Accultura-tion Rating Scale for Mexican AmericansII (ARSMA-II; Cuellar, Arnold, & Maldonado, 1995). Because somefamilies were from Central and South America, the wordHispanic was substituted for Mexican on the question-naire. Six of the mothers were reported as Very His-panic Oriented, and 7 were Very Assimilated (StronglyAnglo Oriented). The remaining mothers (n = 12, 52%)were balanced in cultural orientation, suggesting iden-tification with both the Hispanic and Anglo cultures.

    ProcedureParticipant Recruitment

    Families were recruited on a volunteer basis throughlocal organizations, community events, and Spanish-language radio and television stations. Bilingual re-search assistants (RAs) contacted interested families viatelephone. During this interview, individuals (e.g., par-ents, nannies) were identified as appropriate to partici-pate on the basis of their consistency of contact with thechild over the previous 6 months, their willingness tocome to the university, and their ability to speak andread English or Spanish.

    Language-Learning EnvironmentOne or both parents participated in a Bilingual

    Background Interview (approximately 2030 min) ad-ministered by a bilingual RA in the language most com-fortable for the parent(s). Developed specifically for thisand related studies (Marchman, Martnez-Sussmann,& Price, 2000), this interview was designed to provide acomprehensive picture of the English and Spanish in-put for each child. Parents indicated the preferredlanguage(s) of the home, as well as the language(s) theyused with each other and with the child (English, Span-ish, or both). Parents then described the childs dailyschedule, including wake-up time, nap time, and bed-time, separately for weekdays and weekends. For eachperson with regular contact with the child, both in andoutside the home, parents were probed regarding howmany hours the child spent with that person per weekand what language(s) were spoken to the child (English,Spanish, or both). When both languages were reported,parents estimated the relative proportion of English andSpanish used. Additional questions probed other rel-evant factors (e.g., frequency of trips outside the UnitedStates). In order to estimate the amount of input in eachlanguage, the number of hours for each language wassummed across all sources (mother, father, grandpar-ents, siblings [over 3 years of age], relatives, nonparentcaregivers, adult friends, and child friends [over 3 yearsold]). This estimate was thus affected by the number ofindividuals with whom the child was in regular contact.In order to compare across individuals, proportion scoreswere computed as the amount of Spanish, English, ormixed input divided by total hours of talk. Mixed inputwas added to the appropriate totals on the basis of re-ported relative proportions of English to Spanish in thatcontext. Results indicated that the relative proportionof English to Spanish was balanced on average (M =45% English, 55% Spanish), although a given childsenvironment ranged from 5% English (95% Spanish) to90% English (10% Spanish).

    Caregiver/Parent ReportOne or more adults who were familiar with the

    childs use of that language completed the CDI: Words& Sentences and/or the IDHC: Palabras y Enunciados.The forms were mailed to the families and returned atthe laboratory visits. During the contact phone call, theforms were described and all instructions were reviewedverbally. Reporters were specifically told to report spon-taneous productions on the forms rather than elicitedrepetitions or imitations. General instructions were alsosummarized in a separate cover letter in both Englishand Spanish (e.g., Words do not have to be pronouncedperfectly in order for the child to get credit for under-stands & says; Marque las palabras pronunciadas de

  • 988 Journal of Speech, Language, and Hearing Research Vol. 45 983997 October 2002

    diferente manera [ej. pato en vez de zapato]). Care wastaken to remind reporters that dialectical variants orother words could be substituted for words on the form.The printed instructions on the questionnaires weremarked with a highlighter so that participants wouldbe more likely to notice and read them. A preprintedlabel was added to the front page of each form instruct-ing reporters to indicate who filled out the form (Pleasemark with an X to indicate who helped fill out this form:Mother, Father, Others [please indicate]; Por favormarque con una X quienes ayudaron a llenar esta forma:mam, pap, otros [por favor, especifique]). In 10 cases(38.5%), more than one person filled out one or bothforms. In more than half of the cases (n = 15, 57.7%),different individual(s) completed the CDI and IDHC.Mothers did not participate in filling out the forms for 7of the IDHCs and 4 of the CDIs. Upon receipt, the formswere examined following general guidelines to ensuretheir completeness (i.e., no blank pages). Apparent in-consistencies were noted and clarified or corrected overthe phone as soon as possible after receipt.

