Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2014 Jun 1.
Published in final edited form as: J Speech Lang Hear Res. 2012 Dec 28;56(3):933–944. doi: 10.1044/1092-4388(2012/12-0099)

Linguistic Constraints on Children's Overt Marking of BE by Dialect and Age

Joseph Roy 1, Janna B Oetting 2, Christy Wynn Moland 2
PMCID: PMC3700578  NIHMSID: NIHMS481421  PMID: 23275400

Abstract

Purpose

Overt marking of BE in nonmainstream adult dialects of English is influenced by a number of linguistic constraints, including the structure's person, number, tense, contractibility, and grammatical function. In the current study, we examined the effects of these constraints on overt marking of BE in children as a function of their nonmainstream English dialect and age.

Methods

The data were language samples from 62 children, aged four to six years; 24 spoke African American English (AAE) and 38 spoke Southern White English (SWE). Analyses included analysis of variance and logistic regression.

Results

Rates of overt marking varied by the children's dialect but not their age. Although the person, number, tense, and grammatical function of BE influenced the children's rates of marking, the nature and magnitude of the influence differed by the children's dialect. For AAE-speaking children, contractibility also influenced their marking of BE.

Conclusions

Consistent with the adult literature, the AAE- and SWE-speaking children marked BE in ways that differed from each other and from what has been documented for child speakers of Mainstream American English. These findings show stability in the use of BE in AAE and SWE that spans different generations and different dialect communities.

Adult African American English (AAE) and Southern White English (SWE) differ from Mainstream American English (MAE) in the production of copular and auxiliary forms of BE. Unlike MAE, overt marking of BE is not obligatory for adults who speak AAE or SWE. In fact, it is perfectly acceptable in these two dialects for both overtly marked (e.g., he is happy) and zero marked BE (he Ø happy) to be produced. However, overt versus zero marking of BE does not occur randomly. Instead, the type of marking produced is probabilistic and tied to the linguistic characteristics of the context surrounding the BE form.

Effects of linguistic contexts on AAE and SWE speakers' overt marking of BE are commonly referred to as linguistic constraints because various contexts are thought to constrain (or influence) when a speaker produces or doesn't produce an overtly marked form (for review of constraint literature and earlier work by Labov and others, see Sankoff, Taliamonte, & Smith, 2005). Linguistic constraints have been documented in a number of studies involving adult speakers of AAE and to a much lesser extent SWE (Bailey & Maynor, 1985; Baugh, 1980; Blake, 1997; Childs & Malinson, 2004; Labov, 1969; 1972; Romaine, 1982; Rickford, Ball, Blake, Jackson, & Martin, 1991; Rickford, 1998; Walker, 2000; Winford, 1992; Wolfram, 1969; 1974). The presence of some linguistic constraints on children's overt marking of BE has also been documented in a few studies. However, these studies have often focused on either the copular or auxiliary BE form, and the analyses have either been highly descriptive or limited to analyses of variance. Given this, the goal of the current study was to extend the study of children's overt marking of BE by including copular and auxiliary forms, speakers of two nonmainstream dialects of English (AAE and SWE) and two age groups (four and six years), and by utilizing two types of statistical analyses, analysis of variance and logistic regression.

Although previous adult studies have examined a variety of linguistic constraints, the current work focuses on three: the person, number, and tense of the BE form, the contractibility of the BE form, and the grammatical function of the BE form.1 These constraints were chosen to maximize the number of BE tokens available for the analyses. Each of these linguistic constraints has also been studied by others with children who speak either AAE or MAE; however, they have not been examined together in a single study nor have they been used to compare different child dialects to each other. Studies of BE marking (and grammar in general) have also not been conducted on children who speak SWE, even though this dialect is spoken by many children who live in the rural South (Wolfram & Schilling-Estes, 1998), and the socioeconomic levels and the academic achievement of children who live in the rural South are lower than those of children who live in other regions of the United States (Harris & Zimmerman, 2003; Miller & Weber, 2004).

A detailed study of BE and the constraints that influence its use in child AAE and SWE is important because zero marking of this structure has been repeatedly shown to be more frequent than any other nonmainstream grammar structure within these dialects (for child AAE and SWE, see Oetting & Pruitt, 2005; for child AAE, see Horton-Ikard & Weismer, 2005; Jackson & Roberts, 2001; Washington & Craig, 1994). Speech-language clinicians who work with diverse groups of children in the rural South (and perhaps elsewhere) have likely heard overtly marked and zero marked forms of BE within their community, yet the field lacks empirical data from children to help clinicians understand and articulate when and why a child might produce an overtly marked form over a zero marked form and vice versa.

As background for the study, we operationally define the three linguistic constraints and review what is known about their effects on adults' marking of BE. Findings from the adult studies provide a baseline from which to consider the children's dialects. If the children's marking of BE is found to be similar to that of previous adult studies, we will have evidence of stability in BE across different generations of speakers and time. If differences are identified, we will have evidence of linguistic evolution and change.

Findings by Wolfram and Thomas (2002) indicate that modern day changes in language are likely to be structure specific and dialect specific. Their study focused on six grammar structures (i.e., weren't leveling, was leveling, verbal –s marking with third person plural subjects, zero is, zero are, and zero verbal –s) in the conversations of 49 adults (35 African American and 14 white) who resided in Hyde county, North Carolina. The African American adults reflected four age groups (14–23, 32–43, 55–70, and 77–102 years), and the white adults reflected two (15–27 and 77–94 years). Results showed that the African American and white groups presented varying degrees of change in the six grammar structures. For the African Americans, weren't leveling and verbal –s marking showed significant decreases in use across generations, zero is showed a significant increase, and was leveling and zero are showed minimal change. In comparison, the white participants showed a significant increase of use across generations in weren't leveling and a significant decrease in verbal –s marking with third person plural noun phrases; all other grammar structures showed minimal change. Dubois and Horvath (1998; 1999) have documented similar structure specific changes across generations in phonological patterns that have been historically tied to Cajun English in Louisiana. In their studies, some structures show a decrease in use from older to younger speakers, others show an increase, and still others show a v-shape pattern of change, with high rates of use in older and younger speakers, and low rates of use in middle-age speakers.

In our own work with the same children studied here, we have identified some structure specific and dialect specific changes through separate analyses of children's nonmainstream productions of verbal –s and relative clause markers (Cleveland, 2009; Oetting & Newkirk, 2011). For verbal –s marking, the data show stability across generations for both AAE and SWE when the children's data are compared to previous studies from adults. For relative clause markers, the data show a significant reduction of nonmainstream productions in child AAE when compared to previous adult AAE studies, with stable rates of nonmainstream productions in SWE. The current study allows for further examination of language evolution and change by focusing on children's marking of BE.

Following a review of the adult studies, we present findings from the child language literature. These studies are important for two reasons. First, although these studies have varied in the age of the child examined and in their methodologies, they suggest dialect differences between AAE, SWE, and MAE. In the current work, we directly compare child AAE to SWE while also considering MAE child studies that have employed methods similar to ours. Second, these previous studies highlight the developmental nature of child language and motivate the need to consider age effects within our data. Child studies that have examined age effects repeatedly show higher rates of overt marking of BE in five- and six-year-olds than in three- and four-year-olds, and in MAE, adult levels of BE marking are not expected until children are beyond the age of four years (Rice, Wexler, & Hershberger, 1998). In the current work, we examine the developmental nature of BE marking in AAE and SWE by examining children who are aged four and six years.

