English BESA Morphosyntax Performance Among Spanish–English Bilinguals Who Use African American English

Abstract

Purpose

This study examined the use of African American English (AAE) among a group of young Latinx bilingual children and the accuracy of the English Morphosyntax subtest of the Bilingual English–Spanish Assessment (BESA) in classifying these children with and without developmental language disorder (DLD).

Method

Children (N = 81) between the ages of 4;0 and 7;1 (years;months) completed a narrative task and the BESA Morphosyntax subtest. We identified DLD based on four reference measures. We compared specific dialectal features used by children with DLD and their typically developing peers. We also conducted an overall analysis of the BESA subtest and subsequent item-level analyses to determine if particular items were more likely to contribute to the correct classification of the participants.

Results

Children with DLD used three AAE forms in their narrative samples (subject–verb agreement, zero copula/auxiliary, or zero past tense) more frequently than their typically developing peers. Area-under-the-curve estimates for the cloze, sentence repetition, and composite scores of the BESA indicated that the assessment identified children with DLD in the sample with good sensitivity. Item analysis indicated that the majority of items (84%) significantly differentiated typically developing children and children with DLD.

Conclusions

The BESA English Morphosyntax subtest appears to be a valid tool for the identification of DLD in children exposed to AAE and Spanish. We provide practical implications and suggestions for future research addressing the identification of DLD among children from culturally and linguistically diverse backgrounds.

The intersections of race, ethnicity, health, education, and language are at a pivotal moment in the history of our country. Stakeholders acknowledge the need to be better prepared to provide services that are culturally and linguistically responsive as U.S. children become increasingly diverse (Gregory, 2020). Nearly a quarter (22.5%) of school-age students across the country today speak a language other than English at home (Federal Interagency Forum on Child and Family Statistics, 2019). Across public schools in the United States, 15.9% of kindergarten students are English learners, and Spanish is the number one home language of preschool and K–12 school enrollees learning English as a second language (Administration for Children and Families, 2020; Hussar et al., 2020).

Children learning English have exposure to different varieties or dialects of English (e.g., mainstream American English [MAE] and African American English [AAE]), depending on the neighborhoods and communities in which they live and visit and/or the school that they attend. For example, Latinx and Black/African American children generally attend schools that have a low percentage of White students and, from 1993 to 2015, were progressively more likely to attend school together (Richards et al., 2020). Languages and dialects are constantly evolving and borrowing from one another, and young children may be especially susceptible to subtle and obvious influences on language variation (e.g., Tagliamonte & Molfenter, 2007). When MAE, ¹ which is the variety of English used by government, business, education, and science (American Speech-Language-Hearing Association, 1983), is used as the norm, the use of dialects that are considered to be “nonmainstream” (e.g., AAE and Spanish-influenced English) may affect outcomes on tests of speech and language ability, potentially resulting in the misclassification (i.e., over- or underdiagnosis) of developmental language disorder (DLD; e.g., Oetting et al., 2013; Sullivan & Bal, 2013). Grammatical features that are characteristic of DLD for one group of speakers may be representative of expected normal variation for other groups of speakers. An integral part of the speech-language pathologist's job is to determine disorder “within” linguistic diversity (Oetting, 2018). For this study, we explored AAE forms in the spontaneous language of young bilingual Latinx children with and without DLD. In addition, we investigated the impact of AAE forms on the accuracy of the English Morphosyntax subtest of the Bilingual English–Spanish Assessment (BESA; Peña et al., 2018) in classifying bilingual (English/Spanish) speakers with and without DLD. The BESA is a language assessment that was designed to assess speech and language ability among Spanish–English bilingual children who are exposed to a variety of English and Spanish dialects.

African American English

AAE is the most widely studied (Rickford & Rickford, 2013) and, perhaps, the most well-known racial or ethnic variety of American English. It is a rule-based, complex dialect of English that is associated with individuals in the Black/African American community in all regions of the United States and across socioeconomic status levels (Green, 2002). It is important to note, however, that not all Black or African American individuals speak AAE in all contexts. Furthermore, like languages, dialects are sociocultural phenomena; one does not necessarily have to identify as Black or African American to understand and use AAE forms. Prerequisites for comprehension and use of languages and dialects include exposure and opportunities for interaction with others who use the language varieties (e.g., Hoff, 2006; Wolfram & Schilling, 2016). Although AAE is generally considered a nonmainstream dialect, there are multiple forms of AAE (e.g., Gullah and Southern AAE), and all of those forms have significant overlap with MAE. For a more extensive discussion on AAE, including some researchers' preference for using the term African American Language as opposed to AAE, see Lanehart and Malik (2015).

AAE is characterized by a specific set of phonological and prosodic patterns, morphological and syntactic rules, and vocabulary (Wolfram & Schilling, 2016). The study of AAE has largely focused on variation in pronunciation (phonological) and word- and sentence-level (morphosyntactic—a combination of morphological and syntactic) features. Some examples of phonological forms of AAE include variable production of the final consonant (“four” pronounced as /fɔ/) and reduction of consonant clusters (“hand” pronounced as /hæn/). Among others, morphosyntactic features of AAE include the invariant/habitual BE to signify reoccurring activities (“She be here on Thursdays”), variable subject–verb agreement (“They were here” as “They was here”), and zero marking of the copula and/or auxiliary (i.e., zero copula/auxiliary) in sentences where they would be overtly marked in MAE. For example, the phrase “She is nice” may be stated as “She nice”, and the sentence “They have only been there a few times” might be stated as “They only been there a few times” in AAE. When compared to phonological features, the majority of research on AAE among children has focused on morphosyntactic features because of concerns that general developmental speech patterns might be confounded with phonological features of AAE (Washington et al., 2013).

Some researchers argue that certain AAE forms should be referred to as combination phonological–morphosyntactic features because a sound may be reduced or not fully pronounced, resulting in the production of a morphosyntactic feature (e.g., Craig et al., 2003). For instance, the final consonant at the end of “Jessica's” in “We went to Jessica house” is not pronounced (phonological feature), which results in a zero possessive marker (morphosyntactic feature). Similarly, the zero plural marker (morphosyntactic) is produced when the final /s/ is not fully pronounced (phonological) in words that signify more than one in quantity (“I have fifty cent”) or when the final /t/ sound is not fully pronounced (“He jump back and fell down”), resulting in the zero past tense marker (morphosyntactic). For a more comprehensive list and further discussion of morphosyntactic forms thought to be influenced by phonology, see Labov (1969). For this study, we investigated a subset of morphosyntactic and combination features that are commonly produced among preschool and kindergarten children in previous research studies (e.g., Craig & Washington, 2004; Green, 2010; Oetting & McDonald, 2001; see Table 1). We chose to focus on these particular forms based on previous research studies in which these were the most commonly and frequently occurring forms in language samples among children around the same age as our sample.

Table 1.

African American English (AAE) features coded in narrative language samples.

