Running head: Dialect awareness and lexical comprehension of MAE Dialect Awareness and Lexical Comprehension of Mainstream American English in African American English-Speaking Children

Purpose: This study was designed to examine the relationships among minority dialect use, language ability, and young AAE-speaking children’s understanding and awareness of MAE. Methods: 83 4to 8-year-old African American English-speaking children participated in two experimental tasks. One task evaluated their awareness of differences between Mainstream American English (MAE) and African American English (AAE), while the other evaluated their lexical comprehension of MAE in contexts that were ambiguous in AAE but unambiguous in MAE. Receptive and expressive vocabulary, receptive syntax, and dialect density were also assessed. Results: The results of a series of mixed-effect models showed that children with larger expressive vocabularies performed better on both experimental tasks, relative to children with smaller expressive vocabularies. Dialect density was a significant predictor only of MAE lexical comprehension; children with higher levels of dialect density were less accurate on this task. Conclusions: Both vocabulary size and dialect density independently influenced MAE lexical comprehension. The results suggest that children with high levels of nonmainstream dialect use have more difficulty understanding words in MAE, at least in challenging contexts and suggest directions for future research. Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 3 Dialect Awareness and Lexical Comprehension of Mainstream American English in African American English-Speaking Children American English is spoken in a variety of dialects associated with different racial/ethnic groups, geographic regions, and income strata (Wofram & Schilling-Estes, 1998). Minority dialects such as African American English (AAE) or Southern White English (SWE) are often contrasted with Standard (or Mainstream) American English (e.g., Oetting, 2003). Considerable research has examined whether use of a minority dialect has an impact on school achievement. Most of this research has focused on AAE and whether it is related to the Black-White achievement gap in reading (Labov, 1995; Washington, Terry & Seidenberg, 2013). AAE and MAE overlap but also differ with respect to phonology, morphosyntax, and pragmatics (e.g., Craig & Washington, 1994; Rickford, 1999; Washington & Craig, 2002). Sociolinguistic research by Labov (1972) and others (see Rickford, Sweetland, & Rickford, 2004, for review) has established that AAE is not a linguistically deficient version of the mainstream dialect. Rather, it is representative of the kind of dialectal variation that occurs in most spoken languages (Chambers, 1992). The unresolved questions are not about the linguistic validity of the dialect but rather about the socio-cultural conditions under which it is used. In many cases, children speak the minority dialect in the home and community, but the mainstream dialect is used in school. There are further questions about which differences between dialects have a significant impact (positive or negative), and how they affect tasks such as learning to read. Determining whether dialect usage has an impact on children’s learning is particularly important because it can potentially be addressed more readily than other factors that contribute to the achievement gap, such as poverty and its Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 4 various sequelae (e.g., poorer nutrition, health care, schools, access to educational resources). For example, recent research suggests that awareness and understanding of the mainstream dialect can be enhanced in short term programs using contrastive analysis that are aimed at children in the pre-kindergarten to first grade years (Craig, 2013; Edwards, Rosin, Gross, & Chen, 2013) and can also be taught successfully to older middle and high school children (Lybolt, Gottfred, Anderson, & Olszewski, 2009). Some early studies found that the use of AAE had no impact on school achievement (e.g., Harber, 1977 as cited in Washington et al., 2013). However, in light of the persistence of the achievement gap there has been a revival of research on the topic, using more advanced experimental methods and quantitative analysis tools. A growing body of research suggests that differences between the home and school dialects have important effects on children’s performance (e.g., Charity, Scarborough, & Griffin, 2004; Connor & Craig, 2006; Craig, Zhang, Hensel, & Quinn, 2009; Craig, Kolenic, & Hensel, 2013; Terry, Connor, Thomas-Tate, & Love, 2010; Terry & Connor, 2012; Terry, Connor, Petscher, & Conlin, 2012). These are sometimes termed “dialect mismatch effects,” but the term has been used in different ways and should be interpreted cautiously. In one of these more recent studies, Terry et al. (2010) found that first grade children’s use of non-mainstream dialect features (as measured using the Diagnostic Evaluation of Language Variation [DELV]; Seymour, Roeper, & de Villiers, 2005), was negatively correlated with standardized measures of vocabulary size and phonological awareness. In more recent work, Terry and colleagues (Terry & Connor, 