Within the literature review, we report all findings from previous studies as a function of the speakers' rates of overtly marked forms as opposed to their rates of zero marked forms. Although the field of sociolinguistics focuses on the zero marked forms, the field of child language focuses on the overtly marked forms. For those interested in rates of zero marking, percentages reported in this paper can be converted using the formula, 1 – the rate of overtly marked forms (e.g., 1 − 40% = 60%).

Another methodological difference across fields is in the way rates of marking are calculated. Within the adult sociolinguistics literature, rates of marking are often calculated by pooling data across speakers (e.g., Blake, 1997; Labov, 1969; Rickford, 1999; Rickford et al., 1991). Thus, rates reflect the proportion of time a particular form is produced out of all possible contexts available, and speakers who produce more data carry more weight in the calculation than those who produce less. In contrast, within the child language literature, rates are typically calculated for each child and then these individual rates are averaged to reflect a mean percentage (e.g., Cleave & Rice, 1997; Garrity & Oetting, 2010; Leonard et al., 2003; Rice et al., 1998). Thus, each child's data contributes to the mean in the same way regardless of the number of tokens produced. In the current work, we specify the method of calculation for each study reviewed as well as consider both methods within the analyses.

Linguistic Constraint: Person, Number, Tense

The first linguistic constraint involved the person, number, and tense of the BE form. As shown in 1–4 below, this variable included four levels based on the grammatical subject and temporal reference of the verb. For the study here, we combined BE tokens for was and were because there were low numbers of these forms in the data and predictions about rates of overt marking for these forms were the same. Inspection of the data also suggested similar findings for was and were.

  • 1

    Am: First person, singular, (temporal referent: present/habitual)

  • 2

    Is: Third person, singular, (temporal referent: present/habitual)

  • 3

    Are: Second person and first and third person plural, (temporal referent: present/habitual)

  • 4

    Was/were: (temporal referent: past, past counterfactual)

The adult dialect literature repeatedly shows that both AAE and SWE speakers are most likely to show variable marking (i.e., both overt marking and zero marking) of BE forms that involve are but rarely show variable marking of BE when the forms involve am, was, or were (Blake, 1997; Labov, 1969; Rickford, 1999; Rickford et al., 1991; Wolfram, 1974). For these latter three contexts, both AAE and SWE adult speakers produce such high rates of overt marking that Blake (1997) and others view these contexts as categorically non-variable and recommend that they be excluded from analyses.

The adult literature shows that AAE speakers also variably mark is in ways that are similar to are; however, rates of overt marking are typically higher for is than are. For example, in Rickford et al.'s (1991) study of is and are within AAE, is was overtly marked 47% of the time compared to are, which was overtly marked 22% of the time. These calculations pooled data from 30 speakers and were proportional in nature.

For adults who speak SWE, the literature is less clear. Labov (1969) and Fasold and Wolfram (1970) claim that white southerners do not zero mark is in the variable manner that has been documented for varieties of AAE, but Wolfram (1974), Cukor-Avila (2001), and Hazen (2001) present data from southern white speakers which show some variable marking of is. In all cases, however, SWE-speaking adults have been shown to overtly mark is (and even are) at relatively high rates. For example, rates of overt marking by Hazen's (2001) white speakers from North Carolina were 96% for is and 84% for are. As evidenced by these data, the magnitude of the constraint effect in adult SWE appears minimal, but the direction of the effect (is > are) is consistent with the adult AAE literature.

Linguistic Constraint: Contractibility

The second constraint was the contractibility of the BE context. To be consistent with other child language studies (e.g., Cleave & Rice, 1997; de Villiers & de Villiers, 1973; James & Kahn, 1982; Johnston & Schery, 1976), we followed Brown (1973) and coded contexts as contractible if contractibility was possible, regardless of whether the child actually contracted the BE form. Brown's rationale for this decision was based on the difficulty of coding contractibility in a reliable manner across examiners when the data are samples of children's conversational speech. Thus, utterances such as, They are walking and They're walking were coded as contractible, and utterances such as Sis is walking and Yes she is were coded as uncontractible for phonetic and/or syntactic reasons. In the former example, Sis and is cannot be contracted and produced as a single syllable for phonetic reasons; in the latter example, she and is cannot be contracted and produced as a single syllable for syntactic reasons. When children zero marked a BE form, the contractibility of the context was determined based on the contractibility of the structure had it been overtly marked. Thus, Now they Ø walking was coded as contractible and This Ø being put right here was coded as uncontractible.

The adult dialect literature is inconsistent in the coding of contractibility and zero marking of BE. In fact, since Labov's (1969) seminal study of BE, sociolinguists have debated not only the processes and order by which speakers contract and zero mark various forms of BE but also the formulas used to calculate rates of contraction and zero marking within and across various speaker groups (Blake, 1997; Rickford et al., 1991). As an example, Rickford et al. (1991) identifies three different calculations for determining the rates at which a speaker contracts BE and two different calculations for determining when a speaker zero marks BE. Although the effect of contractibility on BE marking (overt vs. zero) has not been the focus of the adult studies, Labov, Rickford, and others have shown that for AAE, zero marking is most likely to occur in contexts that are contracted. Given this, overt marking in adult AAE is more likely to occur in uncontractible contexts as compared to contractible contexts.

In most studies of SWE, the effect of contractibility on speakers' overt marking of BE has not been examined, and presumably this is because rates of overt marking across contractible and uncontractible contexts are high and invariable. Data by Hazen (2001) is consistent with this claim because in his study, SWE speakers did not vary their overt marking of BE by the contractibility of the form.

Linguistic Constraint: Grammatical Function

The third constraint was the grammatical function of the BE structure, and this variable involved two levels to represent copular functions (e.g., T'Lane is a girl) and auxiliary functions (e.g., T'Lane is walking). Within the adult literature, grammatical function has not been studied in this way. Instead, rates of overt marking have been examined in five different contexts which vary in the content that follows the BE form. As shown in 5–9 below, three of these contexts reflect a copular BE function and two reflect an auxiliary function.

  • 5

    The book is on the table: Locative context following BE (copular)

  • 6

    The book is brown: Adjective context following BE (copular)

  • 7

    Her name is T'Lane; Noun context following BE (copular)

  • 8

    T'Lane is gonna read it: Gonna following BE (auxiliary)

  • 9

    T'Lane is going to the store: Verb + ing following BE (auxiliary)

In the adult AAE literature, rates of overt marking have been documented to be higher in copular contexts than in auxiliary contexts. Again using AAE data from Rickford et al. (1991) as an example, factor weights which were derived from proportional data were above .50 for all three copular contexts and below .40 for the two auxiliary contexts. Similarly, factor weights provided by Wolfram and Thomas (2002) for African Americans were higher for the three copular contexts (> .46) than for the two auxiliary contexts (.24 and .35). Across both of these AAE adult studies, the factor weights indicate higher rates of overt marking in copular contexts as compared to auxiliary contexts. Limited data exist for adult speakers of SWE; however, Hazen (2001) reports that his SWE speakers overtly marked BE in copular contexts at higher rates than auxiliary contexts (96% vs. 78%). Again, the direction (copular > auxiliary) of the effect in adult SWE is consistent with the adult AAE literature.

