LANGUAGE DEVELOPMENT OF SAME-SEX MULTIPLE-BIRTH SIBLINGS: A RETROSPECTIVE NATURE/NURTURE STUDY Except where reference is made to the work of others, the work described in this disertation is my own or was done in collaboration with my advisory commite. This disertation does not contain proprietary or clasified information. __________________________ Linda W. Holladay Certificate of Approval: ______________________ ______________________ Thomas E. Nunnaly Robin Sabino, Chair Asociate Profesor Asociate Profesor English English ______________________ ______________________ Elaina M. Frieda Joe F. Pitman Asistant Profesor Interim Dean Communications Graduate School LANGUAGE DEVELOPMENT IN SAME-SEX MULTIPLE-BIRTH SIBLINGS: A RETROSPECTIVE NATURE/NURTURE STUDY Linda W Holladay A Disertation Submited to the Graduate Faculty of Auburn University in Partial Fulfilment of the Requirements for The degre of Doctor of Philosophy Auburn, Alabama December 15, 2006 ii LANGUAGE DEVELOPMENT IN MULTIPLE-BIRTH, SAME-SEX SIBLINGS: A RETROSPECTIVE NATURE/NURTURE STUDY Linda W. Holladay Permision is granted to Auburn University to make copies of this disertation at its discretion, at the request of individuals or institutions at their expense. The author reserves al publication rights. ________________________ Signature of Author ________________________ Date of Graduation iv VITA Linda Holladay is a graduate of Spearfish High School in Spearfish, South Dakota. She received her Master?s from the University of Nebraska, Omaha, and entered the Ph.D program at Auburn University, Alabama, in 1996. She is maried to retired Air Force Colonel Cecil Holladay and has two sons, Andrew and Mathew. She currently teaches linguistics and composition at Alabama State University in Montgomery, Alabama. v DISERTATION ABSTRACT LANGUAGE DEVELOPMENT IN SAME-SEX MULTIPLE-BIRTH SIBLINGS: A RETROSPECTIVE NATURE/NURTURE STUDY Linda W. Holladay Doctor of Philosophy, December 15, 2006 (M.A. University of Nebraska at Omaha, 1981) (B.S. Black Hils State University, 1971) 122 Typed Pages Directed by Robin Sabino This retrospective study considers the extent to which nature and nurture influence early first language acquisition. Data sets that addres rate of gramatical acquisition, acquisition of regional dialect features, and temperament are examined. The subject population consists of four multiple-birth, same-sex siblings born to a European-American, college-educated couple living in the South. Two of the siblings are monozygotic (identical DNA), and the other two are dyzygotic (at least 50% shared DNA). Pre-existing audio tapes recorded when the children were ages 2 through 6 were examined using Mean Length of Uterance (Brown, 1973) and Developmental Sentence Structure (Le, 1974) to determine gramatical development. Tapes also were examined vi for acquisition of two regional phonological features: monophthongization of /ai/ and variants of (ing), and two lexico-syntactic features, yall and fixin to. Carey Temperament Scales, completed by the parents, were used as a third estimate of similarity or diference. Although the retrospective nature of the study limits the data available for analysis, results show that the monozygotic siblings are more similar to one another in both gramatical acquisition and regional forms than are the dyzygotic siblings. Acquisitional diferences appear more pronounced at earlier ages, indicating that genetics may be more influential in earlier years. vii ACKNOWLEDGEMENTS This project could never have been without the help, support, and encouragement of a myriad of family, friends, and co-workers. I first want to thank Dr. Cindy Bernstein for piquing my interest in linguistics and Dr. Nathan Moore, whose very sage advice, ?Pick something you love,? I remembered often in the early morning hours. Many thanks go to Dr. Teri Teranova, clinical psychologist, and Mrs. Jeannie Lanier, child counselor, both of whom advised me in the area of child temperament. Thanks to Dr. Richard Emanuel and Mr. Mark Hil for statistical help and to Mrs. Anna Head Oggs, for research help. Thanks also to my wonderful commite members, Dr. Robin Sabino, Dr. Thomas Nunnaly and Dr. Elaina Frieda, for being enormously helpful and a joy to work with. Family and friends are too numerous to name, and some of them ay never know the times they were there with an encouraging word just when needed. There is no way I can adequately thank my swet husband for his love, patience, support and fast food runs. My very, very, special thanks goes to these amazing children and their wonderful parents. They are, or course, the dream stuff of the linguistic researcher. Finaly, many, many thanks to Dr. Robin Sabino, my commite chair, who not only guided me along the way, but was also teacher, advisor, and friend. I look forward to continued collaboration with her. vii Style manual or journal used: Publication Manual of the American Psychological Asociation Computer software used: Microsoft Word 2004 for Macintosh Excel 2004 for Macintosh JMP IN 4 for Windows CTS Quickscore Software V4.2P for Windows Speech Analyzer 2.6 for Windows ix TABLE OF CONTENTS LIST OF TABLES xii LIST OF FIGURES xiv CHAPTER 1 INTRODUCTION 1 CHAPTER 2 REVIEW OF THE LITERATURE 4 2.0 Introduction 4 2.1 Twin studies with implications for the efects of nurture and nature on language development 4 2.2 Temperament studies with implications for language development 8 2.3 Asesment of language development 12 2.4 Temperament Asesment 14 2.5 Acquisition of regional dialect features 14 2.5.1 Phonological development of /ai/ 18 2.5.2 Acoustic analysis of /ai/ 20 2.5.3 Southern monophthongization of /ai/ 20 2.5.4 The variable (ing) 23 2.5.5 Yall and fixin to 24 2.6 Summary 25 CHAPTER 3 METHODOLOGY 28 3.0 Introduction 28 3.1 Selection of subjects 29 3.2 Data collection 30 3.3 Transcription 30 3.4 Analysis of gramatical development 31 3.4.1 Mean Length of Uterance (MLU) 31 3.4.2 Developmental Sentence Scores (DS) and Developmental Sentences Types (DST) 32 3.5 Analysis of regional dialect features 35 3.5.1 Analysis of variable /ai/ 36 3.5.2 Analysis of variable (ing) 39 x 3.5.3 Yal and fixin to 40 3.6 Temperament Analysis 40 3.7 Comparison of language development 42 CHAPTER 4 RESULTS: GRAMATICAL DEVELOPMENT 43 4.0 Introduction 43 4.1 MLU results 43 4.2 DS results 47 4.2.1 DST scores 48 4.2.2 DS scores 52 4.3 Summary 53 CHAPTER 5 RESULTS: REGIONAL DIALECT FORMS 55 5.0 Introduction 55 5.1 Description of /ai/ variants 56 5.2 Analysis of /ai/ variation 58 5.2.1 Pre-diphthongal analysis 58 5.2.2 Post-diphthongal analysis 62 5.2.3 Comparison of post-diphthongal analysis of D2 and M1 67 5.3 The variable (ing) 67 5.4 Yall and fixin to 70 CHAPTER 6 RESULTS: TEMPERAMENT ANALYSIS 71 6.0 Introduction 71 6.1 Scoring 71 6.2 Temperament profile analysis 73 6.3. Comparison of temperament characteristics and language acquisition scores 76 CHAPTER 7 DISCUSION 81 7.0 Introduction 81 7.1 Gramatical Analysis 82 7.1.1 MLU analysis 82 7.1.2 DST and DS analysis 83 7.1.3 Comparison of MLU and DST/DS scores 84 7.2 Regional dialect forms 85 7.2.1 Analysis of monophthongization 86 7.2.2 Analysis of (ing) 88 7.2.3 Analysis of yall and fixin to 88 7.2.4 Acquisition of regional dialect forms 88 xi 7.3 Temperament analysis 89 7.4 Limitations of the study 90 7.5 Implications for future research 92 7.6 Conclusions 93 REFERENCES 95 APENDIX A Computation of MLU 105 APENDIX B Developmental Sentence Scoring Rubric 107 xii LIST OF TABLES CHAPTER 3 Table 3.1 DS computation of Can you do this and this? 35 CHAPTER 4 Table 4.1 Number of utterances for each child at each age 44 Table 4.2 MLU for each subject at each age 44 Table 4.3 T-test results of MLU scores 46 Table 4.4 T-test results on LU for the pooled MZs and D2 47 Table 4.5 Total utterances and usable utterances for DST and DS scoring 48 Table 4.6 Sentence structures (S), unique utterances (U), and DST scores 49 Table 4.7 Chi-Square analysis of paired DST scores 49 Table 4.8 D1?s pre-sentence constructions 51 Table 4.9 M1?s pre-sentence constructions 51 Table 4.10 2?s pre-sentence constructions 52 Table 4.11 Reliable DS scores for each child at each age 53 CHAPTER 5 Table 5.1 Monophthong production at age 2 56 Table 5.2 [!] production at ages 3 and 4 57 Table 5.3 Percent of diphthong use for each sibling at each age 58 Table 5.4 Pre-diphthongal analysis of /ai/ variation by following phonological environment 59 Table 5.5 Pre-diphthongal analysis of /ai/ variation by word clas 60 Table 5.6 Post-diphthongal analysis of /ai/ variation by following phonological environment 63 Table 5.7 Comparison of hierarchies of following sound 64 Table 5.8 Post-diphthongal analysis of /ai/ variation by word clas 65 Table 5.9 Velar and alveolar variants of (ing) for ages 2-6 68 Table 5.10 Distribution of (ing) variation by word clas 69 CHAPTER 6 Table 6.1 Raw scores for temperament characteristics 72 Table 6.2 Interpretation of temperament characteristics 72 xii Table 6.3 Temperament scores for PERSISTENCE, SENSORY, APROACH and ACTIVITY 74 Table 6.4 Ranking of raw scores on temperament characteristics 76 Table 6.5 MLU, DS, acquisition of [a:], and CTS rankings of the DZ siblings 78 Table 6.6 MLU, DS, acquisition of [a:], and CTS rankings of the DZ siblings 78 xiv LIST OF FIGURES CHAPTER 4 Figure 4.1 MLU progresion by age 45 Figure 4.2 Z MLU patern 46 Figure 4.3 DZ LU patern 46 CHAPTER 5 Figure 5.1 Acquisition of Southern /ai/ monophthongization 66 CHAPTER 6 Figure 6.1 Comparison of sibling temperament scores 73 Figure 6.2 Temperament comparison for PERSISTENCE, SENSORY, APROACH, and ACTIVITY 74 1 CHAPTER 1 INTRODUCTION Nurture vs nature is a central isue among linguists, who often disagre as to the relative contributions of genetics and environment to the acquisition of language. Nurturists argue from the perspective of the Vygotskyan theory that language acquisition is largely a product of social interaction, and that the environment of the learner is of more importance than heredity. Nurturists acept the premise that heredity may provide cognitively healthy children with an ability to produce language, but do not believe children are geneticaly endowed with a cognitive structure dedicated solely to the acquisition of language. From the nurturist perspective, linguistic behavior, modeled by the speech community, provides adequate linguistic input. As children interact with their speech communities, they produce increasingly complex linguistic structures. Therefore, the linguistic environment in which a child is socialized is the prime impetus for language acquisition. From this perspective, children learn not only linguistic functions, but also social norms and language values. The naturist theory, in contrast, derives from Chomskian theory which holds that the ability to learn language is biologicaly innate; al healthy individuals have the ability to learn language since the brain is geneticaly encoded to acquire language. Acording to naturist thought, the environment surrounding the child merely produces the necesary data that the language acquisition device uses to set language parameters. As evidence 2 for their position, naturists point to the fact that children produce constructions to which they have not been exposed, such as overegularization of past tense morphemes (I runned fast). Naturists argue that children create a complex system of rules despite the poverty of the input. This biological approach explains the universal nature of language acquisition. Although many first language acquisition researchers reject the nativist position, they nevertheles agre that child language acquisition shows remarkable cross-linguistic similarity. Al children, regardles of language, learn first utterances at about one year of age with gramar emerging similarly across languages. Moreover, despite the seming polarity of views, neither naturists nor nurturists contend that language acquisition occurs in a vacuum. The debate centers on predominance of domain and which specific aspects of language are afected by either nature or nurture and to what degre. Two problems have been encountered by researchers when atempting to determine the efects of nature or nurture on the acquisition of language. Studies using controlled environments do not acount for possible genetic diferences, and studies using children with diferent genetic codes do not acount for diferences in environments. A controlled environment and a subject data set which contains both children with identical genetic structures and children with non-identical genetic structures is necesary to determine which influence predominates. Such a controlled environment has been dificult to find. No child is born into exactly the same environment as any other child. A first- born child is born into a pre-existing family of two parents. A second-born child, 3 although born to the same parents, wil be born into a diferent environment, one consisting of two parents and an older sibling. The two environments share some but not al characteristics. In addition, the genetic make-up of two such siblings difers, making it impossible to determine whether the diferent environments or the diferent gene structures are the predominant influences on the acquisition of language. Identical twins solve the problem of identical genetic structure, but since identical twins are also born into an identical familial relationship, it is dificult to tel whether the identical genetic structure or the similar environment most afects language acquisition. An additional set of siblings with non-identical genetic make-up (fraternal twins) who are born into the same family at the same time would provide a comparison to the set of identical twins. One possible data set, therefore, would be a corpus produced by a set of same-sex quadruplets consisting of one set of identical twins and one set of fraternal twins. This study examines the language acquisition of just such a data set, a corpus produced by four, multiple-birth, same-sex siblings consisting of one set of identical twins and one set of fraternal twins. Linguistic features of particular interest to this study are rate of language acquisition and acquisition of regional dialect features. In addition to analysis of the language data, this study considers genetic temperament characteristics as a possible explanation for language similarities and/or diferences. 4 CHAPTER 2 REVIEW OF THE LITERATURE 2.0. Introduction The review of the literature for this study addres both non-linguistic and linguistic variables. Studies of non-linguistic variables are those examining external factors such as the environmental and genetic impact on the rate of child-language development. Studies of linguistic variables wil include those examining the phonological and syntactic conditioning of regional dialect forms. Much work has been done concerning the rate of child language acquisition, and these studies can often shed light on the nurture/nature debate. Very litle, however, has been studied concerning the extent to which regional dialect features are afected by either nurture or nature. The review of the literature wil look at twin studies with implications for either nature or nurture, temperament studies with implications for language development, studies on the rate of child language development, and studies on acquisition of Southern monophthongization of /ai/, variable (ing) and lexico-syntactic features yall and fixin to. 2.1. Twin Studies with implications for the efect of nurture and nature on language acquisition Researchers seking to resolve the nature/nurture debate have studied the language development of twins as a means of examining the relative contributions of 5 genetics and environment. Comparing sets of twins or comparing twins with singletons has alowed researchers to identify diferences betwen identical twins (refered to in this study as monozygotic, or MZ) who share 100% of their genetic make-up, and fraternal twins (refered to in this study as dyzygotic, or DZ) who share at least 50% (Dale, Eley, & Plomin, p. 622). Since the twin environment is largely equivalent, diferences in language acquisition have ben atributed to the diferent genetics of the twin sets. Many of these studies concentrate on the divided atention situation endemic to the twin situation (Conway, Lytton & Pysh, 1980; Staford, 1987; Tomaselo, Mannle & Kruger, 1986). That is, while a mother of a singleton can direct al her atention to one child, a mother of twins must direct atention to two. This diminished individual atention has been considered a cause of slower language development in twins. Conway, Lytton, and Pysh (1980) studied 12 sets of twins and 24 singletons, age 2:6, from similar socioeconomic backgrounds. 1 A two-hour spontaneous speech sesion was taped for each twin set. The study reports that the amount of the mother?s speech to the child was the most important predictor of the child?s language performance. Children from singleton/mother dyads verbalized more; children from twin and mother triads verbalized les. The authors interpreted their results as indicating that the environmental factors surrounding the singleton/mother dyads encouraged verbalization. Tomaselo, Mannle, and Kruger (1986) observed six twin pairs and 12 singletons, al white, middle clas children, at age 15 months and again at 21 months. The results of their study were similar those of Conway, Lytton and Pysh (1980) in that twins scored 1 Specific socioeconomic information was not given. 6 lower than singletons on al measures of language development; however, the amount of verbalization in the environment was not found to be a mitigating factor in this study. Rather, Tomaselo et al. report that although mothers of twins spoke as much as mothers of singletons, twin children had les speech directed to them individualy, a diference that developed from the triadic situation itself. Staford (1987) monitored the conversational characteristics of college-educated mothers and 24-36 month-old twins and singletons. This study included 22 mothers of twins and 22 mothers of singletons, al from middle-clas families. As with previous studies, twins scored significantly lower than singletons on tests involving language expresion and comprehension. There was also a correlation betwen both the qualitative and the quantitative maternal input and the children?s test scores. Staford reports that twins received les responsive, conversation-eliciting, maternal speech than did singletons. Although the above studies are useful in identifying contrasts betwen twins and singletons that may indicate an environmental impact, they have not addresed the question of nature. Studies which compare MZ and DZ twins do indicate a genetic efect on language acquisition. Matheny and Bruggemann (1972) tested the articulatory development of 263 twins and 94 singletons from families of twins. Subjects ranged from thre years to eight years old and were from a range of socioeconomic backgrounds. Using the Templin- Darly Screning Test of Articulation, Matheny and Bruggemann found that the mean scores of twins were significantly below the standard norms. Additionaly, they found 7 greater similarity in articulation betwen MZ twins than betwen DZ twins. Since siblings within the same family group share much of their environment, Matheny and Bruggeman atributed the diference in articulation to genetics. 