2012; Terry et al., 2012) followed children from kindergarten to first grade and from first to second Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 5 grade. Both studies found that non-mainstream dialect use, as measured by the DELV, was significantly and negatively predictive of reading ability; the greater the use of nonmainstream dialect at the earlier measurement point, the poorer the reading outcomes at the later measurement point. Researchers have proposed several ways in which dialect mismatch may contribute to the achievement gap. It is possible that dialect mismatch, in and of itself, may make learning more difficult. Children who speak a non-mainstream dialect may need to use greater cognitive resources simply to understand instruction in MAE, resulting in fewer cognitive resources that are available for understanding what is being taught (e.g., Harris and Schroeder, 2013). Furthermore, children who speak AAE, in particular, may have difficulty in decoding at early stages of learning to read, given the phonological differences between MAE and AAE (Labov, 1995). An alternative hypothesis is that the problem is not with dialect mismatch per se, but that difficulty learning to code-switch from a non-mainstream dialect to MAE is a symptom of more general problems with linguistic flexibility and metalinguistic awareness (Craig et al., 2013, Terry et al., 2010; Terry & Connor, 2012; Terry et al., 2012, Terry, 2012). Clearly, these two explanations are not necessarily mutually exclusive. Furthermore, neither of these explanations is intended to explain all of the achievement gap. It is indisputable that there are many risk factors, both environmental and endogenous, that are associated with poverty; these factors contribute to the achievement gap in general and to poor language skills in particular. 1 A third explanation, which will not be addressed in this study, is that teachers may have negative impressions of students who speak non-mainstream dialects (e.g., Labov, 1995) and it is well known that teacher expectations impact academic outcomes (e.g., Cooper, 1979). Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 6 While recent studies have shown that there is a consistent relationship between higher levels of non-mainstream dialect use and poor literacy outcomes, at least in the early school years, it is still unclear why such correlations exist. The current study was designed to investigate factors that might inform our understanding of the relationship between AAE use and school achievement. We measured the “density” of African American English-speaking children’s use of AAE and related this measure to children’s performance on two experimental language tasks. The first task required participants to associate different-colored animated cartoon monsters with either AAE or MAE and was designed to evaluate one component of what is needed to learn how to dialect shift. In real life, in order to shift from one dialect to another (i.e., from AAE to MAE), children must make the implicit generalization that different groups of people defined along one or more dimensions (white people vs. black people, people at home vs. people at school, etc.) speak differently. The dialect awareness task was particularly demanding because it required that children make this generalization in a fairly explicit manner in a short period of time, given limited information about two groups of speakers (the cartoon monsters using different dialects). We would expect that children who performed well on this task would be successful dialect-shifters in social contexts that demand it, such as a school setting. Based on previous research on successful dialect shifters, we would expect that such children would have larger vocabularies and lower levels of dialect density (e.g., Terry et al. 2010). The second experiment was designed to measure directly how well the participants could comprehend MAE. Such a task is difficult to design because there is so much overlap between MAE and AAE. We decided to examine comprehension of Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 7 words that are ambiguous in AAE, but unambiguous in MAE, because of phonological or morphological differences between the two dialects. For example, the word coal is unambiguous in MAE but ambiguous in AAE; it could mean either coal or cold because of final consonant cluster reduction. The MAE lexical comprehension task is also fairly demanding, but for a different reason. This task requires children who are usually AAE users to interpret spoken words in terms of MAE phonology instead of AAE phonology. We hypothesized that children who are better at comprehending words such as coal that are unambiguous in MAE but ambiguous in AAE would be children with stronger language skills and more experience with MAE, as evidenced by lower dialect density. We also predicted that there should be a relationship between performance on the two experimental tasks. Children who performed better on the dialect awareness task should be better able to learn to shift between AAE and MAE; therefore, performance on the dialect awareness task should be a predictor of performance on the MAE lexical comprehension task. Methods Participants The participants were 