Production of Morphosyntax Within and Across Different Dialects of American English

Abstract

Purpose

This study examined the production of morphosyntactic markers by school-age children with and without developmental language disorder. Comparisons were made between students who speak mainstream American English (MAE) dialects and nonmainstream American English (NMAE) dialects.

Method

First- and second-grade students (N = 82) completed assessments of dialect use and language ability, which are designed for students who speak NMAE dialects. Students also completed an experimental production task targeting three morphosyntactic features: past tense –ed marking, third-person singular –s marking, and plural –s marking. Past tense marking and third-person singular are produced differently across MAE and NMAE dialects, whereas plural marking is produced more similarly across dialects.

Results

When comparing across dialects, children with typical language skills who spoke NMAE dialects overtly marked past tense and third-person singular less often compared to MAE peers. However, when comparing to same-dialect peers with language disorders, children with typical language skills who spoke NMAE dialects overtly marked these morphosyntactic markers more often than peers with developmental language disorder.

Conclusion

The results underscore the importance of considering a child's dialect use when assessing language ability, in particular with measures that include features that are variable in NMAE dialects. At the same time, within-dialect comparisons suggest that a broader set of morphosyntactic features may provide useful information for evaluations of language ability. Future research should investigate the source of these differences, including the extent to which students with language disorders have acquired the social and linguistic factors that condition the use of variable features.

As many as 70% of school-age children with language disorders, such as developmental language disorder (DLD), are not identified as having language difficulties (Adlof et al., 2017; Tomblin et al., 1997). Accurate identification of language disorders is particularly difficult in school-age children from culturally and linguistically diverse backgrounds, including those who speak nonmainstream American English (NMAE) dialects, such as African American English (AAE; Stockman, 2010). All dialects are systematic rule-governed communication systems and use of a nonmainstream dialect should not be mistaken as a communication disorder (American Speech-Language-Hearing Association, 2003). Still children from minority backgrounds are overrepresented in special education (e.g., Harry & Klingner, 2014). U.S. Department of Education statistics suggest that a larger proportion of African American children are receiving special education services compared to all other racial groups (Office of Special Education and Rehabilitative Services, 2016). However, using a large national data set, Morgan and colleagues recently found that African American students were less likely to be receiving language services by 5 years of age compared to their White peers, when controlling for other factors such as socioeconomic status and access to health care (Morgan et al., 2015, 2016). Two types of misdiagnoses are problematic. Children who are overidentified may spend time completing potentially unnecessary intervention activities. Furthermore, misdiagnosing language variation as a language disorder may contribute to social stigma against the use of NMAE dialects. On the other hand, children who are underidentified do not receive interventions that would improve their academic progress. Thus, there is an urgent need for research on language disorders in school-age children from culturally and linguistically diverse backgrounds.

Most assessments of language ability have been developed for children who speak mainstream American English (MAE), and therefore, differences between MAE and NMAE dialects constitute one source of difficulty when assessing language in children who speak NMAE dialects. One key difference between MAE and NMAE dialects is that some morphosyntactic features, which are obligatory in MAE, can be produced variably in NMAE dialects. Features that differ across dialects have been called “contrastive features,” while features that are produced similarly across dialects have been called “noncontrastive” features (Seymour et al., 1998). For example, past tense (PT) marking and third-person singular (3SG) –s marking are produced variably in NMAE dialects (cf. She walked the dog, He walkø away; Oetting & McDonald, 2002) but are obligatory in MAE. Therefore, these features are typically considered contrastive features, even though they differ slightly in how frequently they are overtly produced and the factors influencing the likelihood of overt production (Lee & Oetting, 2014). Plural marking, however, is overtly produced much more frequently in NMAE dialects and is therefore considered a noncontrastive feature. The probability that a morphosyntactic feature will be overtly produced is sociolinguistically conditioned and dependent on both linguistic factors, such as the phonological form (Lee & Oetting, 2014; Pruitt & Oetting, 2009) and the formality of the discourse (Thompson et al., 2004; Washington et al., 1998), as well as social factors such as the speaker's gender (Washington & Craig, 1998), age, and socioeconomic status (Horton-Ikard & Miller, 2004; Terry et al., 2010). For example, using production probes and language samples from 45 preschool children, Pruitt and Oetting (2009) found that overt production of PT marking differed based on the phonological form (i.e., more likely to overtly mark when verb stem ended in a vowel) and the child's age. Interestingly, in contrast to other work, Pruitt and Oetting did not find differences based on the child's socioeconomic status. While zero marking of these features is grammatical in NMAE dialects, the omission of the obligatory morphosyntactic features, such as tense and agreement markers, are a hallmark of DLD in children who speak MAE (Bedore & Leonard, 1998; Rice & Wexler, 1996; Tager-Flusberg & Cooper, 1999).