    On the Vocabulary Checklist of both the CDI andIDHC, total vocabulary size is computed as the totalnumber of words (out of 680) reported as understandsand says (comprende y dice). These checklists cap-ture whether or not a child knows a particular word inEnglish and/or Spanish, without regard to the contextin which it is produced or the way it is pronounced. To-tal scores provide general indications of the size of ex-pressive vocabularies in English and Spanish. In addi-tion to the CDI and IDHC taken individually, a compositevocabulary score was also computed. Using Pearson(1992) as a basis, each item on the CDI and IDHC wasmatched if it reflected the same general child-based con-cept (e.g., dog = perro). Each match was evaluated by ateam of native Spanish-speaking RAs in conjunction withthe first author and other CDI Advisory Board mem-bers. Because the CDI and IDHC are not direct transla-tions, a single concept may be represented by more thanone item on the CDI but only one item on the IDHC, andvice versa. For example, the clock/watch concept is rep-resented by two items (clock and watch) on the CDI, butbecause there is no lexical distinction between clock andwatch in Spanish, this concept is represented by onlyone item (reloj) on the IDHC. In other cases, a given con-cept is represented in only one of the languages (e.g.,the tortilla concept is matched to an item on the IDHC,but not the CDI). Finally, a single concept may bematched to more than one item on each form (e.g., in-side/in and adentro/en). A total of 804 concepts werederived, representing the number of ideas indexed bythe CDI and IDHC taken together. (A list of the matchesis available from http://www.utdallas.edu/~vamarch.)

    Using an automated scoring program (Marchman,1999), total conceptual vocabulary was calculated as the

    sum of the number of concepts reported in English only(e.g., the child is reported to produce dog but not perro),Spanish only (e.g., the child is reported to produce gatobut not cat), and both English and Spanish (e.g., thechild is reported to produce both shoe and zapato). Thisscore is not the sum of the CDI and IDHC totals; a childreceives credit for only one concept when equivalentwords are reported in both languages. The score yieldsa general index of lexical production, in terms of thenumber of concepts expressed, regardless of the languageused to express them.

    On both the CDI and IDHC, progress in grammaris measured in two ways. First, reporters provide ex-amples of three longest utterances that you have heardyour child say recently (Por favor, escriba tres ejemplosde las frases ms largas que recuerde que su hijo hayadicho ltimamente). Length of utterance is typicallycomputed in morphemes for English child language data;however, length in words is more common in studies ofRomance languages because of the difficulties of estab-lishing criteria for counting morphemes in morphologi-cally rich languages (e.g., Gutierrez-Clellen, 1996).Therefore, M3L for both languages was computed inwords, rather than morphemes, following standard con-ventions. Children who were reported to not yet com-bine words were assigned an M3L score of 1.0.

    A second index of grammar was derived from re-sponses on two analogous recognition format Grammati-cal Complexity sections. Reporters indicate which of apair of phrases sounds more like what your child is say-ing right now (la que ms se parezca a la forma comohabla su hijo, en este momento). The first option is anexample utterance that lacks grammatical markers oris syntactically simple, whereas the second provides amore grammatically complex alternative (e.g., Kitty sleepvs. Kitty sleeping; Paloma llorando vs. Paloma estllorando). A childs complexity score is the number oftimes the second (i.e., more complex) of the example sen-tences was chosen (37 maximum). Recall that the itemsreflect grammatical constructs that are particularly rel-evant for English or Spanish acquisition. (See Fensonet al., 1993 and Jackson-Maldonado et al., 2001 for moredetail regarding the specific items.)

    Laboratory SessionsLanguage assessments were conducted using two

    parallel behavioral protocols, one in English and onein Spanish, with a bilingual RA and an adult who wascomfortable conversing in the language-of-the-day.Participants were informed of the language-of-the-dayat the time of scheduling and were reminded whengreeted in the parking lot (Remember, today we will bespeaking English only, Hoy vamos a hablar solamenteen espaol). Sessions typically lasted less than one hour

  • Marchman & Martnez-Sussmann: Validity of Parent Report in English and Spanish 989

    and were generally conducted 1 to 2 weeks apart. In allcases, reporters updated the CDI and IDHC at the re-spective sessions. Session order was counterbalanced.