83 AAE-speaking children from 4 to 8 years of age (mean = 6;3 [years; months], SD = 1;3, range = 4;0 to 8;9). As described below, children’s use of 2 The original sample included 105 African American children from 4 to 8 years of age (mean = 6;2, SD = 1;3, range = 4;0 to 8;9). There were 24 to 25 children in each year (4;0 – 4;11, 5;0 – 5;11, 6;0 – 6;11, 7;0 – 7;11) and nine 8-year-olds (all of whom were in second grade). The children were divided approximately evenly between male (n = 54) and female (n = 51) overall, and within each age group. However, 19 children were excluded from the analyses because they did not provide analyzable language samples and an additional three children were excluded because they did not produce any AAE features on their language samples. These 22 children are not included in table 1 or in any of the analyses. Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 8 morphological and phonological features of AAE was evaluated from an informal language sample; all 83 participants used at least one feature of AAE. All children were typically developing, according to parent report, and children with IEP’s were excluded from the study. All children passed a hearing screening (25 dB at 500, 1000, 2000, and 4000 Hz) prior to testing. We asked all primary caregivers to complete a demographic questionnaire that included questions on the level of education of the primary caregiver and their total family income. This questionnaire was done in interview format or filled out independently, depending on the preference of the adult completing the form. All participants completed norm-referenced measures of expressive and receptive vocabulary (Expressive Vocabulary Test, 2 edition, EVT-2, Williams, 2006; Peabody Picture Vocabulary Test, 4 edition, PPVT-4, Dunn & Dunn, 2007, respectively), and a measure of receptive syntax, the Elaborated Phrases and Sentences subtest (EPS) from the Test of Auditory Comprehension of Language, 3 edition (TACL, Carrow-Woolfolk, 1999). Stimuli Dialect awareness task. The stimuli were recorded by six young adult female speakers of AAE and six young adult female speakers of MAE. All speakers of AAE were African American and were fluent dialect-shifters between AAE and MAE. All speakers of MAE were European American and did not speak AAE. We included a 3 With young children (such as the 4-year-olds in our sample), it is always difficult to differentiate between dialect features and developmental language features when a morphological feature is omitted or a consonant substitution or deletion is produced (e.g., Oetting, Cantrell, & Horohov, 1999). For the 9 children who produced only 2 or 3 dialect features, we examined their mean age and expressive vocabulary size to investigate whether these dialect features might actually be age-related developmental errors. This seems unlikely, as their mean age was the same as the group as a whole, but their mean EVT-2 standard score was higher (M = 6;3, SD = 0;3 for age; M = 101, SD = 2 for EVT-2 standard score). All other participants used at least 6 dialect features. Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 9 relatively large number of speakers for the two dialects to facilitate generalization in terms of dialect differences rather than speaker differences and so that we could manipulate the familiarity/novelty of the speakers. As described below, three speakers of each dialect were included in the training phase, while the test phase included the three familiar speakers of each dialect as well as three novel speakers of each dialect. The speakers were recorded reading two children’s books, A Snowy Day and A Letter for Amy (both by Ezra Jack Keats). For the AAE versions of the two stories, we developed AAE scripts (written by an AAE speaker) that contained both morphological and phonological features of AAE. However, as AAE is primarily a spoken rather than a written dialect and use of morphological and phonological features is optional rather than obligatory, we did not require the AAE speakers to follow the scripts exactly. Instead, the AAE speakers were simply instructed to read the stories in AAE, using the scripts as a guide. These recordings were divided into oneto two-sentence chunks. All chunks were normalized for amplitude across all 12 speakers. For the AAE versions of the stories, all oneto two-sentence chunks were rated by an AAE speaker on a five-point scale (one = sparse dialect use and five = dense dialect use). Only stimuli that were rated as four or five were included in the study. Once the AAE oneto two-sentence chunks had been selected, we chose the same MAE oneto two-sentence chunks. The visual stimuli for the dialect awareness task were six blue and six red cartoon “monsters.” Each monster was paired with two voices, one AAE voice and one MAE voice. Assignment of each dialect to blue vs. red monsters was counterbalanced across participants so that red monsters were associated with the AAE dialect for half of the participants and with the MAE dialect for the other half. Each monster/voice pair was Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 10 animated so that it looked as if the monster was “speaking” the sentences (that is, the monster’s mouth opened for vowels and