Some researchers have argued that the assessment of contrastive features (i.e., those that are produced differently between MAE and NMAE dialects) does not provide meaningful information for the assessment of language ability in children who speak NMAE dialects, such as AAE (Seymour et al., 1998), and many assessment approaches aim to avoid assessing contrastive features. Some assessments, such as the Diagnostic Evaluation of Language Variation–Norm Referenced (DELV-NR; Seymour et al., 2005), choose to focus on noncontrastive features (Seymour et al., 1998), such as negation and determiners, which do not differ across dialects of American English. Other standardized assessments include both contrastive and noncontrastive features but allow for modifying scoring to give credit for items that would have been scored as incorrect in MAE if they contain a grammatical use of an NMAE dialect (e.g., Clinical Evaluation of Language Fundamentals–Fifth Edition; Wiig et al., 2013). Implementing modified scoring, however, is not straightforward, and while modified scoring may reduce the risk of false positives, it may also lead to an increased risk for false negatives (Hendricks & Adlof, 2017). Others suggest using dynamic assessment (e.g., Peña et al., 2006) and processing-based assessments, such as nonword repetition (e.g., Campbell et al., 1997; Dollaghan & Campbell, 1998), to inform diagnostic decisions without assessing contrastive features. While nonword repetition was suggested as a dialect neutral measure, others have documented lower performance on nonword repetition among students who use NMAE features more often (McDonald & Oetting, 2019; Moyle et al., 2014).

Whereas contrastive morphosyntactic features have often been avoided in the diagnostic assessment of language ability assessments for children who speak NMAE dialects, some researchers have hypothesized that a broader range of features may be informative for assessments of language ability. However, the utility of contrastive features depends on whether performance is compared to peers who speak the same dialect or a different dialect (see Leonard, 2014, Figure 3.3, p. 92; Oetting, 2018; Oetting et al., 2016). A growing body of research involving younger children supports these hypotheses and suggests that children with language disorders produce both contrastive and noncontrastive features differently than their same-dialect peers without language disorders (Cleveland & Oetting, 2013; Garrity & Oetting, 2010; Oetting et al., 1999; Oetting & Newkirk, 2008). Consistent with previous studies, research by Oetting and colleagues demonstrates that typically developing children who speak different dialects differ in the rate of production of overt markers of tense and agreement, such as PT and 3SG present tense. Importantly, when comparing within students who speak NMAE dialects, typically developing children overtly produce morphosyntactic markers more often than their peers who have language disorders. Among children ages 3–6 years, differences have been noted using broad syntactic measures such as the IPSyn, which assesses whether students produce a set of syntactic structures during a language sample (Oetting et al., 1999, 2010) and specific morphosyntactic markers such as 3SG –s (Cleveland & Oetting, 2013) and subject relative markers (Oetting & Newkirk, 2008). In an analysis of language samples from 30 six-year-old children who spoke AAE, Garrity and Oetting (2010) documented that children with specific language impairment—the most common type of DLD—overtly marked auxiliary BE forms (am, is, are) less often than both age- and language-matched controls. Based on this growing body of research, Oetting and colleagues have argued that analysis of contrastive features may provide useful information about the language ability of young children who speak NMAE dialects as long as the analysis includes the appropriate within-dialect comparison (Oetting, 2018; Oetting et al., 2016).

Studies suggesting that within-dialect comparisons allow for language ability assessments to be expanded to include contrastive and noncontrastive features have primarily focused on younger children in preschool or kindergarten (Cleveland & Oetting, 2013; Garrity & Oetting, 2010; Oetting et al., 1999; Oetting & Newkirk, 2008; but see Seymour et al., 1998). However, the extent to which the research on preschool and kindergarten students can be extended to school-age children is unclear. There is a critical need for additional research on the use of NMAE features in children in early elementary school. For some students, the beginning of elementary school may be their first intensive exposure to MAE, as they encounter peers who speak a different dialect and much of their instruction is delivered in MAE. Due in part to the changes in linguistic input and environment, research on typically developing children has documented substantial changes in the frequency with which students with typical language (TL) skills use NMAE features during the early elementary school grades, with students increasing their use of overt morphosyntactic markers between kindergarten, first grade, and second grade (Craig & Washington, 2004; van Hofwegen, & Wolfram, 2010).

Changes in the overt production of variable structures across early elementary school has been associated with increased language ability and literacy outcomes (Craig & Washington, 2004; Terry et al., 2012; Washington et al., 2018). For example, second-grade students who reduce their use of NMAE features in school settings, such as on a standardized dialect screener, demonstrate higher scores on standardized measures of oral language and reading comprehension (Terry et al., 2016). Given the changes to the use of NMAE dialect features across elementary school, it is necessary to directly investigate language use by school-age children with and without language disorders who speak different varieties of English.

This study builds on previous research through an examination of the production of morphosyntactic markers—including contrastive and noncontrastive features—by children who speak NMAE dialects. In doing so, we extend the previous research on younger children to an important age group: first- and second-grade students. We examine the production of three morphosyntactic features (PT marking, 3SG –s marking, and regular plural marking) using a cloze task production probe. In line with previous research, we hypothesized that, when compared to children who speak MAE, children who speak NMAE dialects will overtly produce morphosyntactic markers less often for contrastive features (PT and 3SG), but will not differ in their production of noncontrastive features (plural marking). Extending research from younger children, we expect that, when students with DLD who speak NMAE dialects are compared to same-dialect peers with TL skills, students with DLD will overtly produce morphosyntactic features less often.