    At least 30 minutes of free-play was conducted withthe adult (15 minutes) and RA (15 minutes). RAs werefully proficient in English and Spanish and were spe-cifically trained to be sensitive to cultural variations inchild-adult interaction and parenting styles. RAs spoketo the child and family members only in the language-of-the-day. Before beginning the session, RAs talked withthe child and adult(s) for several minutes to establishthe context for an informal, family-like interaction (e.g.,use of familiar register). Following Thal et al. (2000),two parallel sets of toys provided a range of symbolicand social-interactive activities. RAs assured the adultsthat the child should feel comfortable playing with thetoys as if they were their own. A different RA partici-pated in each session. Sessions were audio- and video-taped and transcribed later.

    Object naming in a structured setting was assessedusing a task adapted from Thal et al. (2000). Seated ata table, the RA pulled out real objects from a bag, oneat a time, showed the object (Look!, Mira!) and thenhanded it to the child. If the child spontaneously namedthe object, the RA took back the object and pulled outanother toy. If a label was not spontaneously providedafter about 20 seconds, the RA asked the child to do so(Whats this?, Qu es esto?). If the child named theobject in the non-language-of-the-day (e.g., Whatsthis? Avin!), the RA prompted with Yes, but is thereanother word for this? After two prompts, the RA

    moved on to the next item. Three practice items wereadministered.

    A total of 15 common objects were used, eight ofwhich were taken directly from Thal et al. (2000). Two(of the original 10) objects used by Thal et al. (2000)were eliminated because their common names had pho-nological overlap in English and Spanish (car/carroand flower/flor). In these cases, it would be difficultto determine whether a label produced by the childmatched the language-of-the-day. Seven new objectswere selected on the basis of the likelihood that theywere familiar to children in this age range and on thephonological distinctiveness of the English and Span-ish labels. All objects had everyday names listed on theCDI and IDHC. As shown in Table 2, all English labelswere reported to be produced by at least 50% of 24-month-olds in the CDI norming study (Dale & Fenson,1996). Item frequencies for the IDHC were not avail-able when the task was developed, but are provided inTable 2 for comparison (Jackson-Maldonado et al., 2001).Some target names have comparable frequencies inEnglish and Spanish, but others were less likely to bereported for Spanish- than English-speaking toddlers(e.g., hat vs. sombrero).

    Transcription and Data ReductionLanguage samples were transcribed using standard

    protocols by trained bilingual RAs (MacWhinney &Snow, 1985). Transcripts contained approximately 100fully intelligible utterances (or as many utterances as

    Table 2. Proportion of children reported to produce target items based on the norming of the CDI andIDHC.

    English CDI a Spanish IDHC b

    Item 24 months 28 months Item 24 months 28 months

    Airplane 78 93 Avin 71 77Balloon 86 95 Globo 77 80Bed 78 95 Cama 71 87Book 90 97 Libro 41 67Comb 50 74 Peine 52 67Cup 79 93 Taza 49 66Dog 92 98 Perro 85 86Doll 75 88 Mueca 63 75Duck 88 98 Pato 68 73Egg 66 81 Huevo 81 85Frog 64 85 Rana 32 50Hat 78 93 Sombrero 32 51Horse 82 93 Caballo 73 80Shoe 91 97 Zapato 92 88Spoon 77 91 Cuchara 73 80

    M 78.3 91.4 M 64.0 74.1

    aAs reported in the LEX database (Dale & Fenson, 1996). bBased on Jackson-Maldonado et al. (2001).