2 Locke and Mather (1989) analyzed speech samples from 26 pairs of 4-year-old MZ and DZ twins from white, middle-clas Midwestern backgrounds, and found that MZ twins were more likely to produce substitutions, omisions or distortions for the same sounds on an articulation test than were DZ twins, whose utterances were judged to be similar to unrelated children. Pollack and Keiser (1990) also report that a set of identical twins paterned closely in vowel eror production. More recent studies, such as Ganger?s (1998) disertation, show a genetic influence in acquisition of lexicon and inflectional morphology. Her case study of 43 MZ and 33 same-sex DZ twins shows that MZ twins have similar rates of vocabulary growth and similar acquisition of gramatical categories, whereas DZ twins show greater divergence from one another. Ganger, Wexler and Soderstrom (1998) studied an optional infinitive 3 stage of language acquisition in five pairs of MZ and four pairs of DZ twins. The twins were studied longitudinaly from age 1:7 to 3:10. They report that for the children studied, the optional infinitive stage was more similar in MZ twins than in DZ twins and interpret their findings as indicating a posible geneticaly-driven stage of acquisition. 2 Their study indicates that birth position in the family and socioeconomic status can also afect articulation in both twins and singletons. 3 Defined as ?a period in early child language during which children know the properties of tense and agrement but do not treat the use of these features as obligatory? (224). 8 2.2. Temperament studies with implications for language development A second domain of interest to this study concerns personality and the efects of temperament on language development. Temperament diferences, such as shynes and persistence, are thought to be geneticaly influenced (Wilson & Matheny, 1986; Martin, Drew, Gaddis & Moseley, 1988; Plomin, 1994; Dale & Goodman, 2005), and correspond to the ways a child responds to situations, such as curiously or fearfully, being easily distracted or able to focus atention wel. Thus a temperament efect on language development represents a possible genetic influence, and as such, is relevant to the nature/nurture debate. Questions regarding language, personality development, and motivation have been explored with studies on referential and expresive learning styles. Haynes and Shulman (2000) reference studies by Katherine Nelson in 1973 which first showed that children can be roughly divided into what she refered to as two learning tracks: referential learners, whose early vocabulary includes relatively more nominals, and expresive learners, whose vocabulary includes relatively more personal/social words. 4 Studies by Delacorte, Benedict, and Klein (1983) confirmed Nelson?s findings and went on to connect the expresive style of the mother with the learning style of the child. In the Delacorte et al. study, in-home recordings of 60 mothers were used to clasify maternal utterances acording to communicative intent, focus of atention and evaluation. 4 Other researchers, e.g. Bretherton, 1983; Delacorte, Benedict and Klein, 1983, maintain that children are not easily defined by tracks. They report that most children use a fairly balanced approach and that there is no indication of an advantage of one track over another. 9 They found that mothers of referential children produced a greater number of utterances, and those utterances were more descriptive and les prescriptive than the uterances of mothers of expresive children. Although the study can be interpreted as demonstrating an environmental link to the tracking style of the child, there was no control in the study to preclude a genetic connection. A possible genetic connection to language development may have been documented by Bretherton, McNew, Snyder and Bates (1983) who tested the vocabularies of 30 children at 20 months and again at 28 months. They found that expresive learners produce more pronomials in early vocabulary (first 50 words), are more physical, have more activity-oriented words, and are predominately male. Malenes, however, must not be automaticaly asociated with a genetic link. Although a sex diferentiation indicates a genetic influence, a gender diference reflecting diferent social expectations indicates an environmental influence. Other researchers have identified a genetic influence on language development (e.g. Martin and Holbrook, 1985; Martin, Drew, Gaddis, and Moseley, 1988; and Plomin, 1994). Martin and Holbrook found significant correlation betwen the behavior characteristics of persistence, activity level, adaptability, and approach/withdrawal and 104 first graders? reading scores on both clasroom and standardized achievement tests. In thre separate studies reported in Martin et al., distractibility, persistence, and activity level were reported to have the highest degre of correlation to standardized test scores in 243 kindergarten through second graders. These researchers indicate that behavioral characteristics (i.e. temperament) are inherited. Plomin argues that such characteristics 10 are instrumental in creating a learning environment. He maintains, ?People to some extent choose their environments; the environments chosen are influenced by, not ?determined by,? the genotypes of those who choose? (p. 2). Shynes has also been argued to impact language development. Wong-Filmore (1979) observed the social interaction of five Spanish children, age 6 to14, while they learned English as a second language. Although she did not quantify the data, her observations indicated that the child who was most outgoing and who produced language without regard for correctnes learned English more quickly. Those who were les outgoing produced les language and took longer to learn. Wong-Filmore hypothesized that the risk-taking personality socializes more easily and that greater socialization acelerates language learning. Although socialization may alow for the input needed for language development, in Wong-Filmore?s study it is the risk-taking temperament which creates the environment that acelerates the proces. Goldfield and Snow (1989) came to similar conclusions concerning risk taking and first language acquisition. They found that shy children make fewer erors, but that they talk les and produce les complex speech. Faster learners talk more, make more erors, but sem to learn from those erors and achieve greater fluency. Paul and Kelogg (1997) report a relationship betwen shynes and slower rates of acquisition. Twenty-eight children who had been identified by parents and clinicians as delayed in language development 5 were rated acording to approach/withdrawal 5 Fewer than 50 words at 20 months 11 behavior. The children?s behavioral measure positively correlated with mean length of utterance (MLU). In addition, Paul and Kelogg found a positive correlation betwen physical activity and language development. They suggested that the results were due to the same basic source: shynes resulted in lower motivation to initiate conversation, which, in turn, resulted in les communication. High physical activity resulted in les time to communicate. Other correlations betwen language acquisition and personality are reported by Dixon and Smith (2000). Toddler Temperament Scales were completed on 40 toddlers at age 13 months. This was followed by the Communicative Development Inventory at 20 months. Positive correlations were found to exist betwen stable emotional temperaments at 13 months and later language development at 20 months. Dixon and Smith suggested that children with more stable temperaments are beter able to focus atention, alowing faster acquisition. The authors theorized that focused atention alowed the child to initiate more mother-child interaction, facilitating acquisition. Although the studies discussed here on rate of acquisition show a connection betwen language learning and personality, they generalize among geneticaly unlike subject pools. Thus questions concerning the efects of nature or nurture remain. It is possible, for example, that risk-taking is a learned trait rather than an inherited one, or that temperament shapes mother/child interaction. What is needed is study on geneticaly similar subjects from socialy similar environments. 12 2.3. Asesment of language development A number of instruments have been used to ases child language development. Two of interest to this study are Mean Length of Uterance (MLU) and Developmental Sentence Scoring (DS). MLU, developed by Brown in 1973 as a diagnostic tool to ases child language acquisition, determines the average number of morphemes a child speaks per speech episode. 6 Brown considered evaluation of MLU a valid measure of morphemic and syntactic level up to age 4. However, this widely used measure of language development is not without critics. Although Brown gives fairly comprehensive guidelines for computing MLU, Crystal (1974) criticizes Brown?s guidelines as incomplete. Kle and Fitzgerald (1985) studied 18 normaly developing two and thre-year-old, white children from English-speaking homes to determine if MLU alone was a valid indicator of syntactic development. Excluding single-word morphemes (e.g. yes, no, etc.), Kle and Fitzgerald report that MLU increases predictably as a child ages, but that the results were not consistent across an age range. They argue that there is no predictable relationship betwen MLU in the 24 -48 month range and suggested that MLU is of limited value beyond Brown?s stage I (age 2.0-2.5). Using a larger sample, Blake, Quartaro and Onorati (1993) refuted Kle and Fitzgerald?s results. Recording spontaneous speech samples from 87 English-speaking children, Blake et al. found an age correlation through age 4.5, reporting, ?The measure 6 An episode is the approximate equivalent of a sentence. (Se appendix A). 13 of syntactic complexity was not found to have a ceiling in terms of reflecting clausal complexity; and thus, it is a promising measure of gramatical complexity which is also relatively easy to apply to spontaneous speech samples? (151). High correlations betwen age and MLU are also reported by Miler and Chapman (1981) and Rondal, Ghiotto, Bredart, and Bachelet (1987). Interestingly, Miler and Chapman?s results with 123 children validated MLU through age 5, although they reduced the number of utterances by half observing that ?a minimum of 50 inteligible uterances or 20 minutes of conversation was required? (156). In spite of some criticism, then, MLU has become a standard measurement of child language development, as witnesed by its use in the studies of Conway, Lytton and Pysh (1980), Paul and Kelogg (1997), Roberts (1994), Shatz (1994) and others. A second highly respected and widely used asesment tool of child gramatical development, normed on a group of 200 normaly developing children, is Developmental Sentence Scoring (DS) (Le, 1974). Larsen (1975) of the University of Texas, Austin, describes the DS as ?a significant work on asesing children?s use of gramatical rules in Standard English, and is useful to any profesional working with children with language disorders? (467). However, although DS is used by clinicians to diagnose language delay and aceleration, Le, the creator of the testing measure, does not limit DS to that purpose. She writes, ?Developmental Sentence Analysis is a method for making a detailed, readily quantified and scored evaluation of a child?s use of standard English gramatical rules from a tape-recorded sample of his spontaneous speech in conversation with an adult? (xix). 14 Although to date, DS has not been used in twin studies, it has been used to ases normal children in conversation and picture-description experiments (Haynes, Purcel, & Haynes, 1979). It was also used in Paul and Kelogg?s (1997) study of ?Temperament in Late Talkers? as one means of establishing level of language development. 2.4. Temperament assesment Many of the studies referenced in section 2.4 are based on categories established by the New York Longitudinal Study (NYLS) of Thomas, Ches, Birch, Hertzig and Korn (1963). This seminal work (Wilson & Matheny, 1986) is no longer available. However, the Carey Temperament Scale (CTS), which is highly correlated with the NYLS (Carey Temperament Scale, 1996-2000) and which has been adapted for use on children from one month to 12 years of age, is available. CTS uses parental input to obtain data, a technique increasingly acepted as providing valid data on language and child development. Acording to Dale and Goodman (2005), ?One of the most striking developments in the study of child language, and indeed developmental psychology more generaly over the past 20 years, is the revival of parent report as a trustworthy, and trusted, research technique? (42). 2.5. Acquisition of regional dialect features Acquisition of child language has generaly been studied by considering groups of children acquiring Standard English forms; litle has been done on the acquisition of regional dialect features since these have generaly been sen as afected solely by environmental constraints (i.e. nurture). However, within the limited dialect acquisition 15 literature, no studies have been conducted which atempt to separate environmental from genetic impact. The possibility of a genetic component, such as temperament, that afects the acquisition of dialect cannot logicaly be ruled out. Among the studies relating dialect choices to social influences are Labov (1972a), Jorgenson (1998) and Berthele (2002), who describe language as an instrument of prestige and belonging. Labov?s study on Martha?s Vineyard shows the local population?s use of vowel centralization as a form of identity marking. His study of New York department stores (1972a) shows a correlation betwen socio-economic status and the perceived prestige of postvocalic [!] usage. His work among inner city youth (1972b) demonstrates how group members use language to establish and enforce group boundaries. Jorgensen?s (1998) study of taped conversations of Turkish imigrant tens in Denmark reported situational code-switching when the tens spoke to Danish adults as opposed to Turkish adults. He reports dialect forms when the Turkish tens are speaking to Turks that are not used in conversations with Danes and interprets his results to mean that speakers choose a linguistic code based on isues of power and solidarity. Berthele (2002) reports similar findings in his study of idiolectal disonance. He observes that speakers? choices of linguistic form negotiate social roles and power. Berthele showed, in his study of, 9-year old Swis, elementary school children acquiring a second dialect (N=14), that acommodation to a second dialect serves to mark identity in a group of peers. He interviewed the children and then taped them in their social clasroom seting to discover if there was a target variety and in what contexts it 16 occurred. Berthele indicates that even in same situations, the choice of variety depends on (1) individual psychological disposition, (2) age, (3) social style and (4) personality traits. He suggests that the degre to which an individual acquires a target dialect depends on how much the individual ses himself as part of the target group, and on the group?s aceptance of him/her. As with Labov?s studies of inner city youth, sociograms (analysis of social status within groups) showed greater acommodation to the dialect where there was greater peer aceptance of the child. The question of a target dialect was also addresed by Deser (1989) in a study of six Detroit families with children ranging from 9 to 20 years of age. Al were born in Detroit but had family members who had relocated to Detroit from the South. Individuals were asesed for degre of southern dialect retention. Within the same family unit, some children acquired Detroit speech while others acquired Southern dialect features. Although there were no clear paterns of parent/child modeling, Deser reports that the kinds of dialect models the child is exposed to?especialy in school?have a great impact on the dialect choice. Her evidence also indicated that in some cases dialect ties are stronger than family ties. Similar to Deser (1989), Starks and Bayard?s (2002) study of acquisition of postvocalic /r/ in four pre-school children in New Zealand found parental modeling to be of limited value in the acquisition of New Zealand dialect forms. They studied four children of rhotic parents ([r] producing) and found that thre of the four exhibited litle or no postvocalic [r] in spite of parental modeling of the form. 17 The above studies on the acquisition of a regional or ethnic dialect concentrate on older speakers or second language learners. Few researchers, however, addres early learners and their perceptions of a target language. Cole?s (1980) disertation on the development of dialect features indicates that the impact of dialect choice appears over time and that some dialect features are not produced by young learners. She studied 3 to 5-year-old (N=60) children of speakers who exhibited what she identified as ?moderate Black English? (BE). She found only eight of 19 BE forms to be present in pre-schoolers, but additional forms were found in older children. She concluded that children do not initialy exhibit ethnic dialect forms, but produce forms found in Standard English until the age of thre or four, at which time dialect forms may emerge. The findings sem to indicate that more than parental modeling is involved in the proces of acquisition and dialect choice. Haris? (2004) study of 60 speakers of African American English (AE) from age 4 to 15 helps to iluminate Cole?s finding. Haris found that although pre-school European American (EA) and African American (A) children develop the same phonological, morphological and syntactic features, A speakers retain features longer if the features are also found in AE dialects. Roberts (1994), who studied pre-school children in Philadelphia, shows that for t/d deletion (as in contractions with n?t or past tense verbs such as passed) and the variation betwen the velar and alveolar nasal (e.g. running vs runnin), children create their own rules before acquiring those modeled by parents. She studied daycare children (N=17), age 3-5, in South Philadelphia and found that 3 and 4-year-old subjects exhibit 18 t/d deletion and /ing/ in contexts diferent from adult usage and subject to diferent phonological and gramatical constraints. They were, she concluded, constructing variable rules and linguistic constraints of their own. In contrast to the above studies, in a study of 108 children in Philadelphia, Payne (1980) showed parental modeling, hence environment, was esential in acquiring Philadelphia short /a/. Only those children in her study whose parents were localy born and raised could afectively produce the complex Philadelphia short /a/ patern. Even those children who were born and raised in Philadelphia, but whose parents were non- local, were unable to completely acquire the phonological form. The above research shows that when several alternatives are available in the environment, dialect acquisition sems to be a mater of choice to some degre. However, some dialect features have been found to be so pervasive in certain regional dialects (Bailey and Tilery, 1996; Bernstein, 2003; Hay, Jannedy, and Mendoza-Denton, 1999; Maynor, 1996) that it is likely the features wil be found when studying the acquisition of child language within that region. The monophthongization of /ai/, the use of alveolar nasal [n] rather than veolar nasal [!] in nouns as wel as in verbs, and the use of yall and fixin to are al wel-researched indicators of Southern American English (Bailey and Tilery, 1996) and were thus considered to be of interest to this study. 