Feature	Definition(s)	Code	Example(s)	Observed on BESA
Zero past	▪ Past tense marked with zero-marked verb	[ZPT]	(then) the dog say|said[ZPT] bow_wow.	Yes
				Cloze regular past
Subject–verb agreement	▪ Use of zero-marked verb singular or plural subjects	[SVA]	They was|were[SVA] walking.	Yes
			And then he look/*3s[SVA]	Cloze regular past or third-person singular
Zero copula/auxiliary	▪ Stative (or copular) and auxiliary constructions are zero marked	[COP]	they *auxbe[COP] leaving	Yes
				Copula or progressive (aux + progressive)
Double mark past	▪ Irregular past tense verb form marked with overt past tense –ed form	[DMK]	(then um) it broked[DMK]	Yes
				Opportunity in sentence repetition (there is one irregular target—had)
Overregularization	▪ Irregular past tense form realized as zero-marked stem with overt past tense –ed form	[ORG]	(and) the frog come/ed[ORG] out	Yes
				Opportunity in sentence repetition (there is one irregular target—had)
Zero possessive	▪ Possessive marked by possessor + possessed item	[POS]	you see the boy/*z[POS] feet.	Yes
				Cloze possessive
Zero plural	▪ Plurality marked by number marker + object	[ZPL]	(and) all the bee/*s[ZPL] went.	Yes
				Cloze plural
Multiple negation	▪ The use of two or more negatives in one sentence for emphasis or with preposed negative (for negative concord)	[NEG]	(and) they said I don't want to see you no[NEG] more	No
			They don't see nothing	Cloze in negative + DO insertion
Invariant/habitual BE	▪ BE form used to mark event taking place over time	[IBE]	they be[IBE] walking	Yes
				Maybe in progressive items

Note. AAE features derived from Craig and Washington (2004), Green (2010), and Oetting and McDonald (2001). A tenth code, [SVA][ZPT], was added and used when bare form use was equivocal. BESA = Bilingual English–Spanish Assessment.

When it comes to research on children, studies have devoted much attention to nonmainstream dialects in general and AAE specifically, particularly in the area of speech and language development. Several years of research have been dedicated to understanding differences between language variation (usually referred to as difference) and language disorder (e.g., Seymour et al., 1998), and more recently, scholars have begun to develop guidelines for the classification of language disorders “within” language differences (e.g., Bedore et al., 2018; Oetting, 2018). This body of research has presented information resulting in a better understanding of the nature of dialect use among children and, importantly, has provided recommendations for practitioners who work with linguistically diverse students (e.g., American Speech-Language-Hearing Association, 2003; Oetting, 2018).

Dialect use varies among children. For instance, younger African American children (preschool and kindergarten) tend to speak with a greater frequency of AAE dialect features than children in upper elementary grades (e.g., Craig et al., 2004). When AAE dialect is measured among the same children using different methods, children, as young as preschool age, use forms of AAE in different frequencies across various contexts such as oral narratives, picture description tasks, and sentence imitation tasks (e.g., Connor & Craig, 2006; Craig et al., 2014). In other words, nonmainstream dialect density, or the proportion of nonmainstream dialect features used relative to overall output in a given language sample, varies both across and within children, depending on the language task given. Language tasks that ask a child to tell a story are often used to obtain a more natural or conversational register from children. For that reason, several researchers use oral narratives to elicit and examine AAE use among young children.

Bilingual Speakers of AAE

Spanish–English speakers living in urban areas of the East Coast of the United States have significant exposure to nonmainstream English and often demonstrate features of AAE (Labov, 1968; Labov & Harris, 1986; Poplack, 1978). Wolfram (1974) noted that children from the Northeast, specifically Puerto Rican children, might use dialect features that are consistent with their AAE-speaking peers. Some features are influenced by Spanish (e.g., variable production of the final consonant) and also overlap with features of AAE (Wolfram & Schilling, 2016; Zentella, 1997). Overlapping grammatical features between Puerto Rican English and AAE are zero copula and zero marking of third-person singular –s (variation in subject–verb agreement; Wolfram & Schilling, 2016). Similar to other AAE speakers, dialect density among bilinguals varies across speakers (Zentella, 1988). As an example, bilingual speakers with regular contact with AAE speakers demonstrate the invariant/habitual BE in their spoken language, while those with little or no contact with AAE speakers generally do not use this feature (Wolford & Evanini, 2006).

While it is well documented that Latinx individuals in some communities in the Northeast demonstrate features of AAE in their spoken language, recent work (Carter, 2013; Dunstan, 2010; Hallett, 2015) suggests that these phenomena are also evident in other geographical areas in which both Latinx and Black/African American children reside. Dunstan (2010) examined the use of three morphosyntactic features of AAE (invariant/habitual BE, zero copula, and zero marking of third-person singular –s) among 65 Latinx children in urban and rural communities in the southern part of the United States. Consistent with the findings of Wolford and Evanini (2006), Dunstan found significant variability in the frequency of the use of AAE features among the participants. It should be noted that two of the features examined (zero copula and zero marking of third-person singular –s) can also be associated with Spanish-influenced English and the frequency of their use was influenced by the participants' English proficiency. Zero marking of third-person singular –s could be a result of a phonological simplification (consonant cluster reduction) since Spanish does not permit clusters in word-final position, and in the radical dialects of Spanish (e.g., Caribbean dialects; Guitart, 1978), syllable-final coda is deleted or weakened. Research on Spanish acquisition documents early and accurate use of the copula verb by children (López Ornat et al., 1994; Sera, 1992). However, the work of Thompson (1991) suggests that Spanish speakers are more likely to delete the copula “ser” rather than the copula “estar” when speaking to a non–Spanish–speaking individual. It is unknown whether this process might extend to a second language, resulting in the deletion of the copula in English sentences. Liceras et al. (2012) reported that two bilingual English–Spanish children who were followed from ages 2;0 to 4;11 (years;months) and were not reported to be exposed to nonmainstream English had low rates of English copula omission (6%–9% of the opportunities).

Collectively, the existing data suggest that features associated with AAE are evident in Latinx individuals and that the frequency that the features are used is highly variable and likely influenced by contact with AAE speakers and proficiency in English. However, current research on nonmainstream dialect among bilingual children is limited by the fact that most studies have focused on older children or adolescents and have examined only a very small number of AAE features (i.e., two to three) within those children's language samples. Although it is common knowledge in the discipline of communication sciences that language differences due to a child's status as a nonmainstream dialect speaker or as a bilingual English learner do not constitute DLDs, little research has addressed the nature of nonmainstream dialect use among English learners, particularly at the age group in which speech and language disorders are often identified (Oetting, 2018). As our schools become increasingly diverse, it is important, perhaps now more than ever, to examine young children's language profiles to improve the accuracy of the identification of DLD among linguistically diverse students.

BESA

The BESA (Peña et al., 2018) was developed specifically for the purpose of identifying phonological and language disorders among Spanish–English bilingual children. The test consists of three standardized subtests in Spanish and English: Morphosyntax, Semantics, and Phonology. In a study of an early version of the BESA English Morphosyntax subtest, the measure accurately identified language impairment and typical language development in 56 English-only proficient Latinx children and 20 bilingual Latinx children with stronger skills in English from the Southwest of the United States who all spoke conservative dialects of Spanish. The subtest also had adequate sensitivity (85.7%) for a group of 35 Latinx children from the Northeast who all spoke radical dialects of Spanish; however, the specificity for this group was inadequate (Gutiérrez-Clellen & Simon-Cereijido, 2007). The low specificity (62%) appeared to be a result of low performance on specific forms: possessive 's (e.g., “clown umbrella” for “clown's umbrella”), third-person singular (e.g., “jump over” for “every day the horse jumps over the fence”), formulation of negative constructions (e.g., “no wear hat” for “they don't wear a hat”), and formulation of passive constructions (e.g., “it pulled by the cat” for “It is/was pulled”).