Method

All study procedures were approved by the University of South Carolina Institutional Review Board. Participants were recruited from first- and second-grade classrooms from four elementary schools in one school district in South Carolina. All students in each classroom were invited to participate in the current study. Parents provided written consent and completed a questionnaire that included demographic information and asked about the student's history of language and literacy impairments. In order to identify children who may have DLD, participants completed a classroom-wide language screen (see Hendricks et al., 2019, for full screening methodology). Testing was prioritized to maximize the likelihood of students with DLD being included in the sample, with all students who scored below the bottom 33rd percentile on the language screen tested. Students who scored above the 33rd percentile were assigned a random number to determine the testing priority. Testing continued until the end of the school year.

Participants

Eighty-two ¹ first- and second-grade students completed a battery of standardized and experimental measures of language and reading ability and general cognition. Students were separated into one of three groups based on language ability and dialect use. Table 1 presents demographic information for students in each group. Overall, the sample was racially diverse: 53.4% of parents reported that their child was White, 35.2% reported that their child was Black/African American, 3.3% reported that their child was two or more races, and 8% did not report race information; 67% of parents reported that their child was not Hispanic/Latino, 2.3% reported that their child was Hispanic/Latino, and 30.7% did not report ethnicity. No families reported that their child heard or spoke any languages other than English at home. No families reported any motor disabilities, but one family noted an “other physical/medical problem” and noted that the child had previously received speech therapy. Two families reported that their child had hearing loss, but we confirmed with teachers that the children were not receiving hearing accommodations in school (e.g., hearing aids, classroom amplification), and thus, testing was conducted in a similar setting as their normal classroom activities. Fifteen families reported that their child had attention-deficit disorder/attention-deficit/hyperactivity disorder. ² Parents reported whether their child had ever received speech, language, reading, or other special educational services. Some students in the TL groups had received educational services (20.5% in TL-MAE and 13.6% in TL-NMAE), and services were often limited to services for articulation. Although more students in the DLD group had received services, more than half of the students in the DLD-NMAE group had not received educational services in the past (for additional detail regarding parent reports of concern and educational services, see Hendricks et al., 2019).

Table 1.

Demographic information by group.

Demographic category	Attribute	TL-MAE n = 44	TL-NMAE n = 22	DLD-NMAE n = 16
Demographic information	1st grade	36.4	18.2	43.8
	2nd grade	63.6	81.8	56.3
	Female	21	13	10
	Has ever received speech, language, reading, or other special education services	20.5	13.6	43.8
	ADD/ADHD	22.7	4.5	25.0
Race information	Race not reported	6.8	13.6	6.3
	White	77.3	13.6	31.3
	Black/African American	11.4	72.7	56.3
	2 or more races	4.5	0	6.3
Ethnicity information	Ethnicity not reported	20.5	45.5	37.5
	Hispanic/Latino	2.3	4.5	0
	Not Hispanic/Latino	77.3	50.0	62.5
Maternal education	Maternal education not reported	2.3	27.3	6.3
	Less than high school	0	0	31.3
	High school diploma/GED	20.5	31.8	56.3
	Some College	36.3	18.2	0
	Associate or technical	11.4	13.6	0
	Bachelor	15.9	4.5	0
	Master or higher	13.6	4.5	6.3

Note. TL-MAE = children with typical language who speak mainstream American English; TL-NMAE = children with typical language who speak nonmainstream American English; DLD-NMAE = children with developmental language disorder who speak nonmainstream American English; ADD = attention deficit disorder; ADHD = attention-deficit/hyperactivity disorder; GED = General Educational Development.

Measures

Language Variation

Part I of the Diagnostic Evaluation of Language Variation–Screening Test (DELV-ST; Seymour et al., 2003) was administered as a measure of the level of variation in the child's dialect use. The DELV-ST Part I includes items probing both phonological and morphosyntactic aspects of dialectal variation, including regular and irregular 3SG, nonagreeing don't (e.g., She don't like it.). Responses on Part I of the DELV-ST are categorized according to whether or not they contain an NMAE feature. Responses that do not contain the elicited grammatical form are excluded from analysis. The DELV-ST categorizes student dialect use into three categories: MAE, some variation from MAE, and strong variation from MAE. The “some variation” and “strong variation” categories were combined to create one NMAE dialect category. The DELV-ST does not provide information as to which NMAE dialect a student speaks; however, given the varieties spoken in rural South Carolina, participants who spoke an NMAE dialect were considered to speak either AAE or Southern White English (SWE). ³ Part II of the DELV-ST is a screen of language ability, which was administered to students, but the results were not analyzed because the DELV-NR was used as a full norm-referenced measure of language ability.

Language Ability

The DELV-NR was administered as a measure of language ability. The DELV-NR is an omnibus assessment of language ability and includes three subtests (syntax, pragmatics, and semantics) and assesses both language production and comprehension. The DELV-NR is designed to provide high diagnostic accuracy for students who speak MAE and NMAE dialects and focuses on assessing the acquisition of noncontrastive features, such as negation and passive forms. The DELV-NR is the only norm-referenced language assessment that assesses only noncontrastive features, allowing for the measurement of language impairment independently from language variation. According to the test manual, using a 1 SD below the mean cutoff for diagnosing language impairment, the DELV-NR has a sensitivity of .95 and a specificity of .93. Thus, in the current study, students who scored below 1 SD below the mean on the DELV-NR were classified in the DLD group.