  • 990 Journal of Speech, Language, and Hearing Research Vol. 45 983997 October 2002

    the child produced) based on a minimum of 24 min offree-play (12 min with RA, 12 min with adult). The totalnumber of child utterances ranged from 35 to 271 in theEnglish (M = 124.6, SD = 25.0) and from 36 to 250 inthe Spanish (M = 115.6, SD = 49.8) sessions. After tran-scription, all child utterances were evaluated as English,Spanish, or a mix of both languages. The total numberof single-language child utterances ranged from 35 to228 for the English (M = 107.0, SD = 53.9) and from 3 to203 for the Spanish (M = 82.9, SD = 49.5) sessions. Atotal of 102 additional child utterances were identifiedas mixed (e.g., Mama I need agua, Hay nios in here).Mixed utterances were observed in 13 English and 17Spanish sessions, with at least one mixed utterance pro-duced by 20 (of the 26) children.

    Four children had too few utterances in the lan-guage-of-the-day (3, 14, 23, and 31 in the Spanish ses-sion; 35 in the English session) to allow stable estimatesof their production abilities. Thus, language sampleanalyses were based on 22 Spanish-language sessions(M = 98.1, SD = 44.2, 46 to 205 total utterances) and 25English-language sessions (M = 112.1, SD = 54.5, 45 to228 total utterances). Lexical and grammatical indicesin the language-of-the-day were derived using the ChildLanguage Analysis System (MacWhinney & Snow, 1990):(a) Number of Different Words (NDW), (b) Total Num-ber of Words (TW), and (c) Mean Length of Utterance(MLU-words). Mixed utterances were excluded from thecomputation of MLU-words. However, identifiable wordsin the language-of-the-day in mixed utterances wereincluded in NDW and TW.

    Reliability estimates for transcription were obtainedby comparing the independent transcriptions for a ran-domly selected subsample of the English (5 of 25) andSpanish (5 of 22) sessions. Reliability was computed asthe number of matches out of the total number of oppor-tunities for agreement per transcript. Before resolvingdiscrepancies, reliability coefficients were 88.9% for theEnglish session (range = 74.2% to 100%) and 97.6% forthe Spanish session (range = 94.9% to 100%). The RAsreviewed the transcripts and videorecordings and soughtto obtain a consensus for all discrepancies. A third coderresolved any remaining disagreements.

    All responses on the naming task were transcribedindependently by two RAs from the audio- and video-recordings. A child was credited for naming the object inthe language-of-the-day if his or her response approxi-mated the target name in English or Spanish at the ap-propriate session. Names in the nontarget language werealso noted. Only spontaneous or prompted responseswere coded; credit was not given for imitations. Syn-onyms or equivalent names (e.g., mona for mueca) wereacceptable. In some cases, parents/caregivers confirmedthat the label produced was used by the child to refer to

    that object in other contexts. Sound effects were ac-cepted only if produced in a naming context (e.g., look-ing at a dog while saying thats woof-woof). Reliabilityestimates were derived by comparing each independentlyprepared transcription and coding sheet, yielding 363agreements (of 390 possible) for the Spanish session and360 agreements (of 390 possible) for the English sessions.The average percent agreement across session was 93%(range = 73% to 100%). The two RAs reviewed each dis-agreement together and obtained consensus wheneverpossible. All remaining disagreements were resolved bythe first author. Two naming scores were obtained: (1)total objects named in the language-of-the-day, and (2)total objects named (regardless of language).

    ResultsCaregiver/Parent Report andLaboratory Measures of Vocabulary

    Table 1 presents reported English, Spanish, andComposite vocabulary scores for each child, and Table 3presents descriptive statistics (mean, standard devia-tion, and range) for the same measures. All children werereported to produce at least a few words in both Englishand Spanish, although there was considerable individualvariation that is typical of children at this age. Themajority of the children (n = 16, 61%) had at least 50reported words in each language, and only threeyounger children (23, 25, and 26 months) had fewerthan 50 reported words in both English and Spanish.Composite scores indicated that only two children werereported to express fewer than 50 concepts when bothlanguages were taken into account. (One of these chil-dren began to receive services from a speech-languagepathologist at 27 months old.) Thus, the majority ofthe children were demonstrating progress in vocabu-lary that was generally appropriate for their age. Fur-ther, size of reported vocabularies was significantlycorrelated with the proportion of English (r = .56, p