As part of the norming portion of the BESA, which is the parent study of the current analyses, Peña et al. (2018) found that children from the Eastern region of the United States (primarily speakers of the radical dialects of Spanish) tended to score more than 10% lower than children from Western and Central regions of the United States on 10 English items. Specific items that were less accurately produced for this group included items that targeted possessive 's, third-person singular present tense, and passive constructions, which are consistent with some of the identified features of AAE (Craig & Washington, 2006; Pruitt et al., 2011; Seymour et al., 1998) and consistent with the findings of Gutiérrez-Clellen and Simon-Cereijido (2007). Adjustment of the cut score for the Morphosyntax section of the BESA raised the sensitivity and specificity of the test for this group to acceptable levels. However, the norming study did not explore the use of AAE features of this subgroup, and it is important to further examine this issue and determine the extent to which they use AAE in their language samples and the relationship between its use and their performance on the individual items of the BESA.

Purpose of the Study

Given the limited research on bilingual English-learning children with and without DLD who have exposure to AAE, we aimed to conduct further fine-grained exploration of the use of forms of AAE in narrative language samples and performance on the BESA Morphosyntax items collected in the BESA parent study. Thus, the purpose of this study was to use existing data on bilingual (balanced bilingual and English-dominant bilingual) children with and without DLD between the ages of 4;0 and 7;1 who have exposure to AAE to address the following objectives:

1. To compare the use of forms of AAE by ability in children's narrative language samples. That is, does the frequency of specific AAE feature use differ between children identified with DLD and those who are considered to be typically developing (TD)?

2. To determine the classification accuracy of the BESA English Morphosyntax subtest in bilingual children who have exposure to AAE.

3. To conduct item analysis of the BESA English Morphosyntax items with bilingual children who have exposure to AAE to determine which items contribute to correct classification of DLD.

Method

Design

The parent study employed a prospective design where participants were tested using reference measures to identify DLD and the field test version of the BESA as the index measure. For this study, we conducted a retrospective analysis of the data to evaluate associations between descriptive measures and performance on the BESA English Morphosyntax subtest.

Participants

Participants were drawn from the field test sample (parent study) used in the development of the BESA (Peña et al., 2018). Of the sample of 744 children (165 with DLD) between the ages of 4;0 and 7;1, there were 183 children from the Philadelphia, PA, area. Note that we oversampled children with DLD to identify items that maximally differentiated them across age levels. The Philadelphia subset of children came from neighborhoods with large Latinx and Black/African American populations and thus presumably had a high degree of contact with AAE. For the current study, we selected those participants from the Philadelphia sample who had complete language sample data and complete BESA item-level data. As will be discussed, the children in the sample were exposed to both English and Spanish in their homes. Because we were interested in the English language performance of the participants, we further selected children for this analysis if they were balanced bilinguals (indicated by percent-correct BESA scores in Spanish and English within 10% of each other) or English-dominant bilinguals (indicated by English scores more than 10% higher in English than in Spanish). This yielded a sample of 81 children (see Table 2 for demographic information). Of these, 28 were identified with DLD, and 53 were identified with typical development.

Table 2.

Participant descriptions.

Variable	TD		DLD
	n = 53 (28 girls, 24 boys, 1 unknown)		n = 28 (11 girls, 17 boys)
	M	SD	M	SD
Age (years;months)	5;11	0;9	5;5	0;10
Percent Spanish exposure	31.83%	(21.28)	35.75%	(21.27)
MLUm	6.29	(0.81)	5.53	(0.81)
Ungrammaticality narratives	19.21%	(11.58)	38.35%	(18.76)

Note. TD = typically developing; DLD = developmental language disorder; MLUm = mean length of utterance in morphemes.

Children were identified with DLD based on the following reference measures based on English and Spanish testing: grammaticality, parent report, teacher report, and clinical observation. Children with DLD met at least three of the following four criteria:

1. ungrammaticality greater than 20%, excluding Spanish-influenced forms, in their English narrative language samples and greater than 20% ungrammaticality in the Spanish narratives if they were able to produce a sample;

2. parent report indicating concern in the area of language in the child's better language;

3. teacher report indicating concern; and

4. clinical observation indicating low responsivity and flexibility during conversation and play.

Materials

Three sets of measures (descriptive, index, and reference) were selected for this study. Descriptive measures were used to characterize children's language performance, including the following narrative measures: number of utterances, mean length of utterance in morphemes (MLUm), total number of words, number of different words, and total nonmainstream dialect density (described below). The English Morphosyntax subtest of the BESA is the index measure we evaluated for its sensitivity and specificity relative to the reference standard for this subset of our population. Reference measures are those used to identify DLD among the participants in the community sample. This set of measures was used to independently classify children into two groups. As stated above, reference measures included grammaticality, parent and teacher observation (Inventory to Assess Language Knowledge [ITALK]), and clinical observation (described below).

Descriptive Measures

Narrative measures. The narrative measures were calculated based on samples obtained using wordless picture books: Frog, Where Are You? (Mayer, 1969) and One Frog Too Many (Mayer & Mayer, 1975) for English and Frog on His Own (Mayer, 1973) and Frog Goes to Dinner (Mayer, 1974) for Spanish. According to Heilmann et al. (2016), the Mayer (1969, 1973, 1974) and Mayer and Mayer (1975) frog stories yield generally consistent data with respect to utterances elicited, mean length of utterance, number of different words, and narrative structure. We coded children's retell and tell narratives from the picture books (described in the Procedure section) for the following measures: number of utterances, MLUm, total number of words, number of different words, and dialect density.

Language exposure. The Bilingual Input–Output Survey (Peña et al., 2018) is used to ask parents and teachers about children's hour-by-hour exposure to Spanish and English. A typical weekday is sampled asking what language the child hears and what language the child responds in for each hour of a typical day. Activities and participants during that hour are identified to facilitate recall of language(s) used each hour. Parents are additionally asked to report on a typical weekend day. The data are projected as a 7-day week to estimate the percentage of input and output in each language. Given that input and output are highly correlated (r = .95; see Bedore et al., 2012), we averaged input and output to yield what we refer to as language exposure. Average percent Spanish exposure for TD children and children with DLD are displayed in Table 2.

Index Measure

BESA English Morphosyntax. This subtest of the BESA (Peña et al., 2018) consists of two task types: grammatical cloze and sentence repetition. As a DLD identification measure, the BESA has 24 grammatical cloze items targeting plural –s, possessive 's, past and present tense, third-person singular, auxiliary + progressive constructions, copulas, negative constructions, and passive constructions. Also, there are nine sentences that children must imitate. Each sentence contains two to six targets, totaling 33 sentence repetition items. Sentence repetition items target complex verb forms, conjunctions, embedded prepositions, and noun phrases. Coefficient alpha ranges from .95 to .96 calculated for ages 4;0–6;11.