Children's Marking of BE

Children's marking of BE has also been documented in some studies, and at least two have included AAE-speaking children. Wyatt (1991) studied overt marking of copular BE in 10 AAE-speaking children, aged 3 to 5 years. Consistent with the adult literature, the children overtly marked is and are at variable rates, and their rate of overt marking for is was higher than their rate for are (81% vs. 55%). Similar to studies completed in the adult literature, rates of use were calculated as the proportion of overtly marked BE contexts.

Garrity and Oetting (2010) examined overt marking of auxiliary am, is, and are in 30 AAE-speaking children, aged 4 to 6 years, and 20 of these children were classified as typically developing.2 The data came from language samples and an elicitation probe that included 30 items (10 for each BE structure). In contrast to the adult studies and Wyatt's study, rates of overt marking were calculated for each child and then averaged.

Using data from children who produced at least three BE contexts within their language samples, rates of overt marking were higher for am (4-year-olds = 96%; 6-year-olds = 100%) as compared to is (4-year-olds = 26%; 6-year-olds = 55%) and are (4-year-olds = 1%; 6-year-olds = 39%). For the elicitation probe, however, rates of overt marking for all forms of BE were high, ranging from 70–90%, and these rates of overt marking were not found to differ by the person, number, or tense of the BE form. There was also a statistically significant age difference in rates of marking for are and a marginally significant difference for is (4-year-olds < 6-year-olds) when the language samples but not elicitation probe data were examined.

Finally, Leonard and colleagues examined the effects of person, number, and tense on children's rates of overt marking in two studies that included speakers of MAE (Leonard et al., 2003; Polite & Leonard, 2007). The first included 45 children, aged 3 to 5 years, and the second included 32 children, aged 4 to 6 years. Data were limited to auxiliary BE forms which were elicited using probes with 18 items for am and 16 items for is and are combined and was and were combined. Similar to Garrity and Oetting (2010), the children's individual rates of marking were calculated and then averaged.

In Polite and Leonard (2007), rates of overt marking for am were above 90% for both age groups of typically developing children. In Leonard et al. (2003), an age but not BE form effect was observed when the children's combined rates of overtly marked is and are were compared to their combined rates of overtly marked was and were. The 5-year-olds' combined rates of is and are marking and was and were marking were 84% and 97%, respectively; and the 3-year-olds' combined rates were 81% and 78%, respectively. These findings show developmental changes in rates of marking between 3 and 5 years which is consistent with age effects identified in Garrity and Oetting (2010). Nevertheless, these findings also show that typically developing MAE-speaking children overtly mark auxiliary BE at high rates regardless of the person, number, and tense of the BE form.

Although the effects of contractibility and grammatical function were not the focus of the above studies, these variables have been examined in other studies that have included MAE-speaking children (e.g., de Villiers & de Villiers, 1973; James & Kahn, 1982; Johnston & Schery, 1976). As a recent example, Cleave and Rice (1997) examined BE marking in 10 MAE-speaking three-year-olds. The data were from language samples which were similar to those collected in the current study. Collapsed for grammatical function, the children's average rates of overt marking were higher for contractible contexts than for uncontractible contexts, although rates of overt marking for both contexts were relatively high (>75%). Collapsed for contractibility, the children's average rates of overt marking were higher for copular contexts than for auxiliary contexts, although again rates of overt marking for both contexts were relatively high (>80%).

Although linguistic constraints have been documented in young MAE speakers' marking of BE, by the age of six years, effects for contractibility and grammatical function on MAE-speaking children's marking of BE likely disappear. Data to support this claim come from longitudinal work by Rice, Wexler, and Hersberger (1998). In their study, 41 typically developing MAE-speaking children, aged 2;6 to 8;9, were administered language samples and probes at six-month intervals. By the age of six, the children's average percentage of overt marking for auxiliary BE was at or near 100%. This high rate of overt marking in BE contexts by MAE-speaking six-year-olds would be difficult to achieve if the appearance of these overtly marked forms were linguistically constrained by contractibility or grammatical function.

In summary, the linguistic constraints of person, number, tense, contractibility, and grammatical function have been examined in a number of adult and child studies, and these studies have included various groups of AAE, SWE, and MAE speakers. Although the age of the participants and the focus, data, and methods of these studies have varied, findings suggest dialect differences for the constraints reviewed. For AAE, the adult literature suggests that all three constraints should affect the children's rates of overt marking, with the direction of the effects showing: am and was/were > is > are; uncontractible > contractible; copular > auxiliary. For children who speak SWE, the adult literature suggests effects for the constraint of person, number, and tense, with the direction of the effect showing: am, was/were and is > are (with possibly a slightly lower rate of marking for is than am and was/were). The literature also suggests that the effects of contractibility on child SWE should be absent, and effects of grammatical function, if present, should parallel the effects observed for child AAE by showing copular > auxiliary. If these predictions are confirmed, the findings will demonstrate stability of BE marking across different generations of AAE and SWE speakers and dialect differences between child AAE and SWE. These findings will also demonstrate dialect differences between these two child dialects and child MAE given the relatively high rates at which child MAE speakers overtly mark all forms and contexts of BE. Note also that when effects for contractibility have been documented for child MAE, the direction of the effect (contractible > uncontractible) has differed from what has been documented in the adult nonmainstream English literature (uncontractible > contractible).

Method

Data

The data were spontaneous language samples from 62 children who had previously been studied by Oetting and McDonald (2001, 2002; see also Oetting & Garrity, 2006; Oetting & Newkirk, 2008; 2011; Oetting & Pruitt, 2005). All of the children lived in a rural area of Southeastern Louisiana and attended either public kindergartens or child development centers and Head Starts in close proximity to the kindergartens. For the 43 families who provided parental occupation and/or educational levels, estimates of social strata were based on Hollingshead (1975). Parental occupations ranged from skilled craft, clerical, and sales groups to small business, minor professional, and technical groups; maternal education level averaged 13 years (SD = 1.80; median = 12 years; range = 9 – 16 years).

Twenty-four children were classified as African American and speakers of AAE, and 38 were classified as non-African American (37 white; 1 Asian-Pacific) and speakers of SWE. As detailed in Oetting and McDonald (2002), the children's dialect status as an AAE or SWE speaker was confirmed through blind listener judgments and the children's use of 35 different nonmainstream English grammar structures during an elicited language sample. Within each dialect group, half of the children were four years of age (4N) and the others were six (6N); 56% were male and all were considered to be developing language typically.