closed for consonants). Each monster had distinct visual characteristics in addition to color in order to maintain children’s interest in the task and to help children make the generalization that there are different blue monsters who speak one way and different red monsters who speak a different way. Comprehension task. The stimuli for the comprehension task were chosen to highlight either a phonological or morphological contrast between AAE and MAE. The phonological contrast is that, in AAE, the final /t/ or /d/ in a word-final consonant cluster may be deleted if the previous consonant agrees in voicing (e.g., hold is produced /hol/, mist is produced /mIs/; Guy, 1980; Craig et al., 2003). Final consonant cluster deletion occurs in both AAE and MAE (particularly in spontaneous speech when the subsequent word is consonant-initial), but is more frequent in AAE. The morphological contrast is that the plural morpheme is optional rather than obligatory in AAE if another number word is present (e.g. two cat, fifty cent) (e.g., Washington & Craig, 2002). The stimuli for the comprehension task were 18 word pairs (9 for the phonological contrast and 9 for the morphological contrast) that differed only in the presence or absence of a final consonant cluster. We included word pairs such as goal/gold for the phonological contrast and cat/cats for the morphological contrast. Appendix A provides a list of all stimuli for this task. Insofar as possible, all words were familiar to young children and pictureable. Because there are a limited number of possible word pairs for the phonological contrast, some of the target words were less familiar than others (e.g., bill and coal are less familiar than ball and bus) and some items were less pictureable than others (e.g., start and hold are less pictureable than bald and belt). For this reason, a Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 11 word/picture familiarization phase was included for this task (see procedure below). Recordings of stimulus items by an AAE speaker were used in the familiarization phase because we wanted to familiarize children with the object-name/picture pairings in their native dialect. The target words were recorded in the phrase, “Say _____ please,” so that the final consonant cluster was always followed by a word beginning with a consonant, as this is a phonological context that encourages final consonant cluster reduction. For the phonological contrast, final consonants in clusters were deleted or produced as glottal stops (e.g., gold was produced as “gol”). Nonetheless, words with singleton final consonants (mean duration = 5525 ms) were consistently shorter than words with reduced final consonant clusters (mean duration = 6019 ms). For the morphological contrast, the plural /s/ was produced on all items, but some consonants were deleted (e.g., clouds was produced without the /d/; lights was produced without the /t/). As with the phonological contrast, words in the singular form (mean duration = 5852 ms) were consistently shorter than words in the plural form (mean duration = 6281 ms). In the test phase, the stimuli were presented in MAE and all final consonants and consonant clusters were clearly articulated. Color photographs representing each word were used as visual images. The words were recorded by a young adult female speaking AAE for the training phase and by another young adult female speaking MAE for the test phase. All words were spoken in the carrier phrase say ______ please (familiarization phase) or show me ______ please (test phase), as word-final consonant cluster reduction is more frequent preceding a stop consonant. Words were normalized for amplitude, separately for the AAE and the MAE speakers. Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 12 Procedure General. All children participated in two or three test sessions of about an hour each with breaks. Primary caregivers came with their children and completed the demographic questionnaire. The first session began with the hearing screening and a language sample. Dialect awareness task. For each participant, all AAE monster-voice dyads were assigned to one color (either red or blue) and all MAE monster-voice dyads were assigned to the other color. Color assignment varied randomly across participants, with 50% of the participants receiving red-AAE/blue-MAE monster-voice dyads and the other 50% receiving the opposite pairing. The dialect awareness task included a training phase and a test phase. In the training phase, three monster-voice pairs for each dialect were presented. On each training trial, a red monster and a blue monster were presented on the opposite sides of a computer touch screen. Each monster, first the one on the left side and then the one on the right, would “say” the same oneor two-sentence chunk of the book, A Snowy Day. The first 20 sentences of the story (approximately two-thirds of the story) were presented in order, with each one or two-sentence story chunk presented twice (once in MAE and once in AAE). After each monster “talked,” the participant was asked to “point to the monster that talked” on the touch screen. The participant could always tell which monster was “talking” from the animated lip movements. After a oneor