2.5.1. Phonological development of /ai/ Although there are several studies of children with phonological disorders, litle research has been done concerning normal acquisition of English vowel forms. Even les has been done concerning the monophthongization of an existing diphthong. Fikkert (in 19 pres) holds that a child?s vowel system is not fundamentaly diferent from that of an adult?s. She ses the developing system as not alowing for al the contrasts in adult speech. Rather, children tend to produce sounds which are easier to articulate. Otomo and Stoel-Gamon?s (1992) study of four normaly developing children ages 22-30 months found that two of the earliest vowels to develop are [a] and [i]. Kehoe and Stoel- Gamon?s (1999) studies of 14 English-speaking children ages 1 to 3 found early development not only of [a] and [i], but also of the central lax vowel /!/. Diphthong development has been shown to follow initial monophthong development in studies by Bernhardt and Stemberger (1998). Thus, the development of [a] and [i] precedes the development of /ai/. They also found that diphthong production in closed sylables (those containing a final consonant sound) was dificult for children, and one child in their study produced no diphthongs until after age 2.5. Otomo and Stoel-Gamon (1992) found that substitutions of /ai/ for monophthongs were rare in young children. Pollack and Keiser (1990) studied phonological disorders in vowel production, but cite studies on normaly developing children as wel. Their findings based on 15 children, ages 3.8 to 6.4, studied for vowel production erors show that vowel reduction erors were more common than diphthongization. Pollack and Keiser also cite an older study by Welman (1931) which indicates that diphthong erors typicaly involve the substitution of a monophthong for a diphthong, and that monophthongization is a common vowel production eror in children. The above studies agre that ease of production is involved in early vowel development, that the low and mid-central vowels [a] and [!] appear early, and that 20 children are likely to substitute a monophthong for a diphthong, especialy in closed syllables. 2.5.2. Acoustic analysis of /ai/ Anticipating that spectrographic analysis might be useful in distinguishing monophthongal vowels from diphthongal vowels, several sources were consulted for conventions regarding formant analysis: Borden, Haris, and Raphael, 1994; Kent and Read, (1992); Ladefoged, Peter, 1996; and Carnel, 1997. Acording to Carnel, an adult southern monophthongal [a:] is indicated by F1 around 710 and an F2 around 1100 Hz. Diphthongal /ai/ is indicated by F1 around 640 Hz and F2 around 1190 Hz with marked movement toward 2250. Anderson?s (2002) formant analysis of monophthongization in African-American Detroiters found similar Hz levels. Anderson defined an upward movement of greater than 250 Hz as indicating a strong diphthongal vowel. 2.5.3. Southern monophthongization of /ai/ In adults, Southern monophthongization of /ai/ has been shown to be conditioned by social and stylistic factors such as such as sex, age, socio-economic status (SES) (Head, 2003), social atitudes (Bailey and Tilery, 1996), ethnicity of speaker (Hay, Jannedy & Mendoza-Denton, 1999), and peer and caregiver influence (Starks & Bayard, 2002). Monophthongization of /ai/ has been shown to be influenced by internal factors as wel. Bailey and Bernstein (1989), Hazen (2000), Bowie (2000), Head (2003), and Fridland (2003) have reported several internal variables, including following phonologival environment, part of speech, and word frequency as correlating with the production of monophthongal [a:]. Of those factors, following environment has received 21 the most atention with a few researchers reporting the efects of part of speech (Bowie, 2001), or word frequency (Hay, Jannedy & Mendoza-Denton, 1999). Research indicates that monophthongization is highly influenced by the sonority 7 of the following sounds. The greater the sonority of the following segment, the higher the probability that the second element of the diphthong wil be weakened. Fridland?s (2003) study of 30 adult European-American (EA) and African-American Vernacular English (AVE) speakers in Memphis shows that glide weakening, or a decrease in the diphthong /ai/, is increasing in the South and most likely to be found before voiced obstruents, nasals, and in word final, or open, position. Hazen (2000) also reports that syllable position of the following sound afects monophthongization of /ai/ as much as sonority. Monophthongs in his data set were more likely to be produced in multisyllable words if the vowel and the following obstruent shared a common syllable, such as lightning (la:t ?ni!). When the following sound appeared in the onset of the following syllable, for example lighting (laj " ti!), monophthong production was les likely. He hypothesized that when the following sonorant and the coda share the same syllable, the sonorant had greater impact on the vowel, thus increasing sonority. Hazen?s findings of greater sonority in closed syllables are in contrast with Fridland?s findings of greater sonority in word final position. Hazen (2000) also reports a hierarchy of following environmental factors which influence monophthongization of /ai/ in West Virginia and North Carolina: liquids > 7 Sonorance refers to the fre pasage of air through either the oral or nasal cavity. 22 nasals > voiced obstruents > voiceles obstruents. 8 His results are similar to findings by Bailey and Bernstein (1989), Labov and Ash (1997), Bowie (2001), and Anderson (2002). Head (2003) expands Hazen?s hierarchy to include glides, vowels and word boundary, and reports a hierarchy in Elba, Alabama, of glides > liquids > word boundary > voiced obstruents > nasals > voiceles obstruents > vowels. In adult speech, both Hazen and Head found that following liquids were more likely to produce monophthongization than other following environments such as nasals, or voiced and unvoiced obstruents. Al of the researchers above found greater monophthongization of /ai/ preceding voiced obstruents than preceding voiceles obstruents. Bowie (2001) indicates that word clas has an efect on monophthongization. Nouns, especialy non-subject nouns, adverbs, and verbs are more likely to be monophthongal than adjectives and other syntactic categories, including prepositions and determiners. Hay, Jannedy and Mendoza-Denton (1999) found that lexical frequency, defined by them as five or more occurrences in the corpus of their study, was an indicator of monophthongization. They suggested that frequent production led to ?semantic bleaching, phonological reduction? (n.p.). Some studies indicate possible changes in progres in traditional Southern English, some toward expanded monophthongization and some toward reduced monophthongization. Comparing data from 1996-2001 to data in the Linguistic Atlas of 8 English liquids include [l][ and [r]; nasals include [m] [n] [!]; voiced obstruents include [b] [d] [g], voiceles obstruents include [p] [t] [k]. 23 the Gulf States which shows no pre-voiceles glide weakening from speakers in Memphis, TN in the early 70?s, Fridland (2003) suggests that monophthongization in pre- voiceles contexts is increasing. Thomas?s (1997) evidence from the Phonological Survey of Texas indicates a split in the monophthongization paterns among Texas Anglos with urban areas showing les /ai/ monophthongization than rural areas. Bowie (2001) found a similar decrease in monophthongization in his study of 25 white, middle- clas speakers in Southern Maryland. 2.5.4. The variable (ing) A phonological variant with wide-spread English distribution is the alternation of alveolar [n] and velar [!], particularly in the present participle morpheme (ing) (Wald and Shopen, 1985). Labov (1972) and Shuy, Wolfram and Riley (1968) show that the [n] variant is often sen as les prestigious than [!], with many people not admiting to using the [n] variant at al. In non-Southern speech, greater use of the [n] variant has been documented in lower social clases, in younger speakers, and in informal situations. It is interesting to note, however, that Southern speech alows for the use of alveolar variant in formal situations as wel. The variant has been noticed in State of the Union speeches by Presidents Clinton and Bush. 9 Shuy, Wolfram and Riley (1968) show that production of [!] may also be influenced by phonological conditions such as preceding and following sounds. Preceding alveolar stops (t,d) favor the [!] variant, while preceding velar stops (k,g) favor [n]. Following velar stops favor the [!] variant, while following alveolar stops 9 R. Sabino, personal conversation, August 24, 2006 24 favor [n]. Wald and Shopen?s (1985) article on the factors surrounding (ing) indicates that, in addition to sociolinguistic variables of gender, clas and formality, gramatical category can afect [n] production. In most dialects of English, verbs are the most likely gramatical category to contain the [n] variant, followed by the pronouns something, nothing, and anything. Frequency of production also influences the production of the alveolar [n] form. Although litle research into monopthongization has been conducted on child speakers, some researchers have studied (ing) in children. Fisher?s early (1958) study of (ing) in New England included an equal number of boys and girls (N=24), age 3-10, divided equaly into two groups by age. His findings are similar to Labov?s, (1972a) that [n] was more likely to be produced by boys, by children from lower socio-economic groups, and in les formal situations. Acording to Roberts? (1994) study, 3 and 4 year olds acquire gramatical and phonological constraints before social constraints. Gramaticaly, the Philadelphia children whom she studied were more likely to use [n] in verbs/complement structures 10 and les likely to use [n] in subject forms. 2.5.5. Yall and fixin to Two lexico-syntatic features readily asociated with Southern American English that have received the atention of researchers are yall and fixin to. In fact, acording to Bernstein (2003), ?No feature has been more closely identified with southern speech than the use of yal? (p. 107). Studies by Tilery, Wikle and Bailey (2000) and Maynor (1996) 10 Roberts found no significant diference in (ing) variation betwen verbs and complements, and diferentiates betwen nouns, adjectives, and subject complements. 25 indicate that yall is beginning to lose its Southernes based on evidence from the Southern Focus Polls and Linguistic Atlas of the Gulf States showing that the feature is difusing to speakers in northern and western states in spite of the stigma usualy atached to Southern speech. Tilery, Wikle and Bailey (2000) report that over 40% of non- Southerners surveyed, even those with no apparent Southern roots, admit to using yall. Al the above researchers report that yall is widely prefered over you all among younger users. Maynor (1996) reports that age is of more importance than race, gender, or SES in determining which variant is used. Although not as widely recognized as Southern outside the region, anecdotal evidence suggests the inchoative aspect marker fixin to is a feature often acquired in common with a positive orientation to the South. 1 It is now a common feature of Southern speech (Bailey and Tilery, 1996). Ching?s (1987) study of 104 students and faculty at the University of Memphis indicates that agrement varies on a firm definition of fixin to, but the core meaning contains elements of priority of action, a slight delay, and anticipated preparatory action. 2.6. Sumary The research discussed above addreses the various questions that are of concern in this study which atempts to shed light on the nature/nurture debate in first language acquisition. Twin studies showing the efects of divided atention, both qualitatively and quantitatively, argue for the efects of nurture, while those twin studies that show the 1 R. Sabino, personal communication, August 24, 2006. 26 similarity of articulatory development and rate of acquisition of MZ twins give evidence consistent with a greater efect of nature. Inherited temperament characteristics such as shynes, persistence, wilingnes to approach new situations, and activity level have al been shown to correlate with language acquisition. Shy children do not interact as frequently as more curious counterparts, and children with high activity levels appear to expend energy on physical rather than verbal activities. Children who are curious and persistent appear to acquire language more quickly. Because most of the above nature/nurture studies have been conducted on middle- clas children, they represent the acquisition of Standardized English. Although studies have been conducted to ascertain the conditioning of dialect forms, studies of monophthongization in young children focus on erors, rather than the acquisition of dialect forms. Studies of adult monophthongization indicate several external and internal conditioning factors: topic, age, gender, ethnicity, peer influence, folowing sound, word clas, stres, and word frequency. Other studies indicate that acquisition of a standardized or regional dialect reflects some level of choice. Some studies (Roberts, 1997) indicate that children construct their own rules for certain variables, others indicate that parental modeling is necesary (Payne, 1980 ). To date, researchers have provided results that indicate the efects of both nurture and nature. However, they do not efectively speak to the relative efects of these factors due to the geneticaly and/or environmentaly disimilar subject pools from which they draw. A subject pool consisting of both geneticaly like subjects and geneticaly unlike subjects raised within the same environment is necesary to control for the efects of both 27 nurture and nature. Siblings in larger multiple-birth groupings are born at the same time and share their environment, efectively eliminating environmental diferences. When two or more multiple-birth siblings are MZ, genetic diferences are also eliminated. When the multiple-birth group contains both MZ and DZ siblings, the circumstances are ideal for testing the efects of nature vs nurture. The study that follows analyzes retrospective data from such a subject pool, 2 MZ and 2 DZ same-sex, multiple- birth siblings. 28 CHAPTER 3 METHODOLOGY 3.0. Introduction The purpose of this retrospective longitudinal study is to determine the relative contributions of nature and nurture to the language development of four multiple-birth, same-sex siblings. The study looks at two monozygotic (MZ) siblings and two dyzygotic (DZ) siblings raised in the same home to determine the degre to which the children are developing language similarly. Language data was available from the four siblings from age 2 to age 6. The data were analyzed by several methods. Mean Length of Uterance (MLU) and Developmental Sentence Scores (DS) were used to determine the rate of gramatical acquisition. Quantitative analysis was used to determine relevant linguistic conditioning for two regional phonological features, the monophthongization of /ai/ and the variable (ing). The lexico-syntactic features yall and fixin to were also examined. A Carey Temperament Scale (1996-2000) analysis, which measures nine personality characteristics, was obtained as an additional means of asesing the impact of nature and nurture on language development. This chapter discusses the selection of the subjects and the procedures and the instruments used to collect and analyze the data. In al cases, 29 the null hypothesis, that there is no diference in the language development among the siblings on any of the measures, is tested. 3.1. Selection of Subjects The subject population for this study consists of four members of a group of multiple-birth, same-sex siblings born to European American, college-educated parents living in the South. Two of the siblings are MZ (with identical DNA), and two are DZ (with at least 50% identical DNA). The DZ twins are identified as D1 and D2; the MZ twins are identified as M1 and M2. Al children are developing normaly with no reported speech, hearing, intelectual, or medical deficiencies. From birth, al of the siblings were tended by multiple caregivers who asisted with feding, bathing, changing, and nurturing. In the first few months, tasks were shared equaly by caregivers and parents. No caregiver was asigned an individual child; each asisted with the child who needed atention at a given time. Therefore, it is not likely that particular caregivers unduly influenced any of the children in the sibling group. An asumption of the study is that the sibling environment is more consistent than environments reported in previous twin studies. The parents are acquaintances of the researcher and her family, and the researcher has known the siblings from their birth. Because the sibling group contains both MZ and DZ subjects, the children constitute an excelent subject pool to study the nature/nurture aspects of language development. 