Table 1 provides available test data from the children, and these included scores from the Columbia Mental Maturity Scale (CMMS, Burgmeister, Blum, & Lorge, 1972), the Peabody Picture Vocabulary Test: Revised (PPVT-R, Dunn & Dunn, 1981), three subtests that make up the Syntax Quotient of the Test of Language Development: Primary (TOLD, Newcomer & Hammill, 1988), and the Goldman Fristoe Test of Articulation (GFTA, Goldman & Fristoe, 1986). Twelve children in the 4N group were too young for the normative data of the TOLD; however, data were complete for all of the other tests and on these tests, all of the children earned a standard score above −1 SD of the normative mean. Also, of those who completed the TOLD, all but two earned a standard score above −1 SD (the exceptions were a SWE 4N child who earned a standard score of 79 and a SWE 6N child who earned a score of 83).

Table 1.

Participant characteristics and description of language samples

AAE SWE
4N 6N 4N 6N
N 12 12 19 19
Males 5 5 12 13
Ages in Months 56.83 (3) 74.5 (4) 48.3 (5) 76.1 (6)
CMMSa 102.4 (7) 101.3 (4) 104.7 (10) 104.3 (9)
PPVTb 97.8 (8) 102.2 (13) 102.2 (7) 104.9 (12)
TOLDc 97.08 (11) 101.83 (11) 96.29 (7.4) 104.8 (13)
GFTAd 80.1 (15) 95.6 (6) 88.8 (8) 92.8 (13)
Mean C & I Utterancese 192 (63) 221 (79) 204 (38) 242 (47)
Total C& I Utterancesf 2302 2652 3876 4613
Total BE Tokensg 559 693 1059 1249
Open in a new tab
a

Columbia Mental Maturity Scale, M = 100, SD = 15;

b

Peabody Picture Vocabulary Test: Revised, M = 100, SD = 15;

c

Syntactic quotient from the Test of Language Development Primary, M = 100, SD = 15;

d

Goldman Fristoe Test of Articulation, percentile scores reported;

e

Mean number of complete and intelligible utterances in each child's sample;

f

Total number of complete and intelligible utterances in samples;

g

Total number of BE tokens coded for the constraints of interest.

Identification and Coding of BE

As part of the original studies, a language sample was elicited from each child by an examiner in a quiet room at the children's schools. Toys and pictures were used as prompts. The samples were then transcribed and coded using Systematic Analysis of Language Transcripts (SALT) software (Miller & Chapman, 1992). SALT software was also used to extract and further code the children's productions of BE from their language samples. As shown in Table 1, the samples contained 13,443 complete and intelligible utterances and 3,560 code-able BE contexts. Coding of these BE contexts focused on whether the context was overtly marked (he is walking) or zero marked (he ∅ walking) and as a function of its person, number and tense (am vs. is vs. are vs. was vs. were), contractibility (+/− whether the form could be contracted in English), and grammatical function (copula vs. auxiliary). To be consistent with other studies, copular and auxiliary BE productions involving ain't (e.g., he ain't gonna do it) as well as productions of the go copula (e.g., here go a doll) and the non-equivalent habitual be morpheme (e.g., he be bad) were excluded from the analyses.

When coding for the person, number, and tense of the BE form, 94 BE forms involved cases of was and is leveling. Leveling occurs when a singular form of BE is produced with a plural subject (e.g., Me and my brother was playing…; And two mans was going…; We's gonna …; They is, see?). The AAE group produced 42 (6N = 21; 4N = 21) was forms and 7 (6N = 6; 4N = 1) is forms, and the SWE group produced 21 (6N = 7; 4N = 14) and 24 (6N = 16; 4N = 8), respectively; 54 served a copular function and 40 served an auxiliary function. When these BE forms were identified in the samples, they were coded based on their surface properties. Thus, the was forms in the first two examples above were coded as overtly marked uncontractible auxiliaries, the is in the third example was coded as an overtly marked contractible auxiliary, and the is in the fourth example was coded as an overtly marked uncontractible copula based on the utterance that preceded it.

When coding for contractibility of BE, it was also necessary to determine how best to code zero marked forms of BE when they occurred within questions. Examples of these included: You gonna come tomorrow?, That with the boy?, This where you go?, and When you coming back?. When questions included a Wh- form (e.g., When you coming back?), the zero marked BE forms were coded as uncontractible (e.g.,When ∅ you coming back?); however, when a Wh- form was not included, the contractibility of the zero marked forms was ambiguous in some cases because questions in AAE and SWE can be formed with or without inversion (e.g., Is he coming tomorrow? which reflects an uncontractible context vs. He's coming tomorrow? which reflects a contractible context). To determine how best to code these types of BE contexts, we examined all of the children's questions that contained an overtly marked BE form. Although the children produced 1,320 questions within their samples, only 56 (AAE = 19; SWE = 37) contained an overtly marked BE form whose coding would have been ambiguous had it been zero marked. For the AAE groups, 79% of these BE forms were in questions that were not inverted, and for the SWE groups, 97% of these BE forms were in questions that were inverted. Given these findings, we coded the AAE-speaking children's zero marked BE forms within these types of questions as if the questions were not inverted and the SWE-speaking children's zero marked BE forms as if the questions were inverted. Other studies of children and adults have also documented the noninversion of questions in AAE (Green, 2011; Washington & Craig, 2002); however, to our knowledge, similar studies have not been completed with children and adults who speak SWE.

Reliability

Reliability of the language sample transcriptions was examined as part of the original studies and was found to be above 90%. Reliability of the BE coding was examined by having a second research assistant independently code 20% (n = 12) of the samples. Within these samples, there were 680 BE contexts. Inter-rater agreement in the identification and coding of these BE contexts averaged 98% (range across children = 94 – 100%).

Results

Analysis of Variance

In Table 2, the children's percentages of overt marking for BE by each dialect and age group are presented. To examine these data, we first completed a mixed ANOVA with dialect and age as between-subject variables and the three constraints (person/number/tense, contractibility, and grammatical function) as within-subject variables. For the between-subject variables, main effects were significant for dialect, F(1,58) = 88.42, p < .0001, partial η2 = .17. For the within-subject variables, main effects were observed for person, number, and tense, F(4,191) = 42.06, p < .0001, partial η2 = .22, and grammatical function, F(1,61) = 24.96, p < .0001, partial η2 = .15.

Table 2.

Percent of overt marking of BE.

AAE SWE
4N 6N 4N 6N
Person/Number
Am 100 (0) 91 (26) 95 (14) 96 (18)
Is 38 (31) 57 (32) 88 (24) 94 (19)
Are 34 (38) 31 (40) 72 (41) 75 (38)
Was/were 94 (17) 94 (19) 97 (17) 99 (6)
Contractibility
Contractible 56 (39) 53 (40) 86 (27) 88 (26)
Uncontractible 62 (42) 76 (36) 88 (31) 91 (26)
Grammatical Function
Copular 62 (37) 72 (34) 95 (14) 91 (24)
Auxiliary 55 (43) 57 (43) 76 (38) 88 (29)
Total 58 (44) 64 (39) 87 (29) 89 (26)
Open in a new tab

Unfortunately, the null effect for contractibility as well as the non-significant interactions between the children's dialects and the three linguistic constraints within the ANOVA are difficult to interpret because the various BE tokens occurred in the different within-subject contexts at labile frequencies that were not controlled. To illustrate, Table 3 presents the total number of BE tokens for all of the children in the AAE 4N group. As can be seen, the lowest number of BE tokens for any given context was 0, and 0 tokens were found for 6 out of the 96 possible cells. Further, the number of tokens per child fluctuated from a low of 23 to a high of 92. Such a highly unbalanced data set is the product of the observational nature of language sample data. Whereas in the aggregate, an ANOVA approach can reasonably assess between-subject differences, it does not accurately assess the effects of within-subject variables such as the linguistic constraints studied here when there is such severe unbalance in the data. The assumption of normality in an ANOVA approach also does not hold with binomial dependent variables (i.e., overt BE vs. zero BE) when the overall proportions for cells are less than 30% or greater than 70%. The relationship, especially outside of this 30 – 70 percent boundary, between the dependent variable and the independent variables is no longer linear, and linearity is a crucial assumption of an ANOVA (see Cedergren & Sankoff, 1974, Sankoff & Labov, 1979, and Sankoff, 1988 for early discussions of these issues as well as more recent work by Quené & van der Berg, 2008, and Tagliamonte, 2006).