Morphosyntax Production Task

The morphosyntax production task assessed three types of morphosyntax: (a) regular 3SG –s marking (3SG marking), (b) regular PT –ed marking (PT marking), and (c) regular plural –s marking (PL marking). In NMAE dialects such as AAE and SWE, PL marking is most likely to be overtly marked, followed by PT marking and then 3SG marking. PT marking and 3SG marking are less likely to be overtly marked in NMAE dialects and were considered contrastive features because they are variable in NMAE dialects but obligatory in MAE. PL marking was considered noncontrastive because it is generally overtly marked in NMAE dialects and obligatory in MAE. ⁴

The task included items from each of three allomorphs for each morphosyntactic feature. Items were selected to be similar in overall frequency and concreteness, although nouns were overall more concrete than verbs (e.g., higher concreteness ratings for nouns such as bed compared to lower ratings for verbs such as wait). Concreteness ratings and ratings of adult familiarity with the items were determined using a published database of adult ratings of concreteness, and the SUBLEX_US database was used to determine the frequency of each item (Brysbaert et al., 2014). One-way analyses of variance comparing familiarity, frequency, and concreteness ratings in the three categories revealed that items did not differ in familiarity or frequency across features (percent known: F(2, 36) = 2.20, p = .127; word frequency: F(2, 36) = .832, p = .444); nouns in the plural feature had higher concreteness ratings than the verbs in both the PT and 3SG features (p < .001). The difference in concreteness ratings between the PT and 3SG items was not significant (p = .759).

Participants were asked to complete cloze sentences, which elicited the target nouns and verbs. The Appendix presents a list of all items. Research assistants described a drawing on a PowerPoint slide, and the child completed the sentence. PT items were elicited using the semantic marker, “yesterday” (e.g., Today the boy is painting. Yesterday he did the same thing. Yesterday he ____.). 3SG items were elicited using the habitual present (e.g., This boy is eating pizza. Every day he ____.). Finally, PL items were elicited by presenting the child with a picture of one object, followed by a picture of three identical objects, and the prompt: Here is one pen. Now there are three. There are some ____. As PL marking was intended as a noncontrastive feature, all items included “some” in the prompt to reduce the potential variability noted with numerals such as “three.” Responses were noted by the research assistant at the time of the administration and audio-recorded for off-line scoring and reliability. Research assistants unfamiliar with participant group membership listened to audio recordings of the sessions and orthographically transcribed the participants' responses. The number of responses including an overt morphosyntactic marker—including responses without the expected vocabulary item—was counted for each participant and feature.

Nonverbal Intelligence

The Test of Nonverbal Intelligence–Fourth Edition (TONI-4; Brown et al., 2010) was administered to students as a measure of nonverbal cognition and was analyzed for descriptive purposes only.

Word Reading

Because many children with DLD may also have significant reading difficulties, students completed the Word Identification subtest of the Woodcock Reading Mastery Test–Third Edition (WRMT-III; Woodcock, 2011) for descriptive purposes. The WRMT-III Word Identification assesses children's ability to read real words of increasing difficulty. Items were scored as either correct or incorrect, and incorrect responses were transcribed. For students who speak NMAE dialects, the authors reviewed each protocol and any variations in pronunciation that may be attributed to use of an NMAE dialect were rescored as correct. ⁵ Neither nonverbal intelligence or word reading was used to determine group membership.

Scoring Reliability

All scorers were trained and required to pass a scoring test before they were allowed to score assessments independently. A random sample of at least 20% of each assessment was double scored for reliability. The total number of assessments selected for reliability scoring was based on the number of participants in a larger study (see Hendricks et al., 2019). Reliability scoring was completed on blank protocols using the video and audio recordings, and reliability scorers were unaware of the participant's initial scores. Reliability was determined through item-by-item comparisons between the initial score and the reliability score. Scoring reliability was measured by the percentage of items in which the initial scorer and the reliability scorer agreed. Scoring reliability was 95% for Part I of the DELV-ST, 99.6% for the TONI-4, 96.7% for the WRMT-III Word Identification, 96.1% for the DELV-NR, and 96.5% for the morphosyntax production task.