Reference Measures

Grammaticality. All narratives were transcribed and coded at the utterance level for grammaticality. Much of the literature on DLD has focused on the challenges faced in MAE learning, for example, the tense marking systems and other grammatical forms such as articles, plurals, and possessives. These are also forms that are variable in early English acquisition for bilingual learners and in nonmainstream dialects. Grammaticality coding, where utterances are rated on overall grammatical well-formedness rather than the presence or absence of particular forms, is sensitive to development and risk for DLD. In this approach, each target language utterance is classified as ungrammatical if it has errors from a pre-established list that contains items such as omissions of articles, prepositions, possessives, or verbs; number agreement errors on articles or demonstrative pronouns; gender errors in the use of possessives; verb tense errors such as omissions, overregularizations, or substitutions; or substitutions of prepositions or pronoun case. Utterances not containing any of these errors are classified as grammatical. See Bedore et al. (2010) for a more complete description. Given the focus of this article on the use of AAE dialect features, it is important to note that some features of AAE may be classified as ungrammatical in this schema. On its own, this has the potential to overidentify children as having risk for DLD. However, within the scope of the convergent use of the referent measures, a child could not be classified as having DLD based on this or any single measure.

ITALK. Parents and teachers completed an experimental version of the ITALK (Peña et al., 2018). On this version of the ITALK, parents and teachers rated (on a 5-point scale) children's language proficiency in Spanish and English (operationalized as how the child uses grammar, comprehension, sentence length, and vocabulary) and language use (operationalized as how much the child used each language). They described any concerns they had about the child's language expression and language comprehension in each language. Ratings of 4 or below in both languages indicating some grammatical or word errors and limited use of both languages were flagged as indicators of DLD. We looked to the descriptions of concerns for verification of the ratings. If concerns were specific to articulation, they were coded as having typical language development. Previous research demonstrates that teacher and parent ratings of child use and proficiency are positively associated with performance on language samples, including grammaticality (Gutiérrez-Clellen & Kreiter, 2003). Bedore et al. (2011) show that teacher and parent ratings on the experimental version of the ITALK significantly correlated with children's performance on morphosyntax and semantics tasks. Finally, Pratt et al. (2020) demonstrate that the questions on the ITALK can be used to accurately flag potential DLD.

Clinical observation. Testers elicited narrative samples during one-on-one sessions with each child. Language samples were obtained in each language on separate days. After each session, clinicians rated child responsivity and clinician effort on a 5-point scale (summed 1–10) based on the Mediated Learning Experience Rating Scale (Lidz, 2002). A summed score of 5 or below indicating low child responsivity with high clinical effort during the narrative retell and tell task was flagged as possible DLD. Previous findings demonstrate that clinical observations of language learning tasks are highly indicative of DLD. Peña et al. (2007) demonstrated that clinical ratings of flexibility and metacognition alone had a 93% accuracy rate with respect to identification of DLD in dual language learners. This is consistent with other findings in dynamic assessment context as applied to various language learning tasks such as narratives (Peña et al., 2014; Petersen et al., 2017), word learning (Kapantzoglou et al., 2012; Peña & Iglesias, 1992; Petersen et al., 2020), and definitions (Ukrainetz et al., 2000).

Procedure

The wordless picture books from the Mercer Mayer frog series were used to elicit oral narratives from each child. Children first heard a story model while looking at the pictures and were asked to retell the story they heard (retell task). Next, they were given a different wordless picture book. They looked through all the pictures with the examiner. Then, they were asked to start at the beginning and to tell a complete story (tell task). Examiners used general prompting and back-channeling to encourage the child to keep going. When the child got to the end of the story, the examiner asked whether they wanted to add more to the story. If the child indicated that they were done, the storytelling session was ended.

For the current analysis, each child utterance was transcribed using Systematic Analysis of Language Transcripts (SALT; Miller & Iglesias, 2019). Utterances were reviewed with a special focus on those that had been marked as ungrammatical relative to mainstream English were reviewed for the presence of any of the features of AAE shown in Table 1. Forms that could be attributed to AAE were assigned a word-level code corresponding to the feature. Word-level codes, as shown in Table 1, identified each of the observed features. For tense-related features, such as subject–verb agreement and zero past tense marker, we used the child's overt tense or adverbial (e.g., today, already) marking of the three previously and/or three following utterances to determine the child's intended temporal reference. In the case that there was no overt tense marking or there was ambiguity (e.g., present and past tenses were used in surrounding six utterances), the utterance was marked as equivocal using a combined code (subject–verb agreement OR zero past tense [SVA][ZPT]) because the production was associated with features of both of these (e.g., “He look around”). It should be noted that tense shifting between present and past was considered to be a typical feature of child narrative (Swasey Washington & Iglesias, 2015) and that zero-marked verbs in these sequences were marked with the combined code. We added this feature as a tenth code (see Appendix A for a language sample using this code). Based on SALT code counts, we calculated the total number of uses of each feature. We note that a traditional approach to calculating the amount of AAE use is to count the number of times a form appears in the language sample divided by the number of opportunities to produce the form. However, for some forms, the number of opportunities cannot be determined, as was the case with our sample. Thus, we coded each language sample for each instance of the feature occurring in Table 1.

Transcription Accuracy

All samples were coded by graduate or undergraduate research assistants. For training purposes, the first author, a native speaker of AAE, coded a set of seven randomly selected samples, and these were employed as the gold standard for training purposes. Each coder coded these samples until they reached 90% accuracy. Both coders reached this standard after the first round of training. Twenty percent of the remaining transcripts were checked for reliability, and differences were resolved as they were identified.

Density of AAE Forms

Because of the potential confounding of the number of words or utterances produced by a student and opportunities for the production of AAE in the language samples, a simple count of nonmainstream forms within language samples may not be the most appropriate measure for assessing frequency of AAE dialect use (Craig & Washington, 2002). According to Ivy and Masterson (2011), dialect density provides a reflection of the proportion or ratio of one's overall dialect use in linguistic production (see also Craig et al., 1998). Density of nonmainstream forms can be calculated in various ways, including counting the number of utterances with at least one nonmainstream dialectal form, dividing the number of forms (or tokens) produced by the total number of words, and dividing the number of forms by the total number of utterances in the language sample (Oetting & McDonald, 2002). For this study, we chose the third method (forms divided by number of utterances) based on suggestions by Oetting and McDonald (2002) regarding greater range among participant scores, and we calculated the density of AAE forms in two ways. First, we calculated the total density of forms by adding the total number of AAE forms produced (see Table 1) and dividing it by the number of utterances produced. The resulting quotient was multiplied by 100, resulting in the percentage of dialect frequency for the language samples or total nonmainstream form density. Next, to analyze the use of each of the individual AAE forms, we calculated dialect density for each feature by counting the number of times the feature was produced in each sample and dividing that number by the number of utterances produced, then multiplying by 100, resulting in a dialect density measure for each individual feature. We examined total nonmainstream form density and density of each individual feature used by students with DLD in comparison to those students who were considered TD.