30 3.2. Data Collection Data used for this study were taken from pre-existing audio tapes which were created by the parents and one of the caregivers during the natural course of the children?s development betwen the years of two and six. The tapes are composed of child-directed conversations in the course of the siblings? play activities. Because the audio tapes were not created with this study in mind, the conversation is not controlled in terms of subject mater, type of interaction, location, privacy of conversations, or amount of data. During the taping, the siblings were not aware that they were being recorded: the recording device was not visible, and the siblings were not wearing microphones. Enthusiastic parental consent was obtained to use the data on the audio tapes in this research study. Al data were collected from tapes which fel within two weks of the siblings? second birthday and within two weks of subsequent six-month intervals with the exception of the period betwen age 4.5 and 6 when no tapes were available for study. Thus, al utterances examined for each six-month interval fal within a thirty-day period. 3.3. Transcription The researcher transcribed al available audiotapes which fel within the designated time frame using an Olympus Pearlcorder micro-mini transcriber. Transcription was enhanced by parental input in instances where utterances or speaker identification was not clear to the researcher. Uninteligible utterances were discarded, as were utterances which could not be positively ascribed to a specific sibling. Al useable 31 utterances were then listed by speaker and age on separate worksheets using standard English orthography. Variants of /ai/ were coded as [!], [a] or [ai]. Variants for the variable (ing) were coded as [n] or [!].!"#$!%$&?()*+,-./(.?(!.)0$-+!1$2$!%?+.$34 3.4. Analysis of grammatical development 3.4.1. Mean Length of Uterance (MLU) As discussed in the Section 2.3, despite some criticism (Crystal, 1974; Kle & Fitzgerald, 1985) MLU has become a standard measurement of child language development (Blake, 1993; Genese, Nicoladis & Paradis, 1994; Roberts, 1994; Shatz, 1994). Using this widely acepted measure of early language development, this study hypothesizes that there wil be no diference in MLU scores of the siblings at the 6 ages for which data has been transcribed. Due to the nature of the pre-recorded data, the number of utterances available for analysis difers across the siblings. The 100 utterances for each child recommended by Brown (1973) were not obtainable at any of the ages studied. However, Miler and Chapman (1981) validated Brown?s correlations in their study of 123 children using only 50 utterances to determine MLU; therefore, the current study is comfortable with a level of 50 utterances per sibling. With the exception of the number of utterances per child, Brown?s guidelines were followed in calculating MLU. Only fully transcribed utterances were used and doubtful transcriptions were eliminated. Inflectional and derivational afixes were counted as single morphemes. Lexical items considered as one morpheme were proper names, compound words (doghouse), diminutive words (choo-choo), iregular past tense 32 verbs (ran, went), auxiliaries, and catenatives (gonna, wanna) because evidence indicates that these words operate as single morphemes for young children (Brown, 1973). The following utterance What are we gonna do with the Blues Clues book? (D2 age 4.5) 1 2 3 4 5 6 7 8 9 would consist of nine morphemes with the catenative gonna and the proper name Blues Clues counting as single morphemes. Since the auxiliary is separated from the main verb, it is considered to be operating as a separate morpheme. Data at age 2 did not contain enough tokens for MLU measurement. Thus, MLU for each subject was calculated at ages 3, 3.5, 4, 4.5, and 6 by dividing the total number of morphemes found per child by the total number of utterances available for that child. (Se Appendix A for a fuller explanation of Brown?s guidelines.) SPS 13.0 for Windows was used to perform a two-tailed t-test to determine if significant diferences existed betwen the siblings? MLU scores. A value of p<.05 is interpreted to indicate significant diference for a pair of siblings. 3.4.2. Developmental Sentence Scores (DS) and Developmental Sentence Type (DST) Developmental Sentence Scores (DS) was used in this study as a second measure of the children?s language development. Although DS is not as widely used by first language acquisition researchers as Brown?s MLU, DS is recognized as valid for asesing language development of both normaly developing children and those with language deficits (Haynes, Purcel, & Haynes, 1979; Haynes & Shulman 1998; Pierce & Bartolucci, 1976; Cole, 1980). 33 DS aseses sentences by asigning a weighted value to eight diferent categories of gramatical forms as the child develops their use. Nouns, pronouns, main verbs, secondary verbs, negatives, conjunctions, interogative reversals and wh-questions are ranked by progresively more advanced usage. Structures which develop early are given lower point values than those which develop later. The points asigned to each sentence at each age for each child for the eight gramatical forms were totaled and then divided by the number of sentences asesed to obtain a mean sentence score. The most recent version of Le?s Developmental Sentence Analysis (1974) was used to describe the siblings? sentence structure development. 12 Until a child?s uterances consist of clausal structures, either dependent or independent, regardles of whether or not the utterances are deemed corect by adult standards, DS is not considered a valid measure. Thus, in order to compute a DS score, it must first be established that the child is producing such utterances. Consistent with Le?s instructions, Developmental Sentence Types (DST) analysis was performed in order to determine at what age to begin DS scoring. For each age, each child?s usable utterances were separated into single-word constructions, two-word constructions, phrases of thre or more words, and clausal constructions. The percent of clausal constructions per sibling was calculated. When DST 12 The problems encountered with insufficient data in the MLU analysis also present themselves in relation to computing DST and DS scores, though to a leser degre. Le (1974) recommends a corpus of 100 uterances to compute the initial DST; however, fewer than 100 utterances can be used to estimate a child?s gramatical ability. 34 indicated that more than half of the child?s uterances were composed of clausal constructions, the child was determined to be ready for DS testing. A secondary indication of readines for DS consists of 1) a decline across data points in the proportion of single-word and two-word constructions and 2) a concomitant increase in the proportion of phrases and sentences. The proportion of increase or decrease in these constructions was compared at ages at ages 3, 3.5, 4, and in one case, 4.5 until readines was determined. For the DS, Le (1974) recommends a corpus of 50 unique clausal constructions in order to acurately ases gramatical development, with scores from fewer than 50 sentences considered an estimate of a child?s gramatical ability. DS scoring was computed with the uterances available for each child. To compute the DS, repetitions were removed from the data sets, leaving only the unique utterances for each sibling beyond the age at which previous analysis determined DS could be used. These utterances were listed for each child on the Excel spreadsheets. Uterances were asigned a score acording to the scoring instrument developed by Le (1974) shown as Appendix B. 13 Personal and indefinite pronouns are asigned one point; negatives, modals and interogative reversals are asigned more points, depending on which gramatical structures they combine with. An extra point is asigned for a complete sentence structure. For example, the utterance Can you do this and this? (D2, age 3) is scored as folows: 13 Complete scoring instructions are found in Developmental Sentence Analysis (Le, 1974) 35 Table 3.1. DS computation of Can you do this and this? word structure points this x 2 indefinite pronoun 1x2=2 you personal pronoun 1 can do modal + main verb 4 and conjunction 3 interogative reversal of modal 6 complete sentence structure 1 Total points for uterance 17 Scores for al sentence structures for each child at each age were totaled and checked by the researcher. For each child at each age, the total score was divided by the total number of sentences per child to achieve an individual DS score for each of the age ranges. SPS 13.0 for Windows was used to perform a two-tailed t-test to determine if significant diferences existed betwen the siblings? DS scores. An alpha value of p<.05 was selected as the criterion for significant diference. 3.5. Analysis of regional dialect features Because the siblings reside in the South and some of the caregivers use regional dialect features, it was decided to analyze the siblings? speech for production of features which linguists have noted as being particularly prevalent in the Southern United States. In the South, the phoneme /ai/ have two alophones, [ai] and [a:], each conditioned by phonological and social factors. Transcripts from ages 2 to 6 were searched for lexical items containing /ai/ in adult standardized English, such as mine, diaper, and right to test for alternation betwen the alophones. The pronunciation of the 36 vowel was coded as monophthong or dipthong. Examination of the monophthongs revealed an unanticipated mid-central variant [!]. A second feature asociated with Southern speech is the widespread substitution of [n] for [!] in nouns, adjectives and adverbs as wel as in verbs. Since variants of (ing) are limited to multi-syllable words, the tapes were examined for multi-syllable lexical items containing the variable (ing). Pronunciation was coded as velar [!] or alveolar [n]. The tapes were also examined for instances of fixin to and yall. 3.5.1 Analysis of variable /ai/ The literature (Bernhardt & Stemberger, 1998; Kehoe & Stoel-Gamon, 1999) indicates that monophthongs are acquired before diphthongs. The percent of diphthong use was found for each sibling by dividing the number of /ai/ found at each age by the total number of tokens at the same age. For the purpose of this study, the diphthong was determined to be fuly acquired when a sibling?s diphthong usage was at the unambiguous 80% level. In order to determine a possible diference betwen pre-monothongal [a] and later [a:] constrained by Southern conditioning, tokens of /ai/ acquired before the diphthong was fully acquired were clasified as pre-diphthongal. Tokens of /ai/ acquired after were clasified as post-diphthongal. JMP IN 4.0 was used to compute multivariate analysis on the lexical items containing [a]. Although many quantitative studies (e.g. Bowie, 2001; Head, 2003) exclude nearly categorical types, in the sibling data, I and my are so prevalent (N=202) that limiting the number would reduce the number of tokens per child to numbers too 37 smal for valid analysis. Al the I and my tokens, therefore, were retained for this study. The vowel sounds were clasified as monophthongs or diphthongs impresionisticaly by repeated listening to the audio tapes. Items that were dificult to clasify impresionisticaly were examined acousticaly with Speech Analyzer 2.6 for Windows and a Pearlcorder Micro/mini transcriber connected to a Del Inspiron 1100 laptop computer running Windows XP. A Y-split audio cable was connected though the headphone jack of the transcriber to the audio-in port of the computer alowing the researcher to hear the segments as they were being recorded by Speech Analyzer 2.6. Questionable items were categorized by comparing the F1 and F2 of the vowel segment to a mean F1 and F2 found for the siblings. To perform the formant analysis, the word containing the vowel was digitized. The segment of interest was isolated on the spectrograph by repeated listening. Individual segments within words were isolated by seting boundary markers outside the target segment and gradualy moving in to isolate the vowel sound. To determine mean F1 and F2 formants for the siblings, five tokens of [ai] and five tokens of [a:] were taken from each child at each age and plotted on the spectrograph. The mean for the F1 and F2 formants was calculated for each of the 40 tokens. Once the mean values were established, the questionable tokens were examined for evidence of an ofset drop of the first formant and an offset rise of the second 38 formant, indicating the patern of the diphthongal [ai]. Vowel segments which could not be categorized were discarded. Each identifiable /ai/ token was coded for child, age, following phonological environment, part of speech, syllable stres, and word frequency. Following phonological environment was coded as pause, voiced obstruent (i.e., /b/, /d/, /g/), unvoiced obstruent (i.e., /p/, /t/, /k/), nasal (e.g. /m/, /n/, /!/), liquid (i.e., /r/, /l/) glide (i.e., /w/, /y/), or vowel. Multivariate analysis was used to determine if the siblings difered along this dimension and the degre to which phonological conditioning matched that reported for adults. Word clas was operationalized as noun, pronouns including contractions of I (e.g. I?l, I?ve and I?m), verb, adjective, adverb, and other. Each token was clasified as belonging to one of these clases. Multivariate analysis determined if sibling monophthongization was following adult paterning reported in the literature. Syllable stres was determined by listening for word stres and categorizing each token as occurring in either a stresed or unstresed syllable. Following Hazen (2000) and Bowie (2001), multisyllable words with monophthongal vowels were analyzed for secondary stres paterns by breaking the words into subsyllabic units consisting of onset, nucleus and coda to se whether or not following sound shared a common syllable with the /ai/ vowel or if it occurred in the onset of the folowing syllable. For example, the word dinosaur has an initial stresed syllable, /dai/ with a following nasal occurring in the onset of the following syllable, /n"/. 39 Words containing /ai/ were also analyzed for word frequency. For this study, word frequency was considered the frequency of use of a word within the corpus of the siblings as a whole as found on the transcribed tapes. Lexical items with five or more occurrences within the sibling data base were analyzed to se if there was significantly more monophthongization with those forms than with words with fewer than five occurrences. Because of the high occurrence of I in the sibling lexicon (175) compared to other words, I was analyzed for frequency separately from other parts of speech. I?m, I?l and I?ve were retained with the rest of the data. As the analysis proceded, when two factors within a factor group were linguisticaly similar and factor distributions were non-significant, the factors were collapsed. The Chi-Square web calculator (http:/ww.georgetown.edu/faculty/balc/webtools/web_chi.html) was used to test for significant diferences betwen the factors. 3.5.2. Analysis of variable (ing) Variation of (ing) as an alternation betwen velar [!] and alveolar [n] occurs only in multisyllable words. In monosyllables, [n] and [!] are contrastive, as in sing vs sin, or thing vs thin. The tapes at age 2, 3, 3.5, 4, 4.5 and 6 were examined for multisyllable words containing /i!/ in adult speech. The items were examined by repeated listening by the researcher for phonetic realization of the nasal segment. Tokens which could not be positively identified as either a velar nasal or an alveolar nasal were discarded. Remaining items were placed on a JNP IN 4 worksheet and coded by speaker, age, word clas and following phonological environment. Word clas was coded as noun, adjective, 40 adjective in complement structure, pronoun, and verb. Following phonological environment was coded as glide, liquid, nasal, unvoiced obstruent, voiced obstruent, vowel, and pause. Totals were tabulated for each category. Bivariate analysis was used to determine if the siblings difered along this dimension and whether their phonological conditioning paterned similarly to Southern adult paterning. The research on phonological development indicates that [n] is acquired prior to [!]; therefore variants for each child were analyzed only after the velar nasal had been acquired. For the purpose of this study, the velar nasal was determined to have been acquired when it first appeared in the transcripts of the audiotapes. When Chi-Square analysis found the distribution betwen two factors within a factor group to be non- significant, the factors were collapsed. Chi-Square web calculator (http:/ww.georgetown.edu/faculty/balc/webtools/web_chi.html) was used to test for significant diferences betwen the factors. 3.5.3. Yall and fixin to Al tapes were examined for instances of the Southern features yall and fixin to. Only two examples of yall were found. There were no instances of fixin to. Thus, no analysis was possible. 3.6. Temperament Analysis In contrast to the linguistic data collected from pre-existing audiotapes, the temperament asesment was done at the time of the current study. Parents were asked to complete the Carey Temperament Scale (CTS) which aseses the following nine 41 temperament characteristics, as defined in the Carey Temperament Scales Test Manual (1996-2000): 1. RYTHMICITY?the predictability of routine 2. SENSORY REACTIVITY?the amount of stimulation necesary to evoke a response 3. PERSISTENCE?the length of time activities are pursued 4. DISTRACTIBILITY?the efectivenes of extraneous stimuli in interfering with behavior 5. INTENSITY?the energy level of responses 6. MOD?the amount of pleasant or unpleasant behavior in various situations 7. APROACH?the nature of initial responses to new situations 8. ADAPTABILITY?the ease or dificulty with which reactions can be modified in a desired way 9. ACTIVITY?the amount of physical motion during activities CTS questionnaires created for ages 3-7 were completed for each sibling. Each questionnaire consisted of 110 items concerning behavioral characteristics which were rated on a six point scale. Computer analysis using Quickscore V4.2P, provided by the CTS publishers, was used to create behavioral profiles showing the siblings? temperament characteristics in the nine areas. The scores provided by the Quickscore V4.2P were placed on an Excel worksheet and graphed. Graphs were than compared to determine if the MZs had similar profiles, and whether they difered from those of the DZs. The siblings were ranked in the areas of ACTIVITY level, PERSISTENCE, SENSORY/reactivity and wilingnes to APROACH new 42 situations, since these areas were sen in the literature review as impacting language acquisition. 