Results

Descriptive results are presented in Table 2. As planned, students in the two TL groups did not differ in their language ability as measured by DELV-NR (p = .159, Hedges' g = 0.489), and students in the DLD-NMAE group scored significantly lower in language ability from students in both TL groups (ps < .001, TL-MAE vs. DLD-NMAE Hedges' g = 2.843, TL-NMAE vs. DLD-NMAE Hedges' g = 2.831). Students in the MAE dialect group used NMAE features significantly less often on the DELV-ST compared to students in both NMAE groups (ps < .001, TL-MAE vs. TL-NMAE Hedges' g = 3.180, TL-MAE vs. DLD-NMAE Hedges g = 3.321), but students in NMAE groups used similar levels of NMAE features on the DELV-ST regardless of language ability status (p = .933, Hedges' g = 0.088). Students in the TL groups did not differ in nonverbal intelligence (p = .882, Hedges' g = 0.122), but the students in the DLD-NMAE group scored significantly lower than students with TL from both dialect groups (p ≤ .001, DLD-NMAE vs. TL-MAE Hedges' g = 1.136, DLD-NMAE vs. TL-NMAE Hedges' g = 1.463). ⁶ A different pattern emerged for word reading, ⁷ with a significant difference across groups, F(2, 74) = 20.63, p < .001. Post hoc comparisons indicated that, despite having similar language ability scores, students in the TL-NMAE group scored significantly lower on word reading compared to their peers with TL who speak MAE (p < .001, Hedges' g = 1.051). Students in the DLD-NMAE group scored significantly lower in word reading than the TL-MAE group (p < .001, Hedges' g = 1.906). Students who differed in language ability but who spoke the same dialect did not differ significantly in word reading (p = .113, Hedges' g = 0.718).

Table 2.

Results of descriptive measures by group.

Group	WRMT-III Word Identification M (SD)	TONI-4 M (SD)	DELV-NR Total Language M (SD)	DELV-ST % NMAE M (SD)
TL-MAE	106.05 (14.35)	103.20 (8.47)	99.13 (9.60)	12.4 (9.4)
TL-NMAE	90.40 (16.03)	104.19 (7.23)	94.68 (7.99)	58.9 (21.7)
DLD-NMAE	80.43 (9.97)	94.13 (6.38)	73.93 (6.29)	60.9 (23.9)

Note. WRMT-III: TL-MAE, n = 43; TL-NMAE, n = 20; DLD-NMAE, n = 14. TONI-4: TL-MAE, n = 44; TL-NMAE, n = 21; DLD-NMAE, n = 16. WRMT-III = Woodcock Reading Mastery Test–Third Edition; TONI-4 = Test of Nonverbal Intelligence–Fourth Edition; DELV-NR = Diagnostic Evaluation of Language Variation–Norm Referenced; DELV-ST = Diagnostic Evaluation of Language Variation–Screening Test; NMAE = nonmainstream American English; TL-MAE = children with typical language who speak mainstream American English; TL-NMAE = children with typical language who speak nonmainstream American English; DLD-NMAE = children with developmental language disorder who speak nonmainstream American English.

We next compared student performance on the morphosyntax production task. Overall, students were familiar with the selected vocabulary items, producing the expected word for over 90% of responses across all three categories (PT: 92%, 3SG: 94%, PL: 99%). The distribution of the data across groups and morphosyntactic categories is presented as pirate plots (Phillips, 2017) in Figures 1 –3. Pirate plots display the mean, distribution, shape, and confidence interval of a data set, while also presenting the individual data points in a data set (Phillips, 2017). Because some studies have suggested that the use of NMAE features changes across elementary school grades (Craig & Washington, 2004), grade level (first vs. second grade) was entered as a covariate in the analyses for both research questions. Table 3 presents the descriptive results of the percentage of overt productions for each morphosyntactic feature in each group of participants.

Figure 1.

Pirate plot of overt production of third-person singular (3SG) –s marking by group. TL-MAE = children with typical language who speak mainstream American English; TL-NMAE = children with typical language who speak nonmainstream American English; DLD-NMAE = children with developmental language disorder who speak nonmainstream American English.

Figure 2.

Pirate plot of overt production of regular past tense marking by group. TL-MAE = children with typical language who speak mainstream American English; TL-NMAE = children with typical language who speak nonmainstream American English; DLD-NMAE = children with developmental language disorder who speak nonmainstream American English.

Figure 3.

Pirate plot of overt production of regular plural marking by group. TL-MAE = children with typical language who speak mainstream American English; TL-NMAE = children with typical language who speak nonmainstream American English; DLD-NMAE = children with developmental language disorder who speak nonmainstream American English.

Table 3.

Descriptive results of production probe results by group.

Group	Third-person singular items M (SD)	Past tense items M (SD)	Plural items M (SD)
TL-MAE, n = 44	96.7 (5.8)	84.5 (13.1)	97.5 (9.3)
TL-NMAE, n = 22	75.2 (21.3)	65.6 (27.9)	97.3 (6.5)
DLD-NMAE, n = 16	57.6 (37.3)	44.6 (21.3)	86.5 (16.1)

Note. TL-MAE = children with typical language who speak mainstream American English; TL-NMAE = children with typical language who speak nonmainstream American English; DLD-NMAE = children with developmental language disorder who speak nonmainstream American English.

For each of the three morphosyntactic features (PT, 3SG, PL), we compared the number of responses including an overt morphosyntactic marker by students in each group. Generalized linear models were used to compare differences in the number of overt productions for each feature. All responses without the target forms (e.g., “skating,” “riding on skates,” “did the same thing”) were marked as unscorable and excluded. The total number of scorable items was entered into the model as the total number of events. Across groups, the number of scorable responses was high for each morphosyntactic feature (TL-MAE = 90%–98%, TL-NMAE = 87%–99.5%, DLD-NMAE = 86%–100%), and the number of scorable responses did not differ by group for any of the morphosyntactic features: 3SG, F(2, 81) < 1, p > .05; PT, F(2, 81) < 1, p > .05; PL, F(2, 81) < 1, p > .05.