two-sentence story chunk had been presented in both dialects, the next trial would begin. In each subsequent trial, the next oneor two-sentence story chunk was presented in both dialects. The position of the red and blue monsters on the right or left side of the screen varied across trials. There were 11 training trials altogether (with each oneor Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 13 two-sentence chunk presented in both AAE and MAE). The children were not shown the books as the monsters were talking, and they were not told that the sentences they heard were from storybooks in either the training or the test phase. We did not ask the children or their caregivers if they were familiar with the two stories. In the test phase, six monster-voice pairs for each dialect were presented: three were familiar because they had been presented in the training phase and three were novel because they had not been presented before. In the test phase, as in the training phase, one red and one blue monster were presented on each trial, and the participant heard a oneor two-sentence chunk of a new story (A Letter to Amy). In the test phase, as in the training phase, the oneor two-sentence chunks were presented in the order of the story. The task was the same (“point to the monster that talked”), but in the test phase, the child heard the story segment only once (in either AAE or MAE) on each trial, and the monsters were not animated so that the child had no visual cues about which monster was talking. The only way that the participant could answer correctly was if he/she had made the generalization from the training phase that red monsters speak AAE and blue monsters speak MAE (or vice versa). There were 34 test trials. The test phase was preceded by six practice trials using sentence chunks from A Snowy Day, three of which had been presented in the training phase and three of which were from a later part of the story that was not presented during training. The voice-monster dyads during the practice phase were the same six speakers from the training phase. The children were provided with scripted feedback (e.g., “That’s not quite right. It was the red monster”). Responses were recorded on the touch screen and scored automatically. Comprehension task. The comprehension task included a familiarization phase Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 14 and a test phase. The purpose of the familiarization phase was to ensure that participants were familiar with all of the object-names and all of the object name/picture pairings. In this phase, a picture was shown on a computer screen and the digitized recording of the name associated with the picture was presented in AAE, “Say ___, please.” Immediately after this prompt, the participant named the picture. On the rare occasion that a participant forgot a picture-name, he/she was prompted again with the picture-name in AAE and asked to repeat it (“Say __, please”). In the test phase, the participant was presented with a randomly sequenced array of three pictures (target, distractor, foil as in goal, gold, bus or cat, cats, bill) and was asked to “Show me ___.” See Appendix A for a list of the foils for each target/distractor pair. The test phase was preceded by three practice trials to ensure that the child understood the task. There were 36 trials altogether; each stimulus pair (e.g., cat/cats) was presented twice, once with each member of the pair as the target. Because the task was presented in MAE, a response was considered correct only if it was correct in MAE (e.g., a child needed to point to cat if the prompt was cat and to coal if the prompt was coal). Responses were recorded by the child touching the picture on a touch screen and were scored automatically. Dialect density. A 50-utterance language sample was elicited from all participants and recorded. The language sample was elicited in a conversational context (e.g., “What did you do last weekend?” “What’s your favorite TV show?” etc.). The language samples were elicited by an AAE-speaking examiner. Data Reduction and Analysis For the two experimental tasks, the dependent variable was accuracy at the trial level (correct/incorrect). Both raw scores and standard scores were obtained from the Downloaded From: http://jslhr.pubs.asha.org/ by a Health Sci Learning Ctr User on 08/14/2014 Terms of Use: http://pubs.asha.org/ss/Rights_and_Permissions.aspx Dialect awareness and lexical comprehension of MAE 15 EVT-2, the PPVT-4, and the EPS subtest of the TACL. Average scores for each age group are presented in Table 1. Information from the demographic questionnaire about both education level and family income category was converted to z-scores. We used the average z-score of these two indicators as our measure of SES in the statistical analyses (NCES, 2001). Four primary caregivers did not complete questions about either education level or total family income (n = 3 for education level and n = 1 for family income). All caregivers completed at least one of these two questions. In these cases, the missing data for these 4 participants were imputed using the regression imputation method across the entire data set of 83 participants (Saunders, Morrow-Howell, Spitznagel, Dore, Proctor & Pescarino,