Results

AAE Feature Use by Ability

Of the 81 children in the study, four had missing retell narratives, and nine had missing tell narratives. We compared the narrative measures in the two narrative conditions to examine possible differences between the retell and tell tasks. Repeated-measures analysis of variance examining narrative condition as the within-subject factor was not significant, F(1, 67) = 0.480, p = .491, η_p ² = .007. Children's stories were similar with respect to number of utterances, MLUm, total number of words, number of different words, and total dialect density. We thus averaged the measures' scores across the retell and tell tasks to include all 81 children in the analysis.

In preliminary analyses examining AAE dialect use in the language samples, we found that, across ability groups (TD and DLD), the AAE forms were used by at least one child. Subject–verb agreement, zero copula/auxiliary, zero past tense marker, and the equivocal subject–verb agreement or zero past tense were the features used most frequently. We then compared the use of each of the 10 AAE forms by the number of utterances across ability groups. The 10 features were entered into a multivariate analysis of covariance to compare children's performance with and without DLD with age in months as a covariate. There was no effect for age, F(10, 70) = 1.476, p = .167. We thus reran the analysis as a multivariate analysis of covariance with ability as the between-subjects factor. Results demonstrated a main effect for ability, F(10, 71) = 3.316, p = .001, η_p ² = .318, a moderate effect size. Post hoc analyses indicated that there were significant differences in use of AAE forms affecting subject–verb agreement, F(1, 80) = 11.189, p = .001, η_p ² = .123; zero copula/auxiliary, F(1, 80) = 12.308, p = .001, η_p ² = .133; and subject–verb agreement or zero past tense, F(1, 80) = 13.519, p < .001, η_p ² = .145. Children with DLD used these forms more frequently than the TD children. There were no differences for the other features including zero past tense, double marking past tense, overgeneralization, multiple negation, zero possessive, zero plural, and invariant BE. The means and standard deviations for each of the forms are displayed in Table 3.

Table 3.

African American English feature dialect density in oral narrative tasks by ability.

Variable	TD (n = 54)			DLD (n = 28)			F(1, 80)	p
	M	SD	Range	M	SD	Range
Total dialect density	14.21%	8.75	0–39.2	25.17%	13.33	0.08–70.9	19.77	< .001
Zero past	4.02%	3.76	0–17.5	3.64%	3.15	0–12.5	0.216	.643
Subject–verb agreement	2.13%	2.88	0–13.0	5.03%	4.98	0–20.5	11.189	.001
Zero copula/aux	1.83%	1.95	0–7.1	4.05%	3.79	0–13.8	12.308	.001
Double mark past	0.58%	1.36	0–7.4	0.44%	1.05	0–3.8	0.222	.639
Overgeneralization	0.51%	1.62	0–10.2	0.52%	1.04	0–3.5	0.001	.974
Zero possessive	0.40%	0.84	0–4.3	0.36%	0.86	0–3.2	0.045	.832
Zero plural	0.22%	0.56	0–2.1	0.29%	0.62	0–2.1	0.29	.592
Double negative	0.06%	0.30	0–1.4	0.04%	0.22	0–1.1	0.061	.805
Invariant BE	0.00%	0.00	NA	0.11%	0.56	0–2.9	1.951	.166
Equivocal subject–verb agreement or zero past	3.00%	4.02	0–24.3	8.10%	8.58	0–41.8	13.519	.000

Note. TD = typically developing; DLD = developmental language disorder.

Classification Analysis

We were interested in the extent to which the Morphosyntax subtest of the BESA accurately classified bilingual children who used and were exposed to Spanish and AAE without adaptation of scoring rules. We evaluated the classification accuracy of the two subscales (cloze and sentence repetition) of the Morphosyntax subtest and the composite derived from equal weighting of the two subscales combined. For this analysis, raw scores were converted to scaled scores for the two subscales to control for expected age-related differences. Here, the scaled scores range from 0 to 20, with a mean of 10 and an SD of 3. For the composite, the two scaled scores are added and converted to a standard score with a mean of 100 and an SD of 15. Children with DLD had significantly lower scores than the children with TD for the three scores (see Table 4).

Table 4.

Comparison of Bilingual English– Spanish Assessment Morphosyntax scores by ability.

Measure	TD		DLD		F	p	Canonical correlation
	M	SD	M	SD
Morphosyntax composite	93.30	11.07	74.43	10.48	55.139	< .001	.641
Cloze	7.89	2.58	4.64	1.85	34.635	< .001	.552
Sentence repetition	9.47	2.71	5.07	3.09	43.862	< .001	.597

Note. TD = typically developing; DLD = developmental language disorder.

We used receiver operating characteristic (ROC) curves to identify optimal cut-points for each of the three scores (cloze, sentence repetition, and composite). Optimal cut-points represent the scores that maximize both sensitivity and specificity. Using the sensitivity and specificity results, we calculated positive likelihood ratios (LR+; the likelihood that children with DLD are accurately classified) and negative likelihood ratios (LR−; the likelihood that children without DLD are accurately classified). We also estimated the area under the curve (AUC) using the ROC analysis. AUC provides an estimate of sensitivity and specificity at each possible cut-point. In our application, it represents the probability that a random participant with DLD ranks more highly than a random participant with typical development. AUC values between .70 and .80 are considered acceptable, those between .80 and .90 are excellent, and values above .90 are considered outstanding (Rice & Harris, 2005). In addition, we compared item responses on the Morphosyntax subtest of the BESA to compare individual responses by ability.

Sensitivity and specificity for the cloze, sentence repetition, and composite scores are displayed in Table 5. Figure 1 shows the ROC curves for the three measures. The AUC values for the composite, cloze, and sentence repetition scores were .88, .85, and .86, respectively, which are all in the excellent range. The best classification accuracy was observed for the composite score, which incorporates both cloze and sentence repetition. Cut scores and likelihood ratios are also displayed in Table 5. The composite score, which consists of an equal weighting of the cloze task and sentence repetition, had the highest classification with 89.3% sensitivity and 81.1% specificity using a cut score of 83. The positive likelihood ratio of 4.72 and negative likelihood ratio of 0.13 are in the very likely range. By itself, the cloze task has high sensitivity (89.3%) but relatively low specificity (62.3%). A positive likelihood ratio of 2.37 indicates that a positive result (indicating DLD) is suggestive but not certain. On the other hand, the negative likelihood ratio of 0.17 shows that a negative result (indicating TD) is very likely certain (with smaller numbers being more certain). The sentence repetition task classification is just below the threshold of 80%, with 78.6% sensitivity and 79.3% specificity. The positive likelihood ratio of 3.78 is again suggestive but not certain, and the negative likelihood ratio of 0.27 is very likely.

Table 5.

Classification accuracy for morphosyntax measures.

Measure	Cut score	DLD as DLD	Sensitivity	TD as TD	Specificity	LR+	LR−	AUC
Morphosyntax composite	83.87	25/28	89.3	43/53	81.1	4.72	0.13	.882
Cloze	6.26	25/28	89.3	33/53	62.3	2.37	0.17	.846
Sentence repetition	7.27	22/28	78.6	42/53	79.2	3.78	0.27	.855

Note. DLD = developmental language disorder; TD = typically developing; LR+ = positive likelihood ratio; LR− = negative likelihood ratio; AUC = area under the curve.