We first asked whether typically developing students who speak different dialects of American English differ in how often they overtly mark morphosyntactic features. We compared the number of overt productions of three morphosyntactic features between typically developing students who speak MAE (TL-MAE, n = 44) and typically developing students who speak NMAE dialects (TL-NMAE, n = 22). The results indicated that students with TL skills who speak NMAE dialects overtly produce morphosyntactic markers less often than their peers with TL abilities who speak MAE. This finding was consistent for both contrastive features: PT marking and 3SG marking (PT: χ² = 26.1, p < .001; 3SG: χ² = 62.97, p < .001). ⁸ The effect sizes of differences between TL-MAE and TL-NMAE were large (3SG: Hedges' g = 1.64, PT: Hedges' g = 0.98). However, students who speak different dialects did not differ in their production of noncontrastive PL marking (χ² = .013, p = .909).

Next, we asked whether students who speak NMAE dialects who have DLD differ in their production of morphosyntactic features compared to their same-dialect peers with TL. Similar to our prior analysis, we compared the overt production of each morphosyntactic marker across students who speak NMAE but who have DLD (DLD-NMAE, n = 16) to students who speak NMAE but have TL abilities (TL-NMAE, n = 22). For both contrastive and noncontrastive features, the students with DLD overtly produced morphosyntactic markers less often than their same-dialect peers with TL skills (PT: χ² = 17.09, p < .001; 3SG: χ²= 14.54, p < .001; PL: χ² = 19.86, p < .001). ⁹ Across all morphosyntactic markers, the effect sizes were medium to large (3SG: Hedges' g = 0.61, PT: Hedges' g = 0.82, PL: Hedges' g = 0.94). Taken together, the results of both analyses indicate that, although children with TL skills who speak different dialects differ in their production of contrastive morphosyntactic features, a within-dialect comparison reveals significant differences in the production of both contrastive and noncontrastive features by students who speak the same dialect but differ in their language ability skills.

Discussion

This study explored how students who speak different dialects of English produce three morphosyntactic features, which have been commonly used in assessments of language ability for school-age children. A comparison of typically developing children who speak MAE and NMAE dialects demonstrated that children who speak MAE overtly produce morphosyntactic markers more often than students who speak NMAE dialects. At the same time, when comparing within students who speak NMAE dialects, students with DLD overtly produce morphosyntactic markers less often than their peers who have TL skills.

The current data support previous research demonstrating that some morphosyntactic features are produced differently across MAE and NMAE dialects and that there are significant differences between the frequency with which children with TL who speak different dialects overtly produce contrastive morphosyntactic features (Seymour et al., 1998). At the same time, there were no significant differences in how children with TL who speak different dialects produce plural marking, which is considered a noncontrastive feature when marked with a quantifier rather than a numeral (“some” vs. “three”). These data indicate that dialectal differences are present in early elementary school, even though research has shown that students with TL tend to use NMAE features less often over the course of elementary school (Craig & Washington, 2004; van Hofwegen & Wolfram, 2010). While the current study was not designed to test for differences between features, it was noteworthy that students did not appear to demonstrate a difference in overt marking of PT and 3SG, in contrast to previous studies that suggest that students overtly mark PT significantly more often than 3SG (see Lee & Oetting, 2014). Future research should consider differences in the linguistic properties of the individual lexical items (e.g., event structure, frequency) and other sociolinguistic variables (e.g., speaker age, conversational context) to determine how they influence the overt production of PT and 3SG.

Differences in the rates of overt production of morphosyntactic features underscore the importance of using an appropriate comparison group for accurate assessments. That is, comparing children who speak NMAE dialects to norms established based on productions by children who speak MAE runs the risk that children with TL who speak NMAE dialects may be erroneously flagged for further assessment or intervention services. Inappropriate comparisons to TL-MAE group norms could contribute to an overidentification of language disorders in children who speak NMAE dialects. Inaccurate diagnostic assessments may cause limited clinician resources to be spent on unnecessary assessments or intervention services. This limits the ability of these clinicians to provide services for other children who may, in fact, require services. There are consequences for students as well, including time away from instruction and increased social stigma surrounding the use of NMAE dialects.

These findings are also in line with previous research, which has shown that younger children with DLD—in preschool and kindergarten—do not overtly produce morphosyntactic features as often as their peers with TL skills. In comparison to TL-NMAE students, their peers who speak NMAE dialects and meet criteria for DLD overtly produced morphosyntactic markers less often, including contrastive and noncontrastive features. This pattern of results suggests that, while assessing a broad set of morphosyntactic features may provide useful information about the language abilities of children who speak NMAE dialects, it is important that comparisons be made to their same-dialect peers with TL ability. These data support the disorder within diversity framework put forth by Oetting and colleagues (Oetting, 2018; Oetting et al., 2016). The results of the current study suggest that the framework can be extended to school-age children, despite substantial changes to the use of NMAE features over the course of elementary school.