3.7. Comparison of language development Temperament profiles of the siblings were compared to the MLU, DS, gramatical development scores, and results of the [a:] and [!] analyses. The MLU and DS scores were ranked for each sibling at each age. The siblings were also ranked at each age for degre of monophthongization of /ai/ and use of the [n] variant of (ing). The ranks were then compared to determine if there was a relationship betwen MLU and DS scores and production of Southern regional forms. The temperament characteristics of ACTIVITY, SENSORY, APROACHABILITY and PERSISTENCE were ranked for each sibling in the same fashion, and the temperament rankings were compared with the rankings of MLU, DS, and Southern regional forms to se which temperament characteristics, if any, could be correlated with language development. 43 CHAPTER 4 RESULTS: GRAMATICAL DEVELOPMENT 4.0. Introduction This chapter reports the results of the analysis of Mean Length of Uterance (MLU) and Developmental Sentence Structure (DS). The purpose of the two asesments of gramatical development was to determine if there were similarities in the language development of the MZ siblings and if that development difered from the development of the DZ siblings. The null hypothesis is that no diference wil be found. 4.1. MLU Results MLU was computed from guidelines set by Brown (1973) using the data transcribed from audiotapes for ages 3, 3.5, 4, 4.5 and 6. Data from age 2 contained too few utterances to be analyzed for MLU and DS. For each age, the total number of morphemes in al uterances was divided by the number of utterances to ascertain MLU for each child. Following the findings of Miler and Chapman (1981), 50 morphemes per child were considered reliable for this study. Reliable numbers to compute MLU scores were not available for each sibling at each age. The number of utterances available for each child is sen in the table 4.1 below, with unreliable numbers of utterances indicated in parentheses. 44 Table 4.1. Number of uterances for each child at each age 3 3.5 4 4.5 6 D1 66 57 51 51 (49) D2 64 76 71 (49) 64 M1 62 51 (38) (47) 80 2 68 (44) 63 (26) 51 Table 4.2 shows the results of the calculations for MLU for each subject at each age based on the available number of uterances given in Table 4.1. The 49 utterances produced by D1 at age 6 and D2 at age 4.5 are considered an estimated MLU and appear in parentheses. Cels with unreliable data are left blank. Table 4.2. MLU for each subject at each age 3 3.5 4 4.5 6 D1 2.5 3.7 4.2 3 (4.1) D2 4.5 6.2 5.1 (6.2) 7.4 M1 2.9 3.2 7.1 2 2.3 4.6 5.9 The data is sufficient to evaluate the hypothesis for the MZs only at ages 3 and 6 and for the DZs only at ages 3, 3.5 and 4. Figure 4.1, based on the data in Table 4.2, shows the comparative progresion of the siblings based on MLU from age 3 through age 6 with estimates from D1 and D2 included. 45 Figure 4.1. MLU progresion by age The figure indicates that D2?s MLU is higher than that of the other siblings at every age, difering least at age 4. The MZs and D1 show a patern of similar development until age 4.5, when the MZs show a greater resemblance to D2. At al ages, the MZs patern more similarly to each other than to the either of the DZs. Figures 4.2 and 4.3, which separate the data for the MZs and the DZs, reveal their comparative development more clearly. 46 Figure 4.2. MZ MLU pattern Figure 4.3. DZ MLU pattern To determine if the diferences shown in Tables 4.1-4.3 were statisticaly significant, t-tests were computed on the MLU scores of the siblings. Results of the t-test are given in Table 4.3 below. Diferences were considered to be significant at the p<.05 level. Cels with unreliable data have been left blank. Results based on les reliable MLU scores are indicated in parentheses. Table 4.3. T-test results of MLU scores 3 3.5 4 4.5 6 M1/M2 .010 ns D1/D2 .000 .000 ns (.000) (.000) M1/D1 ns ns (.001) 1/D2 .002 . 000 ns M2/D1 ns ns (.048) 2/D2 .000 ns ns The data are insufficient to evaluate the hypothesis fully; however, they indicate that an initial significant diference betwen M1 and M2 has disappeared by age 6. The diference betwen M1 and D2 persists through age 3.5. D1 and D2 are significantly diferent at ages 3 and 3.5, and significant diferences based on estimated MLU also appear at ages 4.5 and 6. At age 6 there is no diference betwen M1 and D2 or betwen 47 M2 and D2. Estimated MLU also indicates that D1, who developed similarly with the MZs at early ages is significantly diferent by age 6. It appears that, with the exception of D1, the children are more diferent at age 3 and les diferent at age 6. Based on the visual similarity of the MZs for age 3.5, 4.0 and 4.5 in Figure 4.1, and because of the gaps in the data, it was decided to pool the MZs and run a second t- test for significant diferences betwen the MZ group and D2, the DZ sibling who produced the most utterances. The pooled data includes no fewer than 73 tokens at any age for the monozygots, more than the 50 tokens considered reliable for this study. Table 4.4, which compares the pooled MZs with D2, shows a significant diference betwen the MZs and D2 at 3.5 and 4.5. Table 4.4. T-test results on MLU for the pooled MZs and D2 age 3.5 age 4.5 age 6 M/D2 .000 .000 .412 T-test results suggest that by age 3.5, the MZs can be distinguished from their DZ siblings on the basis of MLU, but, as sen on Table 4.3 on page 46, the statistical diferences disappear by age six. 4.2. DS Results Two additional measures of syntactic development, the Developmental Sentence Types (DST) and Developmental Sentence Scoring (DS), are based upon the same 48 corpus of utterances that were used for MLU scoring. 14 DST is an initial analysis of the number of sentence structures within a child?s corpus and is conducted before DS scoring to determine if there are enough sentence structures within a child?s corpus to advance to the more detailed analysis of the gramatical structures in the DS scoring. 4.2.1. DST scores Table 4.5 shows the total utterances, including repetitions and non-clause constructions, as wel as the number of unique utterances used for DST scoring found in the corpus from age 3 through age 6. Table 4.5. Total uterances and usable uterances for DST and DS scoring 3 3.5 4 4.5 6 Total Usable Total Usable Total Usable Total Usable Total Usable D1 85 50 69 52 56 49 52 48 53 44 D2 80 64 106 90 90 82 56 55 71 66 M1 68 52 63 52 42 31 49 44 101 97 2 78 57 50 34 75 54 26 25 74 52 DST scores are the percentage of sentence structures in each child?s corpus of unique utterances. Because the numbers of unique utterances for some of the siblings were low, DST scores were computed for each child with the intention of comparing the results to a gradual increase in the complexity of phrasal constructions (shown in Tables 14 Since DST scores are based upon subject/verb relationships rather than speech episodes, the number of utterances used to compute DST difers from that used to compute MLU. For example, the complex sentence Put the puzzle in there and I can doos it. (D1/3.0) consists of one utterance for the purposes of MLU scoring, but two utterances for DST scoring because of the two clauses with subject and verb in subject/predicate position. Similarly, No want. (M2/3.0) consists of two morphemes for MLU scoring, but is not aceptable for DST or DS because no subject is present. Refer to Appendix B for a full DS scoring chart. 49 4.8-4.10 below) as a secondary measure to further validate readines for DS scoring. Table 4.6 shows the number of sentence structures (S) that were extracted from each child?s usable utterances (U) and the resultant DST percentage score. Table 4.6. Sentence structures (S), unique uterances (U), and DST scores Scores which indicate a readines for DS testing are shown in bold. Chi-Square analysis was then computed on the DST scores to determine if the siblings were producing sentence structures at a similar rate. Results are shown in Table 4.7 below. Table 4.7. Chi-Square analysis of paired DST scores 3 3.5 4 4.5 6 M1/M2 ns ns .001 ns .025 D1/D2 .001 .05 ns .01 .025 M1/D1 ns ns ns ns ns 1/D2 .01 .01 ns ns .025 M2/D1 ns ns .01 ns .025 2/D2 .001 ns ns .01 ns A comparison of the sibling group as a whole shows a significant diference of p<.001 at ages 3, 4, 4.5 and 6, and a diference of p<.05 at age 3.5. Individual comparisons indicate that the MZs are similar at ages 3, 3.5, and 4.5, while the DZs difer significantly at these same thre points. M2 is significantly diferent from each of the other siblings at two points: from M1 and D1 at ages 4 and 6, and from D2 at ages 3 and 4.5. Results show no significant diference betwen D1 and M1 at any age, indicating 3 3.5 4 4.5 6 SS/U DST SS/U DST SS/U DST SS/U DST SS/U DST D1 17/50 34 28/52 54 31/49 63 23/48 48 24/44 55 D2 45/64 70 64/90 71 62/82 76 41/55 75 50/66 76 M1 22/52 42 25/52 48 14/31 47 27/44 64 50/87 52 2 17/57 30 18/34 53 47/54 87 10/25 40 41/52 79 50 that they are developing more similarly than the other siblings. D1 and D2 difer at every age with the exception of age 4. Although MLU results indicated diferences disappearing at age 6, DST results are opposite, finding the most diference at age 6, with similarity crossing zygotic lines: M1 = D1, M2 = D2. When a child?s corpus consists of at least 50% sentences, the utterances are ready for DS scoring, which considers the gramatical complexity of the utterance. Data obtained from D2 at age thre shows 45 sentence structures out of a possible 64 unique utterances, a DST score of 70%, wel above the 50% suggested for subsequent DS scoring. The other thre siblings, however, had les than 50% usable utterances at age 3. D1 and M2 show readines for DS scoring at age 3.5. M1 shows readines at age 4.5. Because the low number of utterances available for thre of the siblings alowed for only an estimate of readines for DS, a secondary measure was used to enhance reliability. Le (1974) suggests that a decrease in one and two-word constructions and an increase in the number of phrases (non-sentences of thre or more words) and sentences indicates readines for DS scoring. Tables 4.8-4.10 show the pre-sentence constructions produced by D1, M1, and M2 until the age when lower percentages of one and two word constructions occur concurrently with an increase in phrases and sentences. (D2?s constructions were not analyzed since readines for DS scoring had been evidenced by age 3, as noted above.) 51 Table 4.8. D1?s pre-sentence constructions 3 3.5 4 D1 N % N % N % One-word 17 43 11 21 6 12 Two-word 10 20 6 12 4 8 phrases 6 12 7 13 8 16 sentences 17 34 28 54 31 63 total U 50 52 49 D1?s production of sentences increased from 34% at age 3 to 54% at age 3.5, while during the same period production of single-word utterances decreased from 34% to 21%. This progresion shows an increase in higher gramatical structures and confirms the readines for DS scoring at age 3.5 as indicated by the DST scoring. Table 4.9. M1?s pre-sentence constructions 3 3.5 4 4.5 M1 N % N % N % N % One-word 11 21 15 29 10 32 9 21 Two-word 11 21 5 10 5 16 4 9 phrases 8 15 7 13 2 6 2 5 sentences 22 42 25 48 14 45 27 64 total U 52 52 31 42 M1?s production is dificult to ases. The percent of two-word utterances showed an expected decrease from age 3 to 3.5, but one-word utterances showed a continuing increase through age 4. The percent of sentence constructions (phrases containing a subject and a verb in subject/predicate relationship) increased at age 3.5 but dropped at age 4. DST was computed for M1 at age 4.5, and, although M1 stil showed a low number of utterances with a tendency to speak in one and two word constructions, at this 52 age, total U increased to 42 with 64% sentence construction, indicating M1?s readines for DS scoring at age 4.5, validating the DST score of 64 in Table 4.6. Table 4.10. M2?s pre-sentence constructions 3 3.5 4 M2 N % N % N % One-word 20 35 9 26 6 10 Two-word 11 19 1 3 6 10 phrases 8 14 6 19 0 0 sentences 5 9 18 53 47 80 total U 57 34 59 At age 3.5 M2 shows a decrease in one and two-word utterances and a slight increase in phrase constructions of thre words or more. This paterns with M2?s percent of usable utterances (53% at age 3.5, and 47 out of 54 usable utterances (87%) at age 4.0 and confirms the readines estimate of 53 at age 3.5 sen in Table 4.6. This paterning of phrasal development shows a similar development for M2 and D1 for whom Chi-Square diference was not significant at ages 3 and 3.5. The phrasal development patern is quite diferent for D1 and M1. D1 shows DS readines at age 3.5 while M1 does not indicate readines until age 4.5. However, Chi-Square results for D1 and M1 show no significant diference at any age, indicating that their phrasal development is esentialy the same. 4.2.2. DS Scores Having met Le?s secondary criteria of an increase in the number of sentences and phrases and a decrease in the number of one and two word constructions, DS scores were compiled for D1, D2 and M2 beginning at age 3.5 and for M1 at age 4.5. 53 DS was computed for the siblings based on Le?s rubric for asigning higher scores for more complex gramatical structures (Se Appendix B). Table 4.11 shows reliable DS scores for each of the siblings at each age. However, as Table 4.6 above indicates, the number of sentences available for analysis did not met Le?s primary criteria at al ages. Cels with les reliable data have been left blank. Table 4.11. Reliable DS scores for each child at each age 3.5 4 4.5 6 D1 D2 7.58 7.8 10.06 M1 7.02 2 Table 4.11 does not alow for comparisons except at age 6, which shows that D2?s scores are higher than at least one of the monozygots. T-test analysis shows a significant diference of p<.003 betwen M1 and D2, indicating a diference which fals along zygotic lines at age 6. 4.3. Sumary The purpose of the MLU and DST/DS measurements was to determine if the geneticaly identical MZs were exhibiting similar gramatical development to one another and to the DZs who are les geneticaly similar. Al thre measures provide evidence that gramatical development betwen the MZs, as wel as betwen the MZs and DZs is disimilar at some points. The null hypothesis of no diference betwen the siblings does not hold, as T-test analysis for MLU indicates that at age 3, the monozygots and D2 are developing diferently from each other. 54 Results from the MLU and DS asesments disagre as to the ages at which similarities are sen among the siblings. Although Chi-Square analysis on the DST scores indicates that MI and M2 are developing similarly at age 3, the more widely used MLU measurement shows disimilarities at age thre. For this study, MLU is the more valid analysis as the DST scores indicate only a ratio of the number of sentences produced, while MLU indicates increased gramatical development. The visual similarity of MLU scores shown on Figure 4.1 also suggests that the MZs are developing more similarly to each other than to either of the DZs. There are more diferences among al the siblings? MLU scores at earlier ages; however, by age 6 diferences sem to disappear with the exception of D1 and D2 who remain significantly diferent. Both MLU and DS indicate disimilarities in the siblings? development; therefore the null hypothesis of no diference is not supported. 55 CHAPTER 5 RESULTS: REGIONAL DIALECT FORMS 5.0. Introduction A wel-described feature of adult Southern English, the monophthongization of /ai/, has received litle atention in the literature on first language acquisition. In fact, the acquisition of diphthongs in general has received litle atention. The present study provides an opportunity to addres both of these isues as it explores the relativity of the roles of nature and nurture in first language acquisition. This chapter approaches the siblings? acquisition of Southern English /ai/ both qualitatively and quantitatively, first discussing the thre variants produced by the siblings. The quantitative analysis explores the frequency of the variants at ages two years through six years. The linguistic conditioning of monophthongization is also explored. This chapter also discusses the two variants of (ing), the alveolar nasal, [n], and the velar nasal, [!], from ages two through six, as wel as factors that condition these variants. Although variation of (ing) occurs in Northern American English, British, and New Zealand English (Wald and Shopen, 1985), Southerners in the U.S. use the alveolar variant in a wider range of styles and in more word clases than Northerners. 56 5.1. Description of /ai/ variants Four hundred and fifty tokens of /ai/ were located in the sibling data set from ages two through six. 15 Data from the earliest audio tapes shows production of thre variants: a mid-central monophthong [!], a low central monophthong [a], and a diphthong [ai]. The [!] tokens (N=14) al occur in 1 st person constructions, such as [!] got, and [!] think. Eighty-six percent of the [!] tokens from age 2 through 4 were produced as [!] know ?I don?t know? in contrast with [ai] know ?I know? As Table 5.1 below shows, not al the siblings produce tokens of [!]