Table 3.

Frequency of BE contexts for 4N AAE group.

Participant Identification Number
100 101 103 105 110 112 114 115 117 118 119 120
Total BE Tokens
47 26 38 23 38 62 38 92 49 26 94 25
BE Tokens by Person, Number, Tense
Am 7 0 0 5 2 2 0 4 3 4 2 0
Is 34 20 21 9 15 39 23 67 25 19 57 12
Are 2 0 5 3 3 13 6 13 12 3 24 2
Was/were 4 6 12 6 18 8 9 8 9 0 11 11
BE Tokens by Contractibility
Contractible 41 16 26 13 20 37 25 64 30 18 69 11
Uncontractible 6 10 12 10 18 25 13 28 19 8 25 14
BE Tokens by Grammatical Function
Copular 28 19 24 18 22 46 24 64 36 20 44 11
Auxiliary 19 7 14 5 16 16 14 28 13 6 50 14
Open in a new tab

Logistic Regression

An alternative approach to data analysis that allows for the binary nature of a dependent variable (i.e., overt BE vs. zero BE) as well as the unbalanced nature of data is logistic regression (see Sankoff, 1988; Tagliamonte, 2006; Walker, 2010 for in-depth discussion). A Generalized Linear Model (GLM) approach to logistic regression is implemented by most modern statistical programs (Hosmer & Lemeshow, 2000). This approach treats the relationship between a binary dependent variable and the independent variables as logistic and transforms the dependent variable accordingly, where a p context represents the probability of a linguistic variant occurring in a particular context:

log(p1−p)=m+pContext1+pContext2+…

There are two approaches to assessing the statistical significance of a factor (i.e., variable) within a logistic regression analysis. In the first approach, used by many sociolinguists, a step-up and step-down procedure selects a sub-set of statistically significant factors from all factors with a given alpha, such as .05. In the second approach, statistical significance is assessed on each factor individually. Our work here follows the second approach. With this approach, we also employ the comparative method outlined in Poplack and Tagliamonte (2001) and presented in detail by Tagliamonte (2002). As such, we focus on the estimated probabilities for contexts that obtain statistical significance. From these estimated probabilities, we then estimate the strength of each factor's effect on the children's overt marking of BE. In the results discussed below, we use the GENMOD procedure of SAS to produce the results and transform the log-odds into more easily interpretable probabilities.

In Table 4, the results of the multi-factor logistic regression are presented for both dialect groups with the data collapsed for age given that age was not identified as a significant variable in the earlier ANOVA. The factor weights represent the likelihood of the children producing an overtly marked BE form instead of a zero marked form in each context and with respect to the corrected mean (i.e., the estimated average of overt marking across all contexts). Factor weights above .50 are said to favor overt marking of BE whereas weights below .50 are said to disfavour overt marking. However, with the comparative method, the importance of this multi-factor analysis is not the particular numerical value of the weights, but the ordering of the weights. The unbalanced nature of conversational data leads to weights that will fluctuate between studies, but the ordering of the weights is more stable.

Table 4.

Logistic regression for the selection of overt BE for AAE and SWEa.

AAE SWE
FW % N FW % N
Overall Rate .78 64 1248 .95 93 2294
Person, Number, Tense
Are .09 27 175 .16 77 192
Is .22 59 740 .41 95 1518
Am .84 94 67 .59 96 192
Was/were .90 96 263 .84 99 319
Range 81 68
Contractibility
Contractable .41 57 818 [] 93 1562
Uncontractable .59 77 430 [] 94 578
Range 18 --
Grammatical Function
Copular .64 70 807 .63 96 1571
Auxiliary .36 53 441 .37 87 569
Range 28 26
Open in a new tab
a

FW = factor weight. The weight reported reflects the estimated probability for the overall rate of overt BE from the logistic regression. The % reflects the proportion of BE contexts overtly marked for the group as a whole. The N reflects the number of BE contexts included in the analysis.

As can be seen in the table, the results were consistent with the ANOVA approach in that they showed lower rates of overt marking for the AAE group than then SWE group (corrected means for AAE = 64% vs. SWE = 93%). More interestingly, however, were the factor weights associated with the three linguistic constraints. The factor weights indicated that all three linguistic constraints obtained statistical significance for the AAE group: person, number and tense, χ2(3) = 259.96, p <.001, contractibility, χ2(1) = 12.26, p > .001, and grammatical function, χ2(1) = 54.16, p < .001. When considering the effect of person, number and tense of BE, the AAE group were more likely to overtly mark BE in was/were contexts (factor weight = .90) than in am (factor weight = .84), is (factor weight = .22) or are (factor weight = .09) contexts.

As can be seen by these results, the factor weights of the logistic regression offer more information about the children's grammars than the raw percentages or the analyses of variance. As an example, the factor weight of .09 for are is an estimate of the probability of the children overtly marking are while controlling for the effects of contractibility and grammatical function and while also allowing for uneven numbers of tokens across factors. The raw percentages and the group averages examined within the ANOVA do not allow for generalizations about the magnitude of the are effect on the children's marking of BE even though a statistically significant main effect was identified for person, number, and tense within the ANOVA analysis.

The range of the weights represents another important component of a multi-factor logistic regression analysis. The range is the distance between the largest weight and the smallest weight, and the range represents the strength (or magnitude) of the effect on the children's overt marking of BE. For person, number and tense, the range for the AAE group was 81 (.90–.09 = 81, with the decimal removed to be consistent with methods used in the field of sociolinguistics). The taxonomy used to interpret the range is summarized by Horvath and Horvath (2003): <10 is a weak effect; 10–30 is a moderate effect; 30–50 is a strong effect; >50 is a very strong effect. Using these guidelines, the data showed that the person, number, and tense of the BE form had a very strong effect on the AAE group's overt marking. Moreover, each of the BE contexts obtained statistical significance when tested by post-hoc Wald-Z scores: was/were, Z = 3.49, p = .009; am, Z = 4.17, p < .0001; is, Z = −7.07, p < .0001; and are, Z = −8.96, p < .0001.