Furthermore, to date, the research on the disorder within diversity framework has been conducted within a limited number of geographical regions. The current study was conducted in South Carolina, which lends preliminary support to the claim that the disorder within diversity framework can be extended to other geographical regions. Although there has been an attempt in recent years to expand research on child language to other varieties of English (Berry & Oetting, 2017; Rivière et al., 2018), there continues to be limited information about the production patterns of children who speak other NMAE dialects, such as Appalachian English or Chicano English. In order to validate the disorder within diversity framework, additional research is needed from different dialect groups, geographic regions, and age groups.

In contrast to previous research, the results of this study suggest that contrastive features may prove useful for assessments of language ability in children who speak NMAE dialects. The current study was not intended to evaluate the ability for these morphosyntactic features to make reliable diagnostic classification decisions. Furthermore, as illustrated in Figures 1 –3, there was considerable overlap between the production rates across dialect and impairment groups. This suggests that, despite significant group differences, these items alone are not sufficient for classifying students as having a language impairment. With additional research, however, future language assessments may be able to draw on a larger range of morphosyntactic features than has been thought in the past. Critically, the expansion of features included in language assessments will rely on a better understanding of the uses of these features by children with TL who speak NMAE dialects.

It is important to keep in mind that there is variability in the extent to which individuals' dialects differ from MAE. That is, speakers differ in the number of NMAE features they utilize and the frequency with which they produce NMAE features (Washington & Craig, 2002). Furthermore, use of NMAE dialects is sensitive to a number of sociolinguistic factors, such as the level of variation in children's input and the formality of the context, and linguistic factors, such as the phonological form (Pruitt & Oetting, 2009) and argument and event structure. The results of this study underscore the need for more in-depth understanding about typical productions by children who speak NMAE dialects, including not only how often they overly mark morphosyntactic features but also the factors that influence within speaker variability. Ultimately, assessment practices may be created to capitalize on these differences to maximize the informativeness of assessments.

Beyond the question of improving diagnostic accuracy, understanding children's production of a larger set of features can provide useful information for language interventions, such as selecting treatment targets. Often even after a clinician has determined that a child who speak a NMAE dialect has a language disorder, clinicians report that contrastive features are excluded from treatment on the basis that they are produced differently in MAE (Hendricks & Diehm, 2020). However, the overt production of these features has been associated with stronger language and reading abilities (Craig & Washington, 2004; Terry et al., 2012; Washington et al., 2018). Thus, it is important to consider whether including contrastive features alongside noncontrastive features in language instruction may improve both language and reading outcomes for children with language disorders. It should be noted that the goal of introducing contrastive features is not to increase the use of MAE but rather to enable students with DLD to use morphosyntactic features in similar ways to their peers with TL who speak NMAE dialects. Dialect awareness programs represent one way of drawing students' attention to differences in dialect use across context while also validating and appreciating their dialect use (Johnson et al., 2017).

It is interesting to note that students who speak NMAE dialects and have language scores within normal limits scored lower in word reading than their peers who speak MAE. Furthermore, the students who speak NMAE dialects and have language scores within normal limits did not score significantly higher on word reading compared to their same-dialect peers with lower language ability, although the effect size for the group difference was moderate (Hedges' g = 0.718). We also note that these differences were observed despite the use of modified scoring, which gave credit for productions that could be attributed to dialectical variation. These data are in line with previous studies, which have documented a relationship between use of NMAE features and language and literacy outcomes (Terry et al., 2012, 2010) as well as the persistent reading achievement gap between African American children and their White peers at Grades 4 and 8 (National Assessment of Education Progress, 2017). From these data, it is not possible to determine the cause of the group differences in word reading. Future research is needed to determine whether lower dialect awareness may contribute to these group differences (Johnson et al., 2017) or whether instructional changes are needed to meet the needs of students who speak NMAE dialects, including those with and without language disorders.

The current study documented differences in the frequency of overt productions of morphosyntactic features across children who speak NMAE dialects with different language ability levels. However, the current data did not allow for investigating the potential causes for these differences in production patterns. Some have suggested that children with DLD may have lower metalinguistic skills leading to an inability to switch their use of dialect features as appropriate. Another possibility is that, given their difficulty with language acquisition, children with DLD may not have fully acquired the same sociolinguistic conditioning factors as typically developing peers. That is, while typically developing children who speak AAE may be sensitive to linguistic factors such as conversational context, children with language disorders may not be as sensitive to linguistic conditioning factors. For example, linguistic descriptions of AAE note that, while plural marking is obligatory in most contexts, it can be variably omitted when a numeral is present, but not when a quantifier , such as “some,” is present. However, it is possible that children with DLD may not be as sensitive to this complex conditioning. If that was the case, then children with DLD may treat noncontrastive features with a neighboring contrastive feature more like a contrastive feature. The current data set did not manipulate this, but the lower accuracy in the DLD-NMAE group on plural markings might suggest that this is an area for future investigation. While the current data were not able to explore the underlying causes of the group differences, future studies might compare children's comprehension of morphosyntax as a means to better understanding what may contribute to the differences seen in production.