. Moreover, [!] does not appear in the data after age 4, suggesting this is the earliest of the thre forms. Table 5.1. Monophthong production at age 2 Table 5.1 shows the distribution of variants of /ai/ at age 2. Table 5.2 shows the gradual decline of [!] until it disappears after age four. 15 Only those containing /ai/ in adult Standard English were considered in the analysis. [!] N % [a] N % [ai] N % D1 I 2 33 I 2 33 right 2 33 D2 0 I 15 37.5 right (2) night(1) rice (3) bite (5) light (1) mine (3) I (11) 26 62.5 M1 I 2 25 I 2 25 tiger (2) I (2) 4 50 2 I 2 28.5 I 2 28.5 I (1) eyes (1) 2 43 57 Table 5.2. [!] production at ages 3 and 4 Sibling age 3 age 3.5 age 4 D1 25% (N=2) 0 0 M1 11%(N=2) 0 25% (N=2) 2 31% (N=4) 0 5% (N=1) The full set also provides the strongest evidence for an order of acquisition: early monophthongal [!] is followed by the monophthong [a]. Next, the diphthong [ai] is acquired, and last to appear is [a:] constrained by the phonological and social conditioning of adult Southern speech. As Table 5.1 indicates, several variants may appear at the same age. The smal number of [!] tokens makes it dificult to draw conclusions about the phonologies of the siblings at age 2-4; however, the diference betwen D2 and the other siblings is not consistent with the hypothesis that no diferences wil be found for the language development of the four siblings. By age 6, al siblings produce the variant [a:] to some extent, with one of the siblings developing a preference for the monophthong. The data for this child also reveal the emergence of some phonological conditioning. An atempt to determine at what point [a:] was acquired by diferentiating the early monophthongal forms by duration proved unsuccesful as length semed conditioned by situation, mood, and topic. Since the published research indicates that the earliest vowels to appear are the mid-vowels, [!] and [a] with diphthongs appearing later (Bernhardt & Stemberger, 1998; Kehoe & Stoel-Gamon, 1999), it was decided to separate the data into two categories, pre-diphthongal and post-diphthongal, and to 58 analyze each category acording to the internal constraints of following phonological environment, word clas, stres, and word frequency. A level of 80% diphthongal use was defined as indicating acquisition of the diphthong. Table 5.3 shows the percentage of diphthong use for each sibling. The ages at which the diphthong appears to be fully acquired is marked in bold. Table 5.3. Percent of diphthong use for each sibling at each age age 2 3 3.5 4 4.5 6 D1 3% (N2) 37% (N3) 46% (N6) 71% (N10) 90% (N9) 73% (N12) D2 38% (N25) 91%(N10) 61% (N20) 63% (N2) 36% (N8) 12% (N3) M1 43% (N4) 68% (N13) 82% (N13) 75% (N6) 10% (N19) 96% (N52) 2 43% (N3) 21% (N3) 10% (N6) 7% (N17) 10% (N3) 65% (N25) In view of the nul hypothesis of no diference, it is interesting to note that both MZ?s acquire the diphthong at age 3.5, while D2 acquires it at 3 years and D1 acquires it at 4.5. This is similar to findings of Matheny and Bruggeman (1972) and Locke and Mather (1989) who reported more similarity of articulation betwen MZ twins than betwen DZ twins. 5.2. Analysis of /ai/ variation 5.2.1. Pre-diphthongal analysis Tokens produced by each sibling before diphthongs had been fully established at the 80% criterion were analyzed for conditioning factors. D1?s tokens prior to age 4.5, D2?s prior to age 3 and the MZ?s prior to age 3.5 were clasified as pre-diphthongal and placed on a JMP IN 4.0 spreadsheet. These tokens were coded for following phonological environment, word clas, stres, and word frequency. One hundred and 59 twenty-eight tokens were clasified as pre-diphthongal. Table 5.4 shows the initial distribution by following environment for pre-diphthongal tokens. Table 5.4. Pre-diphthongal analysis of /ai/ variation by following phonological environment Count Row % [ai] [a] glide 3 42.86 4 57.14 7 liquid 1 50.0 1 50.0 2 nasal 12 41.38 17 58.62 29 -v obst 3 6.0 17 34.0 50 +v obst 16 45.71 19 54.29 35 vowel 1 50.0 1 50.0 2 pause 3 10.0 0 0.0 3 total 69 59 128 While following vowels and liquids do not impact the alternation, the direction of the efect of voiced and unvoiced obstruents is consistent with the adult Southern patern. Unvoiced obstruents are negatively correlated with the [a:] variant; voiced obstruents are positively correlated with it. Although the numbers are smal, it appears that the siblings are treating glides and nasals similarly to voiced obstruents. Chi-Square was used to determine if these distinctions were significant. Where no significant diference existed betwen categories, the categories were collapsed. Pause, having only thre tokens and categoricaly producing a preceding diphthong, was omited from analysis. Because Chi-Square analysis indicated no significant diference betwen 60 glides, liquids, and vowels, these categories were collapsed. Finding no subsequent significant diference betwen the collapsed category (glides liquids, vowels) and nasals or voiced obstruents, al categories except unvoiced obstruents were collapsed, leaving two categories: unvoiced obstruents and other. A significant diference of p< .01 was found betwen unvoiced obstruents and al other categories, indicating that the primary indicator of diphthongal [ai] in sibling pre-diphthongal speech is a following unvoiced obstruent. Bowie (2001) indicates that nouns, adjectives, and adverbs are more likely to be monophthongized (/ai/ realized as [a:]) by adults than are other word clases. Head (2003) found that pronouns, verbs and adjectives were more likely to be monophthongized. To se if the siblings behave as the adults in either Bowie?s or Head?s sample, the tokens were coded for word clas , operationalized as adverb, adjective, verb, noun, pronoun, and other. Table 5.5 shows the initial distribution of pre-diphthongal monophthongization acording to word clas. Table 5.5. Pre-diphthongal analysis of /ai/ variation by word class Count Row % [ai] [a] adj 1 25.0 3 75.0 4 adv 8 72.73 3 27.27 1 noun 20 10.0 0 0.0 20 other 5 71.43 2 28.57 7 pronoun 32 39.02 50 60.98 82 verb 3 75.0 1 25.0 4 Total 69 59 128 61 Pronoun, the largest category (N=82), contained diphthongs 39% of the time. Nouns, which were categoricaly diphthongal, were excluded from analysis. Because Chi-Square analysis showed no significant diference among the categories of verb, adverb, and other, these were collapsed, as were the categories of adjective and pronoun. Analysis of the resulting thre categories shows a sibling pre-diphthongal hierarchy for [a] as pronoun/adjective>adverb/verb/other>noun. Although Hazen (2000) and Bowie (2001) indicate that a monophthong is more likely to occur in syllables which contain both the vowel and a following sonorant, none of the pre-diphthongal data contained multi-syllable words with this patern. Results on single syllable words show 84 which are les stresed within the context of a sentence. These produced a monophthong 43% of the time (N=36). The 44 fully stresed words produced a monophthong 75% of the time (N=33), but Chi-Square analysis did not find the diference to be significant. Hay, Jannedy and Mendoza-Denton (1999) found word frequency to be a significant predictor of monophthongization, defining frequency as occurring five or more times within the corpus. Word frequency analysis of sibling pre-diphthongal data shows that in, general, words occurring five or more times in the corpus, (e.g. I, my, right, I?m, like, and right) were monophthongized, but at 51% the diference was not significant. Les frequently appearing words were al diphthongized with the exceptions of white (N=3) and I?m (N=3) which were both categoricaly monophthongized. It would appear that in the pre-diphthong database, neither stres nor word frequency plays a significant role. 62 The pre-diphthongal sibling? corpus exhibits some evidence of the adult conditioning in monophthongization. A tendency to suppres monophthongization before voiceles obstruents may be emerging, but is not significant at early ages. Adult paterning of /ai/ as [a:] regarding word clas indicates a hierarchy of nouns> adverbs/verbs> adjectives (Bowie, 2001) or pronoun>verb>adj>noun >adv>prep> conjunction acording to Head (2003). The sibling paterning shows the pronoun I is frequently monophthongal, as in adult speech. However, the remaining word clases show either Hazen?s or Head?s patern: adjective >adverb/verb/other >noun. 5.2.2. Post-diphthongal analysis The post-diphthongal analysis contained 322 tokens produced after each sibling had demonstrated acquisition of the diphthong at the 80% criterion. These include tokens from D2 beginning at age 3, the MZs beginning at age 3.5, and D1 beginning at age 4.5 Table 5.6 shows the results of the analysis for following phonological environment. 63 Table 5.6. Post-diphthongal analysis of /ai/ variation by following phonological environment Count Row % [ai] [a:] glide 14 70.0 6 30.0 20 liquid 7 43.75 9 56.25 16 nasal 51 5.43 41 4.57 92 - obst 94 78.3 26 21.67 120 + obst 41 78.85 1 21.15 52 vowel 8 10.0 0 0.0 8 pause 1 78.57 3 21.43 14 Total 26 96 32 The analysis shows that the siblings continue to avoid the monophthong when the vowel is followed by a voiceles obstruent. However, voiced obstruents and pause also strongly favor the diphthong. Following vowels, which categoricaly produced diphthongs, were removed from analysis. Chi-Square analysis showed no significant diference betwen the categories of voiced and unvoiced obstruents, pause, and glide. These categories were collapsed. Although the pre-diphthongal analysis shows liquids as having no impact on monophthongization, the post-diphthongal analysis identifies liquids as the only environment favoring the monophthong. Although liquids and nasals appear to have a diferent efect, liquids favoring [a:] and nasals favoring [ai], the distribution of tokens is not significantly diferent, thus the categories were collapsed. The resultant categories 64 showed a post-diphthongal hierarchy (/ai/ realized as [a:]) of nasal/liquid>everything else. These categories are significantly diferent at p<.0001. Table 5.7 Shows the diferent hierarchies asociated with folowing sounds. Sounds on the left of the implicational hierarchy favor monophthongization. Table 5.7. Comparison of hierarchies of following phonological environment Source Hierarchy (/ai/as [a:]) Hazen (2000) liquids>nasals>voiced obstruents>voiceles obstruents Head (2003) liquids>voiced obstruents>nasals>voiceles obstruents Fridland (2003) voiced obstruents>voiceles obstruents Bailey and Bernstein (1989) voiced obstruents>voiceles obstruents Bowie (2001) voiced obstruents>voiceles obstruents siblings pre-diphthongal other>voiceles obstruents siblings post-diphthongal nasal/liquid>other Results for the analysis of word clas were similar to that found for the pre- diphthongal data. Table 5.8 shows post-diphthongal analysis of /ai/ realized as [a:] by word clas. 65 Table 5.8. post-diphthongal analysis of /ai/ variation by word class Count Row % [ai] [a:] adj 1 73.3 4 26.67 15 advb 15 68.18 7 31.82 2 noun 35 89.74 4 10.26 39 other 12 63.16 7 36.84 19 pronoun 135 6.18 69 3.82 204 verb 18 78.26 5 21.74 23 Total 26 96 32 Pronouns (N=204) again comprised the greatest number of tokens (63%). Chi-Square analysis showing no significant diferences, the categories of pronoun, other, verb, adjective, and adverb were collapsed, leaving a post-diphthongal hierarchy ([ai] realized as [a:]) of other>noun significant at p<.01. While al word clases in post-diphthongal analysis are more likely to contain diphthongs than monophthongs, nouns are least likely to be monophthongized. As with the pre-diphthongal analysis, this patern is in contrast to adult paterning where nouns, adverbs and verbs are more likely to be monophthongized than other word clases. Multivariate analysis showed no significant diference in the efects of word stres or word frequency for the post-diphthongal data. Twenty-two of the multi-syllable tokens in the post diphthong data contained /ai/ in open syllables. Four of these were monophthongized (18%). Twelve tokens contained /ai/ in closed syllables. Only one of these (8%) was monophthongized. Although not consistent with Hazen?s (2000) report 66 that adults in West Virginia and North Carolina are more likely to monophthongize in closed syllables in multi-syllable words, this may be due to the smal number of tokens available for analysis. Figure 5.1 shows the post-diphthongal sibling patern of acquisition of monophthongal [a:] from age 2 to age 6. Figure 5.1. Acquisition of Southern /ai/ monophthongization The figure shows that the sibling who acquired the diphthong earliest also produced the greatest amount of [a:], with 89% acquisition at age 6. Slightly les than half (41%) of the total tokens were produced by this sibling. To test whether adult conditioning was acquired by this child, these tokens (N=133) were analyzed for the full set of factors. D2?s [a:] production was like that of the sibling set with the exception of 67 word clas. For this factor group, there was a significant diference for pronouns (p<.001) and verbs (p<.05). Both categories were more likely to be monophthongized by D2 than by the siblings as a whole. Additionaly, al of the monophthongized adjectives (N=4) in the sibling data were produced by this sibling. 5.2.3. Comparison of post-diphthongal /ai/ Comparing the siblings who produced the most tokens reveals that of M1?s 95 post-diphthongal tokens, only 5 are monophthongs, while D2 shows a growing patern of monophthongal acquisition, becoming 89% monophthongized at age 6. Although neither M1 nor D2 shows significant diferences in stres or word frequency, it is apparent that the nul hypothesis of no diference among the siblings in the rate of acquisition of Southern monophthongized [a:] is not supported. 5.3. The variable (ing) O?Grady, Aronoff and Res-Miler (2005) show that the alveolar nasal appears around age 2 and the velar nasal around age 3 or 4. Haynes and Shulman?s (1998) discussion of morphological and phonological development in children indicates that the present progresive form {-ing} begins to emerge betwen 2 and 3 years, although early pronunciation may not include the velar nasal until around age 3. Both these sources indicate that the alveolar nasal is developmentaly prior to the velar nasal. The sibling data are generaly consistent with those sources. For the siblings, [!] appears first in the data of D2 at the age of 2.5, and for the other siblings at age 3. The variable (ing) occurs only in multi-syllable contexts. Since [n] appears developmentaly before [!], only multi-syllable words were analyzed, and only after [!] 68 appeared in the sibling?s data. Sixty-seven words containing (ing) in multi-syllablic contexts were located over the six age groups. These are shown in Table 5.9. Table 5.9. Velar and alveolar variants of (ing) for ages 2-6 [n] [!] % [n] D1 5 6 55% D2 15 17 53% M1 4 10 72% M2 4 6 64% The table shows that al siblings prefer the [n] variant. Chi-Square analysis shows no significant diference among the four siblings or betwen the DZs or betwen the MZs. It is interesting to note that the sibling producing the greatest percent of Southern monophthongization is not the sibling producing the greatest percent of [n]. In fact, D2 shows the lowest percent of the alveolar variant of the four siblings. Since there were no significant diferences among the siblings, their data were pooled. Using guidelines from Wald and Shopen (1983) and Roberts (1994), who show wide distribution of variable (ing) across standard varieties of English, the (ing) tokens were coded for word clas. The analysis considered adjective, the pronouns something, nothing, and anything, and verbs. Based on Roberts (1994), the data were also coded for verbs in complement structures, such as I saw the leaves turning yelow (D2, age 6). Although an informal survey by students in an undergraduate linguistics clas at Auburn University found about one-third of their [n] tokens were noun forms. such as meting 69 [mi t"n] or Birmingham [b# mIn h?m] (R. Sabino, personal communication, July 13, 2006), no nouns containing (ing) were contained in the sibling data set. Table 5.10 shows the distribution of (ing) tokens by parts of speech. Table 5.10. Distribution of (ing) variation by word class Count Row % [!] [n] total adj 1 10.0 0 1 compl 6 6.67 3 3.3 9 pron 4 50.0 4 50.0 8 verb 17 36.95 29 63.04 46 Nouns 0 0 0 Total 28 36 64 Forty-six of the 67 sibling (ing) tokens were verbs, and 63% of those contained the alveolar variant [n]. Thre tokens could not be categorized by word clas: That?s a winkin, and That?s a layin down winkin. (D1 age 4). Roberts? (1994) data showed an average of 82% [n] in verbs in complement structures for the 17 children in her study, with no child producing les than 67%. However, the data for this study shows only 33% [n] use in verb complement structures, suggesting that the Southern siblings may not be acquiring the Northern verb- complement patern found by Roberts in Philadelphia. More data is needed to determine whether or not this is a regional diference. Chi-Square analysis of the factor groups shows no significant diference across gramatical categories. 70 5.4. Yall and fixin to Yall has ben identified by Bernstein (2003) as the dialect form ost identified with Southern speech. Anecdotal evidence shows Southern children as young as five using the form. 16 During the course of this study, D2 at age 6 was heard by the researcher to produce yal several times; however, only two instances of yall were found in the sibling data set, both produced by D2 at age 6. Sabino (personal communication, July 13, 2006) reports hearing a neighbor child age 3 using fixin to. However, no instances of fixin to appeared in the sibling data, and none of the subjects was heard to use the form during the period of the research. 16 Heard by the researcher in a 5-year-old Sunday School clas 6/21/06 71 CHAPTER 6 RESULTS: TEMPERAMENT ANALYSIS 6.0. Introduction Since temperament is thought to be hereditary (Wilson and Matheny, 1986, Martin et al., 1988, Plomin, 1994, Dale and Goodman, 2005), the children?s temperaments were analyzed to determine if the behavioral profiles of the two MZs were similar or disimilar to each other and similar or disimilar to those of the DZs. As described in Chapter 3, at the time of this study, the parents completed 110 questions regarding aspects of the siblings? behavior. The siblings were 6 years old at the time the questionnaire was completed. This chapter reports the results of the temperament asesment for each of the siblings. 6.1. Scoring Quickscore V4.2P asigned each sibling a raw score in each temperament category from -1 to +2.5. The program then plotted the raw scores on a graph, with 0 considered a mean score for a child betwen 3-6, the age range of the questionnaire. Raw scores for the four siblings are shown in Table 6.1. 72 Table 6.1. Raw scores for temperament characteristics Characteristic D1 D2 M1 M2 ACTIVITY 1.1 0.07 -0.3 -0.34 ADAPTABILITY -0.07 -0.76 -0.42 0.51 APROACH 2.02 -0.47 -0.38 -0.09 MOD 1.19 0.28 -0.58 0.7 INTENSITY 0.48 -0.16 -0.03 0.9 DISTRACTIBILITY 0.8 0.8 0.38 -0.48 PERSISTENCE 1.93 0.3 1.64 1.2 SENSORY/REACTIVITY 1.09 0.03 -0.72 -0.27 RYTHMICITY 2.49 -0.12 1.67 0.69 Characteristics with minus scores are not considered negative traits; they are traits which the creators of CTS consider easier to manage than characteristics receiving a plus score. For example, a plus APROACH score would indicate a more curious child, while a minus approach score would indicate a les curious child. Interpretations of the nine temperament characteristics in terms of the siblings? plus and minus scores are given in Table 6.2. Characteristics are shown in terms of + or ? increments from 0. Table 6.2. Interpretation of individual temperament scores Characteristic D1 D2 M1 M2 ACTIVITY ++active +active -inactive -inactive ADAPTABILITY -quick -quick -quick +gradual APROACH ++cautious -approaching -approaching -approaching MOD ++negative +negative -positive +negative INTENSITY +intense -mild -mild +intense DISTRACTIBILITY +often +often +often -rarely PERSISTENCE ++rarely +rarely ++rarely ++rarely SENSORY/REACTIVITY ++reactive +reactive -nonreactive -nonreactive RYTHMICITY +++iregular -regular ++iregular +iregular Quickscore V4.2P also produces a visual representation of the siblings temperament characteristics which is shown in Figure 6.1. 