The second variable that obtained statistical significance for the AAE group was the grammatical function of BE. For this variable, copular contexts favored overt marking of BE (factor weight = .64) and auxiliary contexts disfavoured overt marking (factor weight = .36). As indicated by a range of 28, grammatical function showed an overall moderate effect on the AAE group's overt marking of BE.

The third variable to have obtained statistical significance for the AAE group was the contractibility of BE. For this variable, uncontractable contexts favoured overt marking (factor weight = .59) and contractable contexts disfavoured overt marking (factor weight = .41). As indicated by a range of 18, contractibility had an overall moderate effect on the AAE's group's overt marking of BE. Based on these results, the ordering of the factors can be described in a constraint hierarchy by their magnitude: person, number and tense > grammatical function > contractibility.

For the SWE group, the person, number and tense of BE obtained statistical significance, χ2(3) = 90.32, p < .001, and the effect was very strong (range = 68). Although the direction of the effect is similar to what was observed for the AAE group: was/were > am > is > are, only are obtains statistical significance when tested by post-hoc Wald-Z scores: are, Z = −8.19, p < .001. The SWE child utterances with are zero marked were varied as evidenced by the following examples: Even when you Ø big; We Ø not gonna go; Some of them Ø from my birthday…; Two people Ø on the ground; …didn't bring the toys that Ø the same; They Ø going fishing; What Ø these for? Thus, the effect of the person, number, and tense constraint on SWE-speaking children's marking of BE, while limited to the form are, was robust across various types of utterances and grammatical constructions.

The grammatical function of BE for child SWE also obtained statistical significance, χ2(1) = 41.38, p < .001, and showed a moderate effect (range = 26) on the children's overt marking of BE, with the direction of the effect showing copula > auxiliary. This finding was consistent with the results of the AAE group. Unlike the AAE group, however, the SWE group's rate of overt marking did not show a significant effect for the contractibility of the BE form. For both of these contexts, rates of overt marking were high (93% and 94%).

Discussion

In this study, AAE- and SWE-speaking children's overt marking of BE was examined as a function of their dialect and age, and as a function of three linguistic constraints that have been shown to influence nonmainstream English-speaking adults' overt marking of BE. From the first set of analyses which made use of ANOVA, the results showed that the children's rates of overtly marked BE varied by their dialect but not their age, with lower rates of overt marking observed for the AAE child speakers than for the SWE child speakers. The person, number, tense, and the grammatical function of the BE forms were also found to influence the children's rates of overt marking within the analyses, but after considering the unbalanced and binary nature of the children's BE productions, the ANOVA approach was deemed limited in its ability to evaluate the influence of the three linguistic constraints.

From the second set of analyses which made use of logistic regression, the results again showed AAE-speaking children to produce a lower rate of overtly marked BE forms than the SWE-speaking children but these analyses also revealed differences in the constraint hierarchies manifested within the two dialects. Whereas the constraint hierarchy for the AAE-speaking children involved all three linguistic variables (person, number and tense > grammatical function > contractibility), the SWE-speaking children's constraint hierarchy was limited to two (person, number and tense > grammatical function). Moreover, although BE marking in both child AAE and child SWE was influenced by the person, number, and tense of BE, the pattern of statistical significance within the dialects varied. For child AAE, the proportions of overtly marked BE forms for am, is, are, and was/were were all statistically different from the proportion of all forms marked. For child SWE, only are was overtly marked at a proportion that statistically differed from (and was lower than) the proportion of all forms marked.

The results documented here for child AAE and child SWE are consistent with the adult nonmainstream English literature. Recall that in adult AAE, all three constraints have been shown to influence rates of overtly marked BE, and this is exactly what was found for the child AAE speakers. The direction and magnitude of the constraints within child AAE also aligns with what has been reported for adult AAE. Like adult AAE, the child AAE speakers produced low rates of overt marking for are, higher rates of overt marking for is, and high rates of overt marking for am and was/were. Like adult AAE, the child AAE speakers produced higher rates of overt marking for copular contexts than auxiliary contexts and higher rates of overt marking for uncontractible contexts than for contractible contexts.

In adult SWE, lower rates of overt marking have been consistently documented for are, and effects have also been documented for the linguistic constraint of grammatical function, with higher rates of marking for copular contexts than for auxiliary contexts. Like the adult SWE literature, the child SWE speakers produced low rates of overt marking for are only and produced higher rates of marking for copular contexts than for auxiliary contexts. Moreover, although the effect for grammatical function was statistically significant, the SWE children's rates of overt marking for both the copular and auxiliary contexts were relatively high (96% and 87%), and this finding is similar to Hazen's (2001) results for SWE-speaking adults (copular =96% vs. auxiliary =78%). Finally and as far as we know, no adult SWE study has revealed an effect for contractibility, and an effect for this constraint was not documented for the SWE-speaking children studied here.

The results presented here were made possible by applying two types of statistical analyses to the data. The traditional ANOVA method allowed us to detect important quantitative differences in the children's rates of overt marking, and logistic regression allowed us to uncover important qualitative differences in the children's constraint hierarchies. When compared to adult AAE and SWE, the results show a high degree of consistency across generations of AAE and SWE speakers. This finding occurred even though none of the adult studies reviewed was conducted in or near the dialect communities of the children. Moreover, the earliest adult studies reviewed were published in 1969, which was ~30 years before the child data were collected. These results indicate that BE marking, like verbal –s marking but not relative clause marking, is a grammatical structure that is remaining stable across different generations and different communities of AAE and SWE speakers.

Our results for child AAE and child SWE differ from each other and they also contrast with literature on children who speak MAE. When the three dialects are compared, the most striking difference between them involves the person, number, and tense of the BE form. For this linguistic constraint, AAE shows effects for all forms, SWE shows effects for are only, and MAE shows no effects. Contractibility also shows interesting dialect differences between AAE, SWE, and MAE. For this constraint, AAE shows effects with the highest rates of overt marking for uncontractible contexts, SWE shows no effects, and MAE, when effects are documented, show the highest rates of overt marking for contractible contexts. Thus, for this linguistic constraint, the presence and direction of the effect differs across the three dialects. For grammatical function, all three dialects are similar to each other and show higher rates of overt marking for copular contexts than for auxiliary contexts.

Although the dialect differences identified in the data are consistent with previous studies, we were unable to detect a significant effect for age in the children's marking of BE. In fact, within each dialect, the average percentages of overt marking were near identical for the four- and six-year-olds (AAE: 58% and 64%; SWE: 87% and 89%). These percentages also appear to align with rates reported in the adult AAE and SWE literature (although overall rates of overt BE marking are not routinely reported for adults). If this finding replicates, then it may be the case that children who speak AAE and SWE reach adult levels of use for BE at an earlier age than children who speak MAE. Additional studies are needed to test this hypothesis. Our study was limited to data from 62 children whose language samples were collected as part of a set of previous studies. The standard deviations that accompanied the group averages of the four- and six-year-olds were also large, especially for the AAE group at 44% and 39%, respectively. In our on-going work, we are systematically recruiting larger numbers of AAE- and SWE-speaking children who vary in their maternal education levels and test scores, and in the rates at which they produce nonmainstream grammar structures to determine if these child variables can explain some of the within-dialect variability that occurs in studies of nonmainstream English speakers. For BE, the findings from the current work indicate that any future study of these types of child attribute variables must control for the form's person, number, and tense, contractibility, and grammatical function.