Figure 1.

Receiver operating characteristic (ROC) curve.

Item Analysis of the BESA

To better understand the contributions of individual items to the BESA Morphosyntax classification results, we compared children's responses on the individual BESA items by ability. We submitted the 24 cloze and 33 sentence repetition items to χ² analysis to determine if there were differences in proportion of children with and without DLD responding correctly to each item. We used the Benjamini and Hochberg (1995) procedure to control the rate of false discovery when multiple tests are conducted. In this procedure, we ordered p values from smallest to largest and retained only smaller p values based on the number of comparisons (n = 57). We set a false discovery rate of 0.05 and generated the critical value. Unadjusted values were considered significant if they were less than .043. Of the 57 comparisons, nine showed no significance by ability. Of these, five were cloze items (one passive, one copula, two auxiliary + progressive, and one regular past), and four were sentence repetition items (one auxiliary + progressive, one conjunction, and two article + noun). Item difficulty for children with and without DLD, discriminant values, and individual χ² results are displayed in Appendix B.

Discussion

When children's language production patterns vary based on exposure to more than one language or dialect, it is difficult to know to what extent the use of grammatical forms indicates typical variability or DLD within linguistic variability. It is important to develop a better understanding of dialectal variation (AAE in this case) in all children (e.g., monolingual and bilingual) with DLD and the ways that children's use of AAE dialectal features may influence classification accuracy. In this retrospective study, we reanalyzed BESA English Morphosyntax performance and narrative production in a group of Spanish-English–speaking children with exposure to AAE. Recent work has documented nonmainstream feature use in monolingual children with DLD and has considered the extent to which the forms children produce are reliable clinical markers. Here, our approach differs from past studies in that we focus on nonmainstream dialectal features used by bilingual children exposed to Spanish and AAE. We evaluated typical and atypical children's use of forms of AAE in a narrative retell and tell task and examined their performance on the BESA Morphosyntax items collected in the BESA parent study. In this study, we focused and engaged in a fine-grained analysis of the language of a group of Latinx children from the Northeast who were presumably exposed to AAE.

It is a well-established fact that there is significant overlap between nonmainstream English features and DLD features among MAE-speaking children. Traditionally, researchers have worked to provide data to differentiate DLD features in MAE from typical features of nonmainstream dialects. However, researchers argue the need to reanalyze this notion. We need to examine language and speech patterns among both TD children and children with DLD who speak a variety of nonmainstream dialects and/or are bilingual (Oetting, 2018). Our study directly addresses this concern, examining the accurate identification of language disorder among bilingual children who are also speakers of AAE.

First, we compared the use of nonmainstream forms of AAE by ability in children's narrative language samples. In our preliminary analysis, we found that the bilingual children together used all of the coded nonmainstream forms of AAE in their narratives, and each feature was used at least once by a minimum of one child in each group—TD children and those with DLD. Our sample's total AAE dialect density was similar to that of previous research among monolingual children (e.g., Gatlin & Wanzek, 2017; Oetting & McDonald, 2001). Also, similar to previous findings, children with DLD generally used more features of AAE in their language samples (e.g., Hendricks & Adlof, 2020). Regarding individual AAE dialect features, in the narratives, children used four of the 10 features coded in the samples most frequently: subject–verb agreement variations, zero copula/auxiliary, zero past tense, and the equivocal subject–verb agreement or zero past tense. These are often used by TD African American children. For example, in Connor and Craig (2006), 60% of the children used zero copula/auxiliary, 58% used subject–verb agreement variations, and 39% used zero past tense marker. Only two children in our study did not use any of the AAE features selected for analysis. On one hand, this is consistent with the notion that speakers may not always produce all features associated with the dialect. Furthermore, older children may use fewer features after school entry (e.g., Craig & Washington, 2004; Terry & McDonald Connor, 2012). The two children who did not use the features selected for this study were TD and were of kindergarten age or older (5;8 and 7;1).

Except for the zero past tense, the same most frequently occurring forms coded in the narratives were produced significantly more often by the children with DLD than their peers. A higher rate of occurrence of these features is consistent with patterns of performance reported in studies of monolingual speakers of AAE with DLD (Hendricks & Adlof, 2020) and the notion of “double vulnerability”—vulnerable as a developmental error and a nonmainstream dialectal form. In comparison, forms such as zero plural and zero possessive did not vary in use across the groups. This suggests that some, but not all, of the features that might be expected to be of high frequency in children with DLD due to double vulnerability are produced at a higher rate. There are several possible reasons for this pattern that bear further consideration. Both subject–verb agreement and the equivocal productions are related to marking of tense, which is demanding for English-learning children. One source of difficulty is the multiple manifestations of tense marking (regular and irregular forms alongside overtly marked and unmarked forms create a high level of variability) may make it more challenging for children with DLD to acquire the same patterns as do their peers. The fact that it was some but not all forms with the potential for double vulnerability speaks further to the probabilistic nature of grammatical production. Finally, there was a subset of forms that did not differ across TD and DLD groups, including double marking of verbs or constructions with multiple negation. These forms require overt knowledge of the target so that it can be produced. Bilingual learners of English sometimes, but not always, produce these forms and, in such cases, represent productive knowledge of the grammatical system (see Jacobson & Schwartz, 2005). As such, it may be less likely that the participants in this study will produce these forms or constructions given their relatively young age.

Similar to our study, Hendricks and Adlof (2020) found that children with DLD who use nonmainstream English overtly marked past tense and third-person singular (variable subject–verb agreement) less frequently than their TD nonmainstream dialect–speaking peers. A similar finding for third-person singular, zero past tense, and zero auxiliary was reported by Oetting et al. (2019, 2021) and Garrity and Oetting (2010). Together, these findings provide information on which features of AAE might need to be taken into consideration when developing assessments for identifying students with DLD when also considering dialect use.

Given that these children demonstrated use of forms of AAE, we examined whether the BESA English Morphosyntax subtest had appropriate classification accuracy using ROC curves. As expected, there was a significant difference in group performance on the BESA English Morphosyntax subtests and composite scores. The fair-to-good classification accuracy is in line with that reported in the BESA manual, which includes children representing Western, South Central, and Eastern regions of the United States. However, note that, in the present analysis, we used a cut score of 83, whereas the cut scores for the BESA are 86, 85, and 81 for 4-, 5-, and 6-year-old children, respectively, as reported in the manual. This classification accuracy is an improvement over the previous study of the earlier experimental version of the BESA (Gutiérrez-Clellen & Simon-Cereijido, 2007) where a total score based on 32 cloze items and 31 sentence repetition items yielded 85.7% sensitivity and 61.9% specificity. In contrast, the final version of the BESA includes 24 cloze items and 33 sentence repetition items. The two task types are equally weighted in the composite by converting raw scores for each subsection to scaled scores, summing them, and then converting these to a standard score.