73 Figure 6.1. Comparison of the sibling temperament scores 6.2. Temperament Profile Analysis Previous studies in the area of language acquisition have shown that four characteristics are likely to influence rate of language acquisition: risk-taking (Wong- Filmore 1979), shynes (Paul & Kelogg, 1997), activity level (Paul & Kelogg, 1997), and atentionality (Dixon & Smith, 2000). ACTIVITY level is measured by the CTS and defined by them as ?the amount of physical motion during slep, eating, play, dresing, bathing.? Although the remaining thre characteristics do not map directly on to the CTS categories, they are equivalent to CTS categories. Risk-taking and shynes are equivalent to the CTS category of APROACH, ?the nature of initial responses to new stimuli? people, situations, places foods, toys, procedures.? Atentionality is similar to the CTS categories of PERSISTENCE, ?the length of time particular activities are pursued by the child with or without obstacles? and SENSORY, ?the amount of stimulation, such as 74 sounds, light, taste, smel or fel, necesary to evoke discernable responses in the child.? 17 Because the studies mentioned above indicated a connection betwen these four temperament characteristics and language acquisition, the CTS categories can be used to ases the null hypothesis of no diference among the siblings. In Table 6.3 and Figure 6.2, the siblings were compared specificaly in the areas of PERSISTENCE, SENSORY, APROACH and ACTIVITY. Table 6.3. Temperament scores for PERSISTENCE, SENSORY, APROACH and ACTIVITY D1 D2 M1 M2 PERSISTENCE 1.93 0.3 1.64 1.2 SENSORY 1.09 0.03 -0.72 -0.27 APROACH 2.02 -0.47 -0.38 -0.09 ACTIVITY 1.1 0.07 -0.3 -0.34 Figure 6.2. Temperament comparison for PERSISTENCE, SENSORY, APROACH, and ACTIVITY 17 Carey Temperament Scales Test Manual and User?s Guide, 1996-2000. 75 CTS asesment indicates that the MZ siblings are more similar to each other than to their DZ siblings in al four characteristics. The MZs are most similar in the traits of PERSISTENCE, APROACH, and ACTIVITY: Their raw scores in these areas are within one standard deviation from each other. Their APROACH scores show a diference of .29 points, PERSISTENCE .44 points, and ACTIVITY only .04 points. Their scores on SENSORY REACTIVITY show a diference of .65 points, just over one standard deviation. In addition, these traits graph similarly as plus or minus characteristics. Their PERSISTENCE scores are both on the right, or plus, side of the chart and their SENSORY, APROACH and ACTIVITY scores are al on the left, or minus, side of the chart, indicating that the MZs are similar in their activity levels, their approach to new situations, their persistence in staying on-task, and the amount of sensory input neded to elicit a response to a situation. In comparison, D2 and D1 are disimilar to each other in al four characteristics. Their raw scores are more than one standard deviation from each other in al four areas. D1 is within one standard deviation to M1 in the area of PERSISTENCE, D2 is within one standard deviation of M2 in the area of SENSORY, and within one standard deviation of both M1 and M2 in APROACH and ACTIVITY. Table 6.3 ranks the four siblings acording to the degre of each characteristic they exhibit. Rank order is determined by how much of a characteristic each sibling exhibits, minus raw scores generaly indicating les of a characteristic than plus scores. The MZs consistently rank next to each other in each of the four characteristics. 76 Table 6.4. Ranking of raw scores on temperament characteristics D1 D2 M1 M2 PERSISTENCE 4 1 3 2 SENSORY 1 2 4 3 APROACH 4 1 2 3 ACTIVITY 1 2 3 3 A summary of temperament characteristics indicates the following: ? D1 is the least persistent, the most active, the most cautious in approaching new situations, and requires the least stimulation to evoke a response. ? D2 is the most persistent and the most curious in approaching new situations. D2 is also sen as having scores closest to the mean, an indication of stable temperament acording to CTS (1996-2000). ? M1 and M2 track similarly and are indicated as having low persistence, a moderately low activity level, are mildly curious in approaching new situations and are les sensitive to outside stimulation than the DZs. 6.3. Comparison of temperament characteristics and language acquisition scores Temperament links to language acquisition are ilustrated by comparing the temperament characteristics of the siblings to their rates of language acquisition acording to their DS and MLU scores. Although DS data was too low to provide meaningful statistical analysis, the DS rankings of the siblings when the unreliable data is included is consistent with the rankings of temperament characteristics of the siblings. D2 has the highest DS score at each age, followed by M1 or M2 and then D1, with the exception of age 4 where D1?s score is higher than the MZs. 77 MLU scores do not show as neat a paralel with temperament scores as DS scores do. Tables 6.5 and 6.6 show the MLU, DS, acquisition of monophthongal [a:], and CTS rankings of the siblings when using raw scores. CTS rank remains the same over the five data points as the behavioral profile aseses temperament from age 3 to age 6. Post-diphthongal acquisition of [a:] difers for each sibling and reflects the age at which the diphthong /ai/ was fully acquired. 78 Table 6.5. MLU DS, acquisition of [a:], and CTS rankings of the DZ siblings D1 D2 CST CST A g es MLU DSS [a:] persistence Sensory Aproach activity MLU DSS [a:] Persistence Sensory Aproach activity 3 3 4 1 4 1 1 1 1 1 2 1 2 3.5 2 4 4 1 4 1 1 1 1 1 2 1 2 4 3 2 4 1 4 1 1 1 1 1 2 1 2 4.5 3 4 2 4 1 4 1 1 1 1 1 2 1 2 6 4 4 3 4 1 4 1 1 1 1 1 2 1 2 Table 6.6. MLU, DS, acquisition of [a:], and CTS rankings of the MZ siblings M1 M2 CST CST A ge s LU DS [a:] persistence Sensory Aproach activity LU DS [a:] Persistence Sensory Aproach activity 3 2 3 4 2 3 4 2 3 3 3 3.5 3 3 2 3 4 2 3 2 2 3 2 3 3 3 4 4 3 2 3 4 2 3 2 4 2 2 3 3 3 4.5 2 2 3 3 4 2 3 4 3 3 2 3 3 3 6 2 2 4 3 4 2 3 3 3 2 2 3 3 3 79 The temperament ranking of the siblings in the areas of persistence and approach show the same hierarchy as the DS scores, that D2 is most persistent, most approaching, followed by the monozygots and then D1. The areas of SENSORY REACTIVITY and activity level difer slightly in the hierarchy, showing that D1 is most active and needs the most sensory input, followed by D2 and then the MZs. The results would sem to indicate that persistence and wilingnes to approach a new situation may have a greater bearing on language acquisition than activity level. Although D2?s MLU scores rank the same as the DST scores, the only age which follows the temperament ranking of D2, M1, M2 and D1 is age 6. Although the table shows D1?s MLU score ranking betwen the MZs , it must be remembered that T-test showed no significant diference betwen the MZs and D1 at most ages. More importantly, the MZs show a similar development to each other while the DZs show a disimilar development from each other and, for DS, from the MZs. Interestingly, the rankings for acquisition of monophthongal [a:] at age 6 show D1 ranking betwen the MZs. Although the MZs are nearly identical at ages 4 and 4.5, Chi-Square analysis shows no significant diference betwen the MZs and D1 at age 4.5 or betwen M2 and D1 at age 6, indicating that, as with MLU and DST, the MZs and D1 are developing similarly while D2 is disimilar. Although the remaining areas of temperament have not been studied as to their influence on language acquisition, it interesting to note that the MZs appear to be les distractible than the DZs, who are nearly tied in their raw scores on DISTRACTIBILITY, ?the efectivenes of extraneous environmental stimuli in interfering with ongoing 80 behaviors? (CTS, p. 24). Additionaly, the MZs rank 2 and 4 in the area of INTENSITY, ?the energy level of responses regardles of quality or direction? (CTS, p. 24). 81 CHAPTER 7 DISCUSION 7.0. Introduction The purpose of this retrospective study of first language acquisition was to determine if biologicaly identical children and biologicaly non-identical children acquire language similarly when they are raised in a similar environment. A second research question was whether diferences in temperament could help to explain any diferentiation. A null hypothesis of no diference betwen the siblings was tested for both these questions. Data from two DZ and two MZ same-sex siblings were examined for this study. Previously recorded audiotapes of the siblings from ages 2 to 6 were analyzed for evidence of acquisition of gramatical structure and dialect features. Parental asesment of the siblings? temperament characteristics was used to elucidate acquisition results. The study finds that despite a similar environment for al siblings, they are not developing similarly, either gramaticaly or phonologicaly. Rather, the linguistic data and temperament measures patern to some extent along zygotic lines. Results also suggest that the temperament traits of PERSISTENCE and APROACH may be more important than the traits of ACTIVITY or SENSORY/reactivity. 82 7.1. Grammatical development This study evaluates gramatical development in four ways: 1) MLU, the average number of morphemes per utterance, 2) DST, the development of sentence structures in readines for DS testing, 3) phrasal development, the increase in longer phrases and clauses and the decrease in one and two-word phrases 4) DS, the development of gramatical structures within sentences, such as complements, wh questions, and negative reversals. Gramatical analysis of MLU and phrasal development indicate that diferences in language development tend to disappear as siblings age. In contrast, DST shows diferences increasing with age. The DS analysis was inconclusive due to the retrospective nature of the study. Al measures indicate diferences in gramatical development among the siblings; thus the null hypothesis of no diference is not supported. 7.1.1. MLU analysis 18 MLU results, which did not support the nul hypothesis of no diference betwen the siblings, indicated that at least two of the siblings, D1 and D2, are not developing similarly. Results for the group as a whole indicate that the siblings are more diferent at earlier ages, but that diferences are non-significant by age 6. At the ages where reliable data exists (3, 3.5 and 4), no significant diference is found betwen D1 and either M1 or M2. Although D1 is similar to the MZs at early ages and D2 is similar to them at later 18 Because of the retrospective nature of the study, 50 utterances were not available for al of the siblings at al of the data points. Two MLU scores based on 49 utterances were used to estimate MLU and are indicated as such on the tables in the previous chapters. Scores lower than 49 were considered unreliable and were not used to test for statistical diferences. 83 ages, similarity betwen the MZs themselves cannot be determined as reliable data for the pair is found only at age 3, when they show a significant diference, and at age 6, when no significant diferences are found. Thus, the MLU results from this study are neither consistent nor inconsistent with the findings of Ganger (1998) who reports similar development in MZ twins at ages 1-3. Although data for M1 and M2 was inconclusive, the diferences betwen D1 and D2 indicate that the null hypothesis is not supported by MLU testing. 7.1.2. DST and DS analysis Developmental Sentence Structure (DS), a measure used to determine children?s gramatical development, was determined using Le?s (1974) guidelines for scoring advanced gramatical structures (Appendix B). Scores computed on fewer than Le?s suggested 50 utterances were considered unreliable and were not used for statistical analysis. Although it was not possible to use the DS measurement to fully test the null hypothesis due to gaps in the data, significant diferences in DST scores and scores of phrasal development indicate that the null hypothesis of no diference is not supported by these testing measures. DST, designed to indicate readines for analysis of gramatical development, has not previously been used to characterize development itself. However, although the results cannot be claimed to be definitive, the data provides some counter evidence to the null hypothesis, due to the similarities betwen M1, M2, and D1 at earlier ages and significant diferences betwen D2 and the other siblings at al ages except age 4. Interestingly, the analysis of DST scores, which represent the percent of sentence 84 structures in each child?s corpus of utterances, reveals a patern diferent from the MLU results. Acording to DST, there is les diference at earlier ages and more at age 6. For example, D1 and M1 are not significantly diferent across the five data points, and both are diferent from M2 only at ages 4.5 and 6, indicating that D1 and M1, siblings who share at least 50% of their DNA, are paterning similarly. At age six, however, diferences appear betwen M1/M2, M1/D2, M2/D1 and D1/D2. A secondary analysis of the siblings? corpora for phrasal development, done to determine readines for DS scoring, shows D2 ready at age 3, D1 and M2 ready at age 3.5, and M1 ready at age 4.5. Although the MZs and D1show readines at diferent times, Chi-Square testing showed the diference betwen the thre was not significant. Thus, similar to the DST analysis, which shows early similarity and later divergence, this analysis shows thre of the siblings developing similarly up to age 4.5 when D1 diverges. D2?s readines at age 3 and D1?s divergence at 4.5 indicates the siblings phrasal development is not similar; thus the nul hypothesis is not supported. Because of the retrospective nature of the study and the rigor of Le?s protocol, it was possible to reliably compare the siblings? DS scores only for M1 and D2 at age 6. Although only this one comparison is available, it is interesting to note that, at this age, there is a significant diference betwen the two siblings at the level of p< .003. 7.1.3. Comparison of MLU and DST/DS scores It is also interesting to note that unanticipated decreases in MLU scores occur as the siblings age. D1 and D2 are the only siblings with valid, or nearly valid, MLU data across al ages, and both show a drop in MLU score?D2 at age 4 and D1 at age 4.5. Le 85 (1974) asumes that children cannot speak beter than their gramatical ability alows, but they may choose not to use certain constructions (p. 67). While this may be an artifact of the study?s reliance on retrospective data, nonetheles, it is of interest that these unexpected results appear around age 4. One explanation is that age four may be a pivotal age in child cognitive proceses. The findings in this study in regard to MLU coincide with the cognitive studies of Bloomquist (2003) and Welman (1990) who show divergent results at this age. Although cognitive proceses are beyond the scope of this study, it is possible that the MLU results document a pivotal proces in language acquisition which co-occurs with other changes in cognitive procesing. 7.2. Regional dialect forms Speakers of Southern English variably monophthongize /ai/ to [a:] and use the alveolar nasal in nouns, as in [a:m let t! "! mitn] ?I?m late to the meting.? To test the null hypotheses for the acquisition of these Southern dialect features, tokens were counted for variants of /ai/ and (ing) after the acquisition of the diphthong [ai] and the velar nasal [!] was established. Tokens for both phonological variables were analyzed by following sound, part of speech, word frequency and word stres. The acquisitional paterns of the siblings were also compared to descriptions of Southern adult paterning. The corpora indicated that the MZs acquired the diphthong [ai] similarly, at age 3.5. Chi-Square analysis of the rate of diphthongization showed no significant diference betwen the MZs except at age 3 (p<.025). These siblings difered from both D2 who acquired the diphthong at age 3.0 and from D1 who did not demonstrate diphthong use 86 until age 4.5. Chi-Square analysis showed a significant diference betwen D1 and D2 of p<.025 at age 3.5 and p<.001 at ages 4.5 and 6. The data are inconsistent with the null hypothesis, suggesting that M1 and M2 are developing similarly while D1 and D2 are not. For both acquisition of Southern monophthong [a:] and alveolar variant [n], the siblings were found to be developing a patern unlike adult modeling, a finding in keeping with Roberts (1994) and Cole (1980). 7.2.1. Analysis of monophthongization Although there were not enough tokens of /ai/ to alow for the testing of the hypothesis for the conditioning of this phonological alternation, analysis of the pooled data revealed that following liquids and nasals were more likely to be asociated with a preceding monophthong than al other categories. Several researchers (Bailey & Bernstein, 1989; Hazen, 2000; Head, 2003; Fridland, 2003) have reported monophthongs asociated with voiced obstruents>voiceles obstruents in producing a preceding monophthong. However, the sibling data showed obstruent voicing not to be a conditioning factor. It is interesting to note that D2, who produced more tokens than the other siblings, shows an emerging tendency to produce monophthongs preceding unvoiced obstruents and glides as wel as before nasals and liquids, showing a possible emerging Southern patern similar to that found by Fridland (2003). Fridland (2003) and Head (2003) found that for Southern adults, word boundary, which is equivalent to open syllables, was asociated with preceding monophthongization. Hazen (2000), however, although he did not specificaly addres 87 word boundary, indicates that closed syllables were asociated with increased monophthong production in multi-syllable words. The present study did not specificaly addres word boundary; however, in light of these studies, the results for pause, which is equivalent to open syllables, in both the pre and post-diphthongal sibling data are interesting. Pre-diphthongal results (N=3) show no monophthongization in pre-pause contexts. Post dipthongal results (N=14) show 21% monophthongization asociated with following pause. Whether the pre-diphthongal data indicate a phonological dificulty or the post-diphthongal data indicate the emergence of a patern is dificult to determine. In contrast to Southern adults, the sibling group as a whole produced more diphthongal [ai] than monophthongal [a:] in every speech category. However, as with Head?s (2003) adults and children, the pronoun I was more frequently monophthongized than other lexical items. Only 10% (N=4) of nouns contained a monophthong, indicating a patern unlike Southern adult paterning reported by Bowie (2001) in which nouns were commonly monophthongized. Thre of the four monophthongized nouns produced by the siblings were from the data of D2, who, with a monophthongization rate of 89% at age 6, shows a possible emerging patern in keeping with the adult patern. It may be that the low rate of monophthongization for M1, M2 and D1 reflect the cross-generational decrease in monophthongization reported by Thomas (1997) and Bowie (2001) for Texas and Maryland respectively. If this is the case, thre of the siblings may be acquiring an emerging Southern patern, with the fourth showing greater evidence of the more traditional patern. 