Conclusion

Extending previous research on younger children, the results demonstrated that school-age children with language disorders produce morphosyntactic features differently when compared to their same-dialect peers with TL skills. These results underscore the importance of considering a child's dialect when assessing language ability with measures of morphosyntax that include contrastive features. If a child who speaks NMAE is compared to children who speak MAE, a difference in production rates may inaccurately be interpreted as problematic, potentially leading to an overdiagnosis of DLD. However, within children who speak NMAE dialects, there was also a difference between the children with TL and those with DLD, both for contrastive and noncontrastive features. Therefore, a comparison within the child's dialect may provide useful information for assessments of language ability, even when features are produced differently across dialects of English.

Appendix

Stimulus Production Probes for Morphosyntax Production Task

Example 1. Look! Here's a sun. Now there are three of them. Here are some SUNS.

Example 2. Today the boy is walking the dog. Yesterday he did the same thing. Yesterday he WALKED.

Example 3. Here the girl is reading a book. Everyday she does the same thing. Everyday she READS.

Item	Condition	Allomorph	Item	Stimulus
1	Past tense	t	kick	Today the boy is kicking the rock. Yesterday he did the same thing. Yesterday he ____.
2	Past tense	t	dance	Today the girl is dancing. Yesterday she did the same thing. Yesterday she ____.
3	Past tense	t	kiss	Today the boy is kissing his mom. Yesterday he did the same thing. Yesterday he ____.
4	Past tense	t	cook	Today the mom is cooking dinner. Yesterday she did the same thing. Yesterday she ____.
5	Past tense	d	smell	Today the boy smelled the cookies. Yesterday he did the same thing. Yesterday he ____.
6	Past tense	d	push	Today the dad is pushing the car. Yesterday he did the same thing. Yesterday he ____.
7	Past tense	d	jump	Today the girl is jumping rope. Yesterday she did the same thing. Yesterday she ____.
8	Past tense	d	carry	Today the mom is carrying the box. Yesterday she did the same thing. Yesterday she ____.
9	Past tense	ed	paint	Today the girl is painting the wall. Yesterday she did the same thing. Yesterday she ____.
10	Past tense	ed	lift	Today the boy is lifting the bucket. Yesterday he did the same thing. Yesterday he ____.
11	Past tense	ed	wait	Today the girl is waiting for the bus. Yesterday she did the same thing. Yesterday she ____.
12	Past tense	ed	skate	Today the boy is skating at the roller rink. Yesterday he did the same thing. Yesterday he ____.
13	3SG	s	talk	Here the girl is talking. Everyday she does the same thing. Everyday she ____.
14	3SG	s	sit	Here the girl is sitting on the bus. Everyday she does the same thing. Everyday she ____.
15	3SG	s	eat	Here the boy is eating a hotdog. Everyday he does the same thing. Everyday he ____.
16	3SG	s	pick	Here the girl is picking berries. Everyday she does the same thing. Everyday she ____.
17	3SG	z	run	Here the girl is running. Everyday she does the same thing. Everyday she ____.
18	3SG	z	hang	Here the boy is hanging a picture. Everyday he does the same thing. Everyday he ____.
19	3SG	z	throw	Here the boy is throwing a ball. Everyday he does the same thing. Everyday he ____.
20	3SG	z	sing	Here the girl is singing. Everyday she does the same thing. Everyday she ____.
21	3SG	ez	catch	Here the girl is catching the ball. Everyday she does the same thing. Everyday she ____.
22	3SG	ez	watch	Here the boy is watching TV. Everyday he does the same thing. Everyday he ____.
23	3SG	ez	brush	Here the boy is brushing his teeth. Everyday he does the same thing. Everyday he ____.
24	3SG	ez	wash	Here the girl is washing her hair. Everyday she does the same thing. Everyday she ____.
25	Plural	z	baby	Look! Here's a baby. Here there are three of them. Here are three____.
26	Plural	z	bed	Look! Here's a bed. Here there are three of them. Here are three____.
27	Plural	z	bird	Look! Here's a bird. Here there are three of them. Here are three____.
28	Plural	s	boat	Look! Here's a boat. Here there are three of them. Here are three____.
29	Plural	s	book	Look! Here's a book. Here there are three of them. Here are three____.
30	Plural	ez	bus	Look! Here's a bus. Here there are three of them. Here are three____.
31	Plural	s	clock	Look! Here's a clock. Here there are three of them. Here are three____.
32	Plural	s	cup	Look! Here's a cup. Here there are three of them. Here are three____.
33	Plural	ez	dress	Look! Here's a dress. Here there are three of them. Here are three____.
34	Plural	ez	horse	Look! Here's a horse. Here there are three of them. Here are three____.
35	Plural	ez	house	Look! Here's a house. Here there are three of them. Here are three____.
36	Plural	z	table	Look! Here's a table. Here there are three of them. Here are three____.

Note. 3SG = third-person singular.

Funding Statement

This project was supported in part by grants from the University of South Carolina Vice President for Research Advanced Support for Innovative Research Excellence (PI: Hendricks) and the National Institutes of Health (Grant R03DC013399; PI: Adlof).