Speech-language clinicians should be able to use the current set of findings to better understand and articulate the ways in which children's grammars vary across different dialects of English. As shown here, the dialect variation involves differences in the rates of overt marking and in the linguistic constraint hierarchies. In other words, these child dialects differ in the rates at which they overtly mark BE and in their linguistic conditioning of BE, the latter of which influences when a particular type of marking (zero or overt) will be produced.

The findings also have implications for assessment, treatment of children with language impairments, and consultative services to teachers of nonmainstream English-speaking children without language impairments. For assessment, the findings can be used to advocate for language evaluations to include a wide range of grammar structures. For BE, this evaluation should target multiple forms and contexts to fully capture children's rates of use and their dialect-specific constraint hierarchies. For treatment of children with language impairments and whose rates of BE marking and constraint hierarchies are shown to fall outside of a dialect-appropriate range, the findings can be used to strategically target some forms and contexts of BE before others.

Finally, for consultative services, the findings can be used by speech-language clinicians to evaluate the test and teaching materials of a classroom. For BE, the current findings suggest that classroom materials should not present a limited number and type of BE forms. BE form variability in classroom materials should help nonmainstream English-speaking children demonstrate their knowledge when they are tested. BE form variability in classroom materials should also help these same children expand their understanding and use of grammar in school and when interacting with same dialect- and different dialect-speaking communication partners.

Acknowledgments

This study was made possible by RO3 DC03609 and RO1 DC009811 from the National Institute on Deafness and Other Communication Disorders. The authors would like to thank the children, parents, and teachers who made data collection possible, Tina Villa for completing an analysis of the children's questions, and Nicté Fuller Medina, Kyomi Gregory, Allison Lealess, Ryan Lee, Jessica Richardson, Andrew Riviére, Christy Seidel, and Tina Villa for providing valuable feedback on an earlier draft of this manuscript.

Footnotes

1

Other linguistic constraints that have been examined in some adult AAE studies include: the preceding subject context (it/that/what vs. other pronouns vs. specific nouns), stress (emphatic vs. non-emphatic), the clause position (final vs. non-final), and the preceding and following phonological context (vowel vs. consonant). For discussion and analysis of various combinations of these constraints in adult AAE, see Blake (1997), Rickford et al. (1991), and Wyatt (1996).

2

This study as well as a few other child language studies reviewed includes children classified as specifically language impaired; findings from these children are not discussed because they are not the focus of the work.