88 7.2.2. Analysis of (ing) The variant [#] first appears for D2 at age 2.5 and for the other siblings at age 3. The siblings produced a total of 67 (ing) tokens across al five ages, a number too low to reliably test the null hypothesis. Al used 53% to 72% alveolar variant [n] across the age range. This is somewhat lower than the rate Roberts (1994) finds for 3 and 4 year-olds in Philadelphia, but may indicate that a high percentage of alveolar [n] is a common element of early childhood phonological development. The sibling patern in the pooled data appears to miror adult paterning (Wald and Shopen, 1983) insofar as pronouns (something, nothing) and verbs are most likely to contain the alveolar variant. However, no noun forms with alveolar [n] were found in the sibling data, although anecdotal evidence (Sabino personal conversation, 7/13/06) indicates Southerners commonly produce alveolar [n] rather than velar [!] in noun forms 7.2.3. Analysis of y?all and fixin? to Due to insufficient data, the null hypothesis could not be tested for either fixin to or y?all. 7.2.4. Acquisition of regional dialect forms The siblings are developing similar paterns of monophthongization, but at diferent rates. A Chi-Square analysis of the siblings? monophthongization of /ai/ revealed significant diferences betwen D1 and D2 at ages 3, 4.5 and 6. No diferences were found betwen the M1 and M2 across the five age groups, indicating that the MZs are developing along similar lines while the DZs are not. Since D2 was found to develop 89 [!] by age 2.5 and the others at age 3, the null hypothesis is not supported for acquisition of either /ai/ or (ing). The results also indicate that the siblings are not yet approximating adult speech. This finding is consistent with Cole?s (1980) and Roberts? (2000) hypotheses that children construct dialect rules of their own at this level of phonology. 7.3. Temperament analysis Carey Temperament Scale questionnaires (1996) completed by the parents asesed sibling temperament in nine areas. Four of these, because of previous research in language acquisition, were of interest to this study: PERSISTENCE, SENSORY/REACTIVITY, APROACH TO NEW SITUATIONS, and ACTIVITY LEVEL. Although diferent pairs of the siblings show similarities in one or two characteristics, the MZ siblings? scores in these four areas are much closer to each other than to their DZ siblings. M1 and M2 difer les than one standard deviation in al four characteristics; D1 and D2 show diferences of more than one standard deviation in al four characteristics. Moreover, when the siblings are ranked acording to the degre of each characteristic that each manifests, the MZs are ranked next to each other in every category. In the areas of PERSISTENCE, APROACH, and ACTIVITY level, the MZs rank betwen D1 and D2, indicating that although M1 is similar to M2, D1 is not similar to D2. Previous studies (Paul & Kelogg, 1997) have suggested that more active children acquire language more slowly than their les active counterparts. This finding is consistent with data produced by D1 whose ACTIVITY score is much higher than those of the other siblings, and whose language development is progresing more slowly. 90 However, the MZs scored lowest in ACTIVITY level, yet showed lower MLU than the more active D2. The sibling results for PERSISTENCE and APROACH are similar to Martin and Holbrook?s (1985) correlation betwen those characteristics and first grade reading scores. D1 ranks lowest in both PERSISTENCE and APROACH and also ranks lowest in MLU and DS at age 6. D2, who is acquiring language most quickly, scored higher than the others in both PERSISTENCE and wilingnes to APROACH new situations, and it may be that these characteristics have greater weight than ACTIVITY level regarding the acquisition of language. This is in keeping with the findings of Paul and Kelogg (1997) who found a high correlation betwen wilingnes to approach new situations and higher MLU scores. 7.4. Limitations of the study The primary limitation of the study was its retrospective nature. This resulted in insufficient data at several points. Although the study followed Miler and Chapman (1981) in using 50 utterances in place of Brown?s recommendation of 100 utterances, the prerecorded tapes did not contain 50 utterances for al of the siblings at al five ages. As a result, reliable comparisons could not be made for the MZs except at ages 3 and 6. The available evidence, however, does not support the null hypothesis of no diference betwen the siblings gramatical development. A similar limitation existed in the case of DST and DS data. 100 unique utterances were not available at al ages to determine DST scores, resulting in the necesity of a secondary test for DS readines, an increase in phrasal complexity. Even using both DST and phrasal complexity as readines indicators for DS, some of the 91 results can only be considered an estimate of readines. In addition, the 50 unique clausal structures necesary for DS analysis were not available for al siblings at al ages. As a result, a comparison was possible only for D2 and M1 at age 6. As with the MLU, available results do not support the null hypothesis of no diference. Pooled tokens of /ai/ for the siblings totaled 450 over the 5 ages. More tokens from each sibling would have resulted in greater acuracy in determining the siblings? patern. In addition, existing studies concentrate on adult acquisition of regional dialect features. Few studies addres child acquisition of dialect forms. The recordings contained a mid-central alophone of /ai/ not previously reported in the acquisition literature. This variant emerged early and disappeared by age 4.5. The order of acquisition suggests [!] is acquired before [a:] constrained by Southern phonological conditioning. It is unfortunate that [!] tokens acounted for only 13% (N=17) of the pre-diphthongal tokens, and, thus, did not alow for a ful analysis of the variant in child language development. Pooled tokens of (ing) yeilded only 67 across the five ages, alowing for only a speculative analysis of child variation. Also disappointing to the study is the lack of evidence of lexico-syntactic features yall and fixin to. Although anecdotal evidence indicates use of yall by at least one sibling, only two examples of yall were available on the tapes. The instruments used to ases gramatical development appear also to be a limiting factor. Unexpectedly, MLU and DST results difer. Although this diference may reflect the gaps in the data, it is possible that since MLU measures acquisition of 92 morphemes and DS measures gramatical complexity, the two asesment tools are actualy measuring diferent types of development. It is unknown which one best measures acquisition of gramar. MZ twins share 100% of their DNA while DZ twins share at least 50% of their DNA. Full interpretation of the relationship betwen D1 and D2 was limited because no efort was made to determine the amount of DNA they shared. Thus, it is possible that the similarities in MLU scores and phrasal development betwen D1 and the MZs are a result of a substantialy greater amount of shared DNA. 7.5. Implications for future research Because it is unknown whether the diferences found betwen MLU results and DS results reflect the gaps in the data or diferences in the testing measures themselves, more research is needed on these and other instruments used to ases gramatical development so that the best measurement tools can be used for future investigators. This study found a pre-diphthongal alophone of /ai/ which has not been reported in previous studies on child language acquisition. Tokens of [!] were too few to make a detailed analysis of its acquisition, and no studies were available for reference. Child acquisition of this variant bears further study to determine the ful patern of monophthong acquisition and its possible impact on the acquisition of Southern speech. Future nature/nurture studies in the area of language development would benefit from biological measurements which can determine levels of shared DNA among DZ siblings. Knowing the degre to which DZ siblings share inherited traits would more fully explain the efects of genetics on child language development. 93 7.6. Conclusions If language acquisition were due only to environmental influences, the results of this study would support the null hypothesis of no diference by indicating similarities among al four siblings, due to their consistent environment. On the other hand, if acquisition were due only to biological influences, the results would indicate MZ similarity and DZ disimilarity across al five data points. In fact, the results are mixed. MLU results show some disimilarity betwen the MZs at earlier ages along with some similarity betwen the MZs and D1. Although occasional gaps in the data exist because of the retrospective nature of the study, the sibling data are inconsistent with the null hypothesis of no diference at several points. On the one hand, although the monozygotic siblings share 100% of their genetic make-up, diferences can be sen in their language development and aspects of their temperament. The DZs show significant diferences from each other in every area: MLU, DST, DS, acquisition of regional forms, and temperament. The study also indicates greater similarity betwen the MZs than betwen DZs in MLU scores, DST scores, phrasal development, and acquisition of regional forms. Results also indicate greater similarity betwen the MZs twins in areas of temperament (ACTIVITY level, SENSORY/REACTIVITY, PERSISTENCE and APROACH) that previous studies (Dixon & Smith, 1997; Paul & Kelogg, 1977; Wong-Filmore, 1979) have shown to impact language acquisition. These results suggest that despite an environment that is more consistent than has been reported in previous twin studies, biologicaly identical children do not develop 94 language identicaly. The results also suggest that biologicaly similar children diverge to an even greater extent. At least part of the explanation for the acquisitional diferences may lie in inherited temperament traits; that the combination of wilingnes to APROACH new situations and PERSISTENCE may be of more importance, especialy in language development, than either ACTIVITY level or SENSORY/REACTIVITY. The present study was possible because of the presence of same-sex MZ and DZ siblings within a familial environment that controlled for both nature and nurture. Although similar data sets are rare, multiple-birth siblings are becoming more common due to recent advances in fertility techniques, increasing the possibility of similar sets of multiple-birth, same-sex siblings. Future studies of these subjects would provide additional information to the growing body of research on the efects of nature and nurture on the acquisition of child language. 95 REFERENCES Anderson, B. (2002). Dialect leveling and /ai/ monophthongization among African- American Detroiters. Journal of Sociolinguistics, 6 (1), 91-98. Bailey, G., & Bernstein, C. (1989). Methodology of a phonological survey of Texas. Journal of English Linguistics, 22 (1), 6-16. Bailey, G., & Tilery, J. (1996). The persistence of Southern American English. Journal of English Linguistics, 24 (4), 308-321. Bernstein, C. (2003). Gramatical features of Southern speech: y?all, might could and fixin to. In S. J. Nagle & S. L. Sanders (Eds.), English in the Southern United States (pp 106-118). New York: Cambridge University Pres. Bernhardt, B., & Stemberger, J.P. (1998). Learning the napping from perceived to articulated. Procedings from the regional metings, Chicago Linguistics Society, 34 (4), 359-72. Berthele, R. (2002) Learning a second dialect: A model of idiolectal disonance. Multilingua, 21(4) 327-44. Blake, J., Quartaro, G., & Onarati, S. (1993). Evaluating quantitative measures of gramatical complexity in spontaneous speech samples. Journal of Child Language, 20 (1) 39-52. 96 Bloomquist, J. (2003). Cross-cultural semantic acquisition: Evidence from over- extensions in child language. Doctoral Disertation, University of New York at Buffalo. Borden, G., Haris, K., & Raphael, L. (1994). Speech Science Primer. Philadelphia: Lippincott, Wiliams, & Wilkins. Bowie, D. (2001). The Diphthongization of /ay/: Abandoning a Southern norm in southern Maryland. Journal of English Linguistics, 29 (4), 329-45. Bretherton, I., McNew, S., Snyder, L., & Bates, E. (1983). Individual diferences at 20 months: Analytic and holistic strategies in language acquisition. Journal of Child Language, 10 (2), 293-320. Brown, R. (1973). A first language: The early stages. Cambridge: Harvard University Pres. Carey Temperament Scales: Quickscore software, test manual and user?s guide. (1996- 2000). Scottsdale, AZ: B-DI Publishers. Carey, W. B. (1970). A simplified method for measuring infant temperament. Journal of Pediatrics, 77, 188-194. Carmel, T. (1997). Spectrogram reading: Spectral cues for the broad categories of speech sounds. [On-line]. Available: http:/cslu.cse.ogi.edu/tutordemos/SpectrogramReading/ipa/ipahome.html. Ching, M. (1987). How fixed is fixin? to? American Speech, 62 (4) 332-245. Chomsky, N. (1986). Knowledge of language: Its nature, origin, and use. New York: Praeger. 97 Cole, L. (1980). A developmental analysis of social dialect features in the spontaneous language of preschool black children. Doctoral disertation, Northwestern University, 1980. Conway, D., Lytton, H., & Pysh, F. (1980). Twin-singleton language diferences. Canadian Journal of Behavioral Science, 12, 264-271. Crystal, D. (1974). A first language: The early stages. Journal of Child Language, 1 (2), 289-304. Dale, P. S., Dionne, G., Eley, T., & Plomin, R. (2000). Lexical and gramatical development; a behavioural genetic perspective. Journal of Child Language, 27 (3) 619-42. Dale , P. S., & Goodman, J. C. (2005). Commonality and individual diferences in vocabulary growth. In M. Tomaselo & D. I. Slobin (Eds.), Beyond Nature Nurture: Esays in Honor of Elizabeth Bates. (pp. 41-78). Mahwah, NJ: Lawrence Erlbaum Asociates. Delacorte, M., Benedict H., & Klein D. (1983). The relationship of pragmatic dimensions of mother?s speech to the referential expresive distinction. Journal of Child Language, 10, 35-44. Deser, T. (1989). Dialect transmision and variation: An acoustic analysis of vowels in six urban Detroit families. York Papers in Linguistics, 13, 115-28. Dixon, W. Jr., & Smith, P. (2000). Links betwen early temperament and language acquisition. Meril-Palmer Quarterly, 46 (3) 417-440. 98 Fikkert, P. (in pres) Acquisition of phonology: Handbook of phonological theory. P. deLacy (Ed). Cambridge U. Pres. [online]. Available: ww.fikkert.com/publicationsabel.htm. Fischer, J. L. (1958). Social influences on the choice of a linguistic variant. Word, 14, 47- 57. Fridland, V. (2003). ?Tie, tied and tight?: The expansion of /ai/ monophthongization in African-American and European-American speech in Memphis, Tennese. Journal of Sociolinguistics, 7 (3), 279-298. Ganger, J. B. (1998). Genes and environment in language acquisition: A study of early vocabulary and syntactic development in twins. Doctoral disertation, Masachusets Institute of Technology. Ganger, J., Wexler, K., and Soderstrom, M. (1998). The genetic basis for the development of tense: A preliminary report on a twin study. In A. Grenhil et al. (Eds.) BUVLD 22 Procedings, 224-234. Gelman, S.A., & Welman, H.M. (1991). Insides and esences: Early understandings of the non-obvious. Cognition, 38, 213-44. Genese, F., Nicoladis, E., & Paradis, J. (1994). Language diferentiation in early bilingual development. Journal of Child Language, 22, 611-31. Goldfield, B., & Snow, C. (1989). Individual diferences in language acquisition. In J. Gleason (Ed). The development of language. New York: Meril Pres. 99 Haris, K. L. (2004). A comparison of phonological features in normally developing African American English speakers in thre age groups. Master?s thesis, Auburn University, 2004. Hay, J., Jannedy, S., & Mendoza-Denton, N. (1999). Oprah and /AY/: Lexical frequency, refere design and style. Procedings of the14 th International Congres of Phonetic Sciences. [online] Available: http:/ww.ling.canterbury.ac.nz/jen /documents/icphs.html. Haynes, W. O., Purcel, E. & Haynes, M. D. (1979). A pragmatic aspect of language sampling. Language, Speech, and Hearing Services in Schools. 10 (2), 104-110. Haynes, W. O., & Shulman, B. B. (1998). Communication development: Foundations, proceses, and clinical applications. Baltimore: Wiliams and Wilkins. Hazen, K. (2000). A methodological suggestion on /aj/ ungliding. American Speech, 75 (2), 221-24. Head, A. (2003). The monophthongation of /ai/ in Elba and the environs. Master?s thesis, Auburn University, 2003. Hemeter, M. L. & others. (1986). Paper presented at the Conference on Human Development. Nashvile, TN, April 3-5. Jorgensen, J.N. (1998). Children?s acquisition of code-switching for power-wielding. In P. Auer (Ed.), Code switching in conversation: Language, interaction and identity. (pp. 237-58). London: Routledge. 100 Kehoe, M., & Stoel-Gamon, C. (2001). Development of syllable structure in English- speaking children with particular reference to rhymes. Journal of Child Language, 28 (2). 393-432. Kent, R. D., & Read, C. (1992). The Acoustic Analysis of Speech. San Diego: Singular. Kle, T., & Fitzgerald, M.D. (1985). The relation betwen gramatical development and mean length of utterance in morphemes. Journal of Child Language, 12 (2) 251- 69. Labov, W. (1972a). Language in the inner city: Studies in Black English Vernacular. Philadelphia: University of Pennsylvania Pres. Labov W. (1972b). Sociolinguistic patterns. Philadelphia: University of Pennsylvania Pres. Ladefoged, P. (1996). Elements of acoustic phonetics, 2. Chicago: University of Chicago Pres. Larsen, S. (1975). Developmental sentence analysis. Journal of Learning Disabilities, 8 (7), 467-468. Le, L. (1974). Developmental sentence analysis. Evanston, IL: Northwestern University Pres