The third research question of this study aimed to investigate the discriminant contribution of the individual BESA Morphosyntax items to correct classification of DLD. The majority of the items did differentiate between TD children and children with DLD using χ² analysis. Note that the difference in accuracy between children with and without DLD on these items was at least 24%. A closer examination of the items that did not differ significantly between the two groups (see Appendix B) shows that several cloze items failed to differentiate TD children and children with DLD because they were not produced by either group of children. The production patterns children learn arise from the convergence of input they hear. The children in this study are likely to have learned based on AAE, MAE, and Spanish-influenced English. One factor may be convergence of phonological rules. Neither AAE nor Spanish-influenced English favors final consonant clusters. Comparison of a past tense cloze item that failed to differentiate children, “dropped the balls,” versus one that did, “jumped over the fence,” helps to illustrate this. Children who are dialect speakers may zero-mark the tense by reducing the production of the /t/ sounds in the context of the final “p” in “drop”. In contrast, when producing “jumped over,” they are less to reduce the “t.” Indeed, in post hoc analyses in their study among 5- and 6-year-olds, Pruitt and Oetting (2009) found that past tense marking was affected by the phonological characteristics of the items. Specifically, the researchers found differences among regular past tense verbs that were classified as high-probability items, those that increased the likelihood of overt marking and decreased the probability of zero marking (e.g., dry/dried). Low-probability items were those that decreased the probability of overt marking and increased the probability of zero marking (e.g., walk/walked). High-probability items were zero-marked less frequently than low-probability items. As noted by Wolfram and Schilling (2016), consonant cluster reductions appear more often when the cluster is followed by a consonant (e.g., “best aunt, bes' kid”), even among mainstream English users. Overtly producing the phoneme /s/ to mark the possessive before a noun beginning with a consonant sound, relative to nouns beginning with vowel sounds, may be difficult for children who are bilingual and/or are AAE dialect speakers, regardless of their status as TD or having DLD. However, children with DLD may not be adept at producing consonant clusters, even when the following noun begins with a vowel sound.

Other examples of forms that may be associated with convergence of cues across sources of input include progressives such as “are skating” and “were eating.” In these cases, children responded with the progressive only (“skating”, “eating”) or used zero-marked auxiliaries (“is skating”, “was eating”). These forms are consistent with AAE rules and manifest as being low in accuracy or difficult because neither the children with or without DLD used the MAE target but instead expressed the target with a different phrase to describe the target action. Recording individual student responses and directly comparing them to spontaneous language samples in future studies could provide more information regarding this hypothesis. It is also worth noting that some sentence repetition items failed to differentiate because the children produced different forms than the targets. Some examples were the replacement of the conjunctions or noun phrases in the sentence repetition items.

In summary, the individual items on the BESA that failed to differentiate children with and without DLD potentially have different sources of difficulty. Some morphosyntactic features are challenging to produce (e.g., final consonant clusters associated with past tense or plurals and weak syllables associated with article-noun combinations), while others may have placed greater memory or processing demands on the children (e.g., question inversion or passive constructions). A key difference between the structured test items and narrative performance is that the constrained nature of the test task may make these items difficult for all children. Some of the items and, hence, forms that did not differentiate DLD from typical development on the BESA English Morphosyntax subtest revealed differences in the narrative samples. For example, in the narrative samples, children with and without DLD did not differ in zero marking of past tense but differed in the use of copula and auxiliary forms. Future work should focus on the nature of the child productions to understand how and when children differ.

Limitations

Although the study's findings are informative, it is important to note that because this study is the first, to our knowledge, to examine more than two to three forms of AAE among young, bilingual Spanish-English–speaking children, it is not without limitations. First, the language tasks did not generate equal amounts of data for each form, rendering it difficult to confidently state that frequency of production of individual forms could separate the two groups of students (TD and DLD). In this study, we did not analyze opportunities for production of forms of AAE, but rather the production of each form. Because we do not know how many opportunities each child had to produce a particular form, the data may not be sufficient for finding differences for these forms. For instance, no child in the TD group produced the invariant/habitual BE, but we are unaware if the children's language samples provided opportunities to produce this particular feature, making the difference in the findings a product of the story structures rather than a true difference between TD children and children with DLD. Future research may explore opportunities for production of features of AAE and/or elicit opportunities for particular forms to explore this notion further. Second, the nature of the wordless picture books used in this study elicits ambiguous tense contexts. In both AAE and MAE dialects, children may alternate present and past tenses, making it difficult to accurately assess subject–verb agreement and zero past tense marking. Future work may more explicitly guide students to use one tense for language sample analysis.

Conclusions

Our study highlights the need for culturally and linguistically fair assessment practices in the identification of DLD among monolingual and bilingual children. In the assessment of language disorders among culturally and linguistically diverse students, bilingualism and nonmainstream dialect use cannot be ignored. Recently, researchers have advocated for a disorder within diversity framework to replace the previous difference versus disorder approach (Oetting, 2018; Oetting et al., 2016). This approach calls for researchers and practitioners to take local norms and language varieties of the community into consideration when assessing language abilities of children, especially in the context of diagnosing language disorders. Language differences that are attributable to a child learning English as a second language or speaking a nonmainstream dialect of English do not constitute language disorders. In order to avoid misclassification (overidentification and underidentification) of language disorders, particularly among linguistically diverse students, it will be important for researchers and practitioners to know and understand what is considered typical in a child's language environment. In our study, we examined language profiles of young children in the Philadelphia, PA, area presumably exposed to varieties of English and Spanish. Future research may purport to examine language profiles of bilingual dialect speakers in other geographical regions of the United States to learn more about the norms of those communities. Furthermore, researchers and language test developers may also conduct large studies in which children's language use across various geographical regions is compared to assess differences across the country to better prepare assessments and technical manuals. Moreover, research is needed to explore the relative contribution of phonology and other domains impacting the use of nonmainstream dialectical features. This will help us improve assessment and interventions to promote language learning of all children with DLD.

Language disorders exist among those who are learning English as a second language, students who speak with a nonmainstream dialect of English, and those who fall into both categories. For researchers and clinicians, understanding more about children's lived experiences and local norms is imperative to accurately identify what entails a DLD and, thus, what services should be provided. Anyone who speaks a dialect speaks a dialect of that language, which includes mainstream or what most would refer to as standard or, perhaps, standardized English. However, dialect features that are consistent with nonmainstream or socially stigmatized varieties of English are often confounded with the assessment and identification of language disorders. In speech and language pathology and in education in general, the development of culturally and linguistically responsive assessment practices is just as important as incorporating instructional practices that are sensitive to the needs of our nation's increasingly diverse students. Based on this study's results, the BESA English Morphosyntax subtest appears to be a valid tool for the identification of DLD in children exposed to AAE and Spanish.

Appendix A

Coded Language Sample With Equivocal Subject–Verb Agreement/Zero Past Tense

Gender: F

Chronological Age: 72 months

Student ID# 1008

[BeginFrogtell]

C the boy look[ZPT][SVA] at the frog [SI-1] [U].

C the dog look[ZPT][SVA] in the thing [SI-1] [U].

C the boy was|auxbe sleep/ing [SI-1] [G].

C the frog jump/ed out [SI-1] [G].

C (the) he said|say/ed [+”] uh_oh [SI-1] [G].