References

  1. Bailey G, Maynor N. The present tense of be in southern Black folk speech. American Speech. 1985;60:195–213. [Google Scholar]
  2. Baugh J. A re-examination of the Black English copula. In: Labov W, editor. Locating language in time and space. Academic; New York: 1980. pp. 83–195. [Google Scholar]
  3. Blake R. Defining the envelope of linguistic variation: The case of “don't count” forms in the copula analysis of African American Vernacular English. Language Variation and Change. 1997;9:57–79. [Google Scholar]
  4. Brown R. A First Language: The Early Stages. Harvard University Press; Cambridge, MA: 1973. [Google Scholar]
  5. Burgmeister B, Blum H, Lorge I. Columbia Mental Maturity Scale. The Psychological Corporation; San Antonio, TX: 1972. [Google Scholar]
  6. Cedegren H, Sankoff D. Variable Rules: Performance as a Statistical Reflection of Competence. Language. 1974;50:333–355. [Google Scholar]
  7. Childs B, Malinson C. African American English in Appalachia: Dialect accommodation and substrate influence. English World-Wide. 2004;25:27–50. [Google Scholar]
  8. Cleave PL, Rice ML. An examination of the morpheme BE in children with specific language impairment: The role of contractibility and grammatical form class. Journal of Speech, Language, and Hearing Research. 1997;40:480–492. doi: 10.1044/jslhr.4003.480. [DOI] [PubMed] [Google Scholar]
  9. Cleveland LH. Unpublished doctoral dissertation. Louisiana State University; 2009. Children's productions of verbal –s by dialect type and clinical status. [Google Scholar]
  10. Cukor-Avila P. Co-existing grammars: The relationship between the evolution of African American and White Vernacular English in the South (-) In: Lanehart S, editor. Social and Historical Contexts of African American English. John Benjamins; Amersterdam: 2001. [Google Scholar]
  11. de Villiers JG, de Villiers PA. A cross-linguistic study of the acquisition of grammatical morphemes. Journal of Psycholinguistic Research. 1973;2:267–278. doi: 10.1007/BF01067106. [DOI] [PubMed] [Google Scholar]
  12. Dubois S, Horvath BM. From accent to marker in Cajun English: A study of dialect formation in progress. English World-Wide. 1998;19:161–188. [Google Scholar]
  13. Dubois S, Horvath BM. Let's tink about dat: Interdental fricatives in Cajun English. Language Variation and Change. 1999;10:245–261. [Google Scholar]
  14. Dunn LM, Dunn LM. Peabody Picture Vocabulary Test - Revised. American Guidance Service; Circle Pines, MN: 1981. [Google Scholar]
  15. Fasold R, Wolfram W. Some linguistic features of Negro dialect. In: Fasold R, Shuy R, editors. Teaching Standard English in the Inner City. Center for Applied Linguistics; Washington, DC: 1970. pp. 41–86. [Google Scholar]
  16. Goldman R, Fristoe M. Goldman-Fristoe Test of Articulation. American Guidance Service; Circle Pines, MN: 1986. [Google Scholar]
  17. Garrity A, Oetting J. Auxiliary BE production by African American English-speaking children with and without specific language impairment. Journal of Speech, Language, and Hearing Research. 2010;53:1307–1320. doi: 10.1044/1092-4388(2010/09-0016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Green LJ. Language and the African American child. Cambridge University Press; Cambridge, MA: 2011. [Google Scholar]
  19. Hazen K. Identity and Ethnicity in the Rural South: A sociolinguistic view through past and present BE. Duke University Press; Durham, North Carolina: 2001. [Google Scholar]
  20. Harris R, Zimmerman J. The Southern Rural Development Center Policy. Mississippi State, MS: 2003. Children and poverty in the rural South. (Series, No. 2). [Google Scholar]
  21. Hollingshead A. Four-factor index of social position. Yale University; 1975. Unpublished manuscript. [Google Scholar]
  22. Horton-Ikard R, Ellis Weismer S. Distinguishing African-American English from developmental errors in the language production of toddlers. Applied Psycholinguistics. 2005;26:597–620. [Google Scholar]
  23. Horvath BM, Horvath RJ. A closer look at the constraint hierarchy: Order, contrast and geographic scale. Language, Variation and Change. 2003;15:143–170. [Google Scholar]
  24. Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd Edition John Wiley & Sons, Inc; New York: 2000. [Google Scholar]
  25. Jackson S, Roberts J. Complex syntax production of African American preschoolers. Journal of Speech, Language, and Hearing Research. 2001;44:1083–1096. doi: 10.1044/1092-4388(2001/086). [DOI] [PubMed] [Google Scholar]
  26. James SL, Kahn LM. Grammatical morpheme acquisition: An approximately invariant order? Journal of Psycholinguistic Research. 1982;11:381–388. [Google Scholar]
  27. Johnston JR, Schery TK. The use of grammatical morphemes by children with communication disorders. In: Morehead D, Morehead A, editors. Normal and deviant child language. University Park Press; Baltimore, MD: 1976. pp. 239–258. [Google Scholar]
  28. Labov W. Contraction, deletion, and the inherent variability of the English copula. Language. 1969;45:715–762. [Google Scholar]
  29. Labov W. Language in the inner city. University of Pennsylvania Press; Philadelphia, PA: 1972. [Google Scholar]
  30. Leonard LB, Deevy P, Miller CA, Charest M, Kurtz R, Rauf L. The use of grammatical morphemes reflecting aspect and modality by children with specific language impairment. Journal of Child Language. 2003;30:769–796. doi: 10.1017/s0305000903005816. [DOI] [PubMed] [Google Scholar]
  31. Miller J, Chapman R. Systematic Analysis of Language Transcripts [Computer software] Language Analysis Laboratory; University of Wisconsin; Madison: 1992. [Google Scholar]
  32. Miller K, Weber B. Persistent poverty and place: How do persistent poverty dynamics and demographics vary across the rural urban continuum. 2004 Retrieved from the Southern Rural Development Center: http://srdc.msstate.edu.
  33. Newcomer P, Hammill D. Test of Language Development–Primary. Second Edition ProEd; Austin, TX: 1988. [Google Scholar]
  34. Oetting JB, Garrity AW. Variation within dialects: A case of Cajun/Creole influence within child AAE and SWE. Journal of Speech, Language and Hearing Research. 2006;49:16–26. doi: 10.1044/1092-4388(2006/002). [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Oetting JB, McDonald JL. Nonmainstream dialect use and specific language impairment. Journal of Speech, Language, and Hearing Research. 2001;44:207–223. doi: 10.1044/1092-4388(2001/018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Oetting JB, McDonald JL. Methods for characterizing participants' nonmainstream dialect use in child language research. Journal of Speech, Language, and Hearing Research. 2002;45:505–518. doi: 10.1044/1092-4388(2002/040). [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Oetting JB, Newkirk BL. Subject relatives by children with and without SLI across different dialects of English. Clinical Linguistics and Phonetics. 2008;22:111–125. doi: 10.1080/02699200701731414. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Oetting JB, Newkirk BL. Children's relative clause markers in two nonmainstream dialects of English. Clinical Linguistics and Phonetics. 2011;25:725–740. doi: 10.3109/02699206.2011.553700. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Oetting JB, Pruitt SL. Use of Southern African American English across groups. International Journal of Multicultural Communication Disorders. 2005;3:136–144. [Google Scholar]
  40. Polite E, Leonard LB. A method for assessing the use of first person verb forms by preschool-aged children with SLI. Child Language Teaching and Therapy. 2007;23:353–366. doi: 10.1177/0265659007083640. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Poplack S, Tagliamonte S. African American English in the Diaspora. Blackwell Publishers; Oxford: 2001. [Google Scholar]
  42. Quené H, van den Bergh H. Examples of mixed-effects modeling with crossed random effects and with binomial data. Journal of Memory and Language. 2008;59:413–425. [Google Scholar]
  43. Rice ML, Wexler K, Hershberger S. Tense over time: The longitudinal course of tense acquisition in children with specific language impairment. Journal of Speech, Language and Hearing Research. 1998;41:1412–1431. doi: 10.1044/jslhr.4106.1412. [DOI] [PubMed] [Google Scholar]
  44. Rickford J. The creole origin of African American Vernacular English: Evidence from copula absence. In: Mufwene S, Rickford J, Bailey G, Baugh J, editors. African American Vernacular English Structure, History and Use. Routeledge; New York: 1998. pp. 154–200. [Google Scholar]
  45. Rickford JR. African American Vernacular English. Blackwell; Malden, MA: 1999. [Google Scholar]
  46. Rickford J, Ball A, Blake R, Jackson R, Martin N. Rappin on the copula coffin: Theoretical and methodological issues in the analysis of copula variation in African-American Vernacular English. Language Variation and Change. 1991;3:103–132. [Google Scholar]
  47. Romaine S. Socio-historical linguistics: Its status and methodology. Cambridge University Press; Cambridge: 1982. [Google Scholar]
  48. Sankoff D. Variable Rules. In: von Ulrich Ammon Herausgegeben, Dittmar Norbert, Mattheier Klaus J., editors. An International Handbook of the Science of Language and Society. Mouton De Gruyter; Berlin: 1988. pp. 984–996. [Google Scholar]
  49. Sankoff D, Labov W. On the uses of variable rules. Language in Society. 1979;8:189–222. [Google Scholar]
  50. Sankoff D, Tagliamonte S, Smith E. GOLDVARB X. 2005 [Google Scholar]
  51. Tagliamonte S. Comparative sociolinguistics. In: Chambers JK, Trudgill P, Schilling-Estes N, editors. The Handbook of Language Variation and Change. Blackwell; Oxford: 2002. pp. 729–763. [Google Scholar]
  52. Tagliamonte S. Analyzing Sociolinguistic Variation. Cambridge University Press; Cambridge: 2006. [Google Scholar]
  53. Walker J. Rephrasing the copula: Correction and zero in early African American English. In: Poplack S, editor. The English History of African American English. Blackwell; Oxford: 2000. pp. 35–72. [Google Scholar]
  54. Walker J. Variation in Linguistics Systems. Routledge; New York: 2010. [Google Scholar]
  55. Washington JA, Craig HK. Dialectal forms during discourse of poor, urban, African American preschoolers. Journal of Speech and Hearing Research. 1994;37:816–823. doi: 10.1044/jshr.3704.816. [DOI] [PubMed] [Google Scholar]
  56. Washington JA, Craig HK. Morphosyntactic forms of African American English use by young children and their craegivers. Applied Psycholinguistics. 2002;23:209–231. [Google Scholar]
  57. Winford D. Another look at the copula in Black English and Caribbean Creoles. American Speech. 1992;67:21–60. [Google Scholar]
  58. Wolfram W. A sociolinguistic description of Detroit Negro Speech. Center for Applied Linguistics; Washington, DC: 1969. [Google Scholar]
  59. Wolfram W. The relationship of White Southern speech to vernacular Black English. Language. 1974;50:850–872. [Google Scholar]
  60. Wolfram W, Schilling-Estes N. American English: Dialects and variation. Blackwell; Malden, MA: 1998. [Google Scholar]
  61. Wolfram W, Thomas E. The development of African American English. Blackwell Publishing; Malden, MA: 2002. [Google Scholar]
  62. Wyatt T. Unpublished dissertation. University of Massachusetts; 1991. Linguistic constraints on copula production in Black English child speech. [Google Scholar]
  63. Wyatt T. The acquisition of African American English copula. In: Kamhi A, Pollock K, Harris J, editors. Communication development and disorders in African American children. Paul H. Brookes Publishing Co; Baltimore, MR: 1996. pp. 95–116. [Google Scholar]

RESOURCES