C [+”] where ‘s|be the frog [SI-1] [G].

C the puppy dog look/s back [SI-1] [G].

C he look/s in his boot [SI-1] [G].

C he look/s everywhere [SI-1] [G].

C he ‘s|auxbe call/ing the frog [SI-1] [G].

C the puppy fell|fall/ed [SI-1] [G].

C the boy ‘s|be angry [SI-1] [G].

C he (‘s|auxbe) keeps|auxkeep/3 s[ZPT][SVA] calling|call him [SI-1] [G].

C (the tree) the tree was|auxbe moving|move/ing [SI-1] [G].

C he ‘s|auxbe call him in the hole [SI-1] [U].

C he ‘s|auxbe call him in the beetles [SI-1] [U].

C he ‘s|auxbe call him in the :>.

C he *be mad[COP] [SI-1] [G].

C (he's) he ‘s|be angry [SI-1] [G].

C (he) the dog: is|auxbe stand/ing up [SI-1] [G].

C the bee/s was|auxbe[SVA] get/ing out of his thing [SI-1] [G].

C (the:) the tree was|auxbe moving|move/ing [SI-1] [G].

C (the tree) the boy almost see[ZPT][SVA] the owl [SI-1] [G].

C the dog run[ZPT][SVA] away [SI-1] [U].

C (the boy hear his no) he put his hand up [SI-1] [G].

C he ‘s|be hot [SI-1] [G].

C he keep|auxkeep[ZPT][SVA] call/ing the frog [SI-1] [U].

C he ‘s|be on the x [SI-1] [G].

C he ‘s|be on (the this) this [SI-1] [G].

C (he) the boy fell|fall/ed [SI-1] [G].

C the dog fell|fall/ed [SI-1] [G].

C the boy found|find/ed the water [SI-1] [G].

C the dog was|be scared [SI-1] [G].

C (the boy wasn't the boy was|be/ed n't scare/ed) the boy was|be n't scared [SI-1] [G].

C (the do* the the the boy was) he said|say [+”] be quiet [SI-2] [G].

C (then) they get[ZPT] out the water [SI-1] [G].

C (then) they saw|see the two frog/s (the two frog) [SI-1] [G].

C (they saw they get) they came|come/ed out [SI-1] [G].

C (cough) they said|say/ed goodbye [SI-1] [G].

C (Aw) there ‘s|be a good (li*) little froggie [SI-1] [G].

C ((and the end)).

+ [EndFrogtell]

Appendix B

Item Difficulty Results for Bilingual English–Spanish Assessment (BESA) Items

Task	BESA item no.	Target	Difficulty		D-value	χ2	p	BH
			TD	DLD
Cloze	1	Possessive 1	.49	.14	.35	9.499	.002	sig
	2	Possessive 2	.57	.32	.24	4.391	.036	sig
	3	Possessive 3	.53	.29	.24	4.367	.037	sig
	4	3rd singular 1	.52	.21	.30	6.984	.008	sig
	5	3rd singular 2	.45	.11	.35	9.852	.002	sig
	6	3rd singular 3	.36	.00	.36	13.114	.000	sig
	7	Regular past 1	.70	.32	.38	1.594	.001	sig
	8	Regular past 2	.57	.29	.28	5.781	.016	sig
	9	Regular past 3	.49	.36	.13	1.321	.250	ns
	10	Plural 1	.92	.61	.32	12.231	.000	sig
	11	Plural 2	.38	.14	.23	4.832	.028	sig
	12	Plural 3	.92	.61	.32	12.231	.000	sig
	13	Progressive 1	.32	.19	.14	1.649	.199	ns
	14	Progressive 2	.36	.04	.32	1.265	.001	sig
	15	progressive 3	.17	.07	.10	1.511	.219	ns
	16	Copula 1	.51	.14	.37	1.421	.001	sig
	17	Copula 2	.81	.54	.28	6.845	.009	sig
	18	Copula 3	.32	.14	.18	3.019	.082	ns
	19	Negative 1	.69	.30	.40	11.295	.001	sig
	20	Negative 2	.62	.07	.55	22.685	.000	sig
	21	Negative 3	.28	.00	.28	9.726	.002	sig
	22	Passive 1	.36	.04	.32	1.265	.001	sig
	23	Passive 2	.15	.04	.12	2.463	.117	ns
	24	Passive 3	.34	.11	.23	5.156	.023	sig
Sentence repetition	1	Article + noun	.94	.85	.09	1.801	.180	ns
	2	Pronoun	.75	.26	.49	17.597	.000	sig
	3	Progressive	.77	.59	.18	2.687	.101	ns
	4	Copula	.83	.30	.53	21.87	.000	sig
	5	Article + noun	.92	.59	.33	12.619	.000	sig
	6	Article + noun	.87	.56	.31	9.34	.002	sig
	7	Adjective	.69	.37	.32	7.573	.006	sig
	8	Do question	.73	.26	.47	16.117	.000	sig
	9	Pronoun	.90	.52	.39	14.999	.000	sig
	10	3rd present	.75	.52	.23	4.313	.038	sig
	11	Infinitive	.83	.41	.42	14.46	.000	sig
	12	Past	.81	.33	.47	17.478	.000	sig
	13	Article + noun	.67	.30	.38	1.172	.001	sig
	14	Preposition	.83	.37	.46	16.777	.000	sig
	15	Article + noun	.90	.52	.39	14.999	.000	sig
	16	Conjunction	.44	.30	.15	1.589	.207	ns
	17	Pronoun + irregular past	.65	.22	.43	13.246	.000	sig
	18	3rd singular + infinitive	.56	.11	.45	14.708	.000	sig
	19	Interrogative pronoun	.88	.63	.25	7.154	.007	sig
	20	Conjunction	.83	.52	.31	8.413	.004	sig
	21	Plural	.81	.44	.36	1.841	.001	sig
	22	Pronoun	.77	.37	.40	12.169	.000	sig
	23	Article + noun	.73	.59	.14	1.569	.210	ns
	24	Conjunction	.63	.22	.41	12.092	.001	sig
	25	Pronoun	.77	.30	.47	16.672	.000	sig
	26	Regular past	.44	.11	.33	8.829	.003	sig
	27	Auxiliary	.50	.26	.24	4.235	.040	sig
	28	Article + noun	.94	.63	.31	12.639	.000	sig
	29	Preposition	.88	.56	.33	1.621	.001	sig
	30	Possessive pronoun + noun	.90	.44	.46	19.381	.000	sig
	31	Copula	.73	.19	.54	2.737	.000	sig
	32	Pronoun	.82	.48	.34	9.934	.002	sig
	33	Preposition	.84	.59	.25	6.014	.014	sig

Note. TD = typically developing; DLD = developmental language disorder; Difficulty = proportion of children in group who responded correctly to the item; D-value = the difficulty difference between children with and without DLD; BH = Benjamini–Hochberg critical value = .043; sig = significant; ns = nonsignificant.

Funding Statement

Funding for this study was provided by National Institute on Deafness and Other Communication Disorders Grant N01DC82100, awarded to Aquiles Iglesias. The views expressed are ours and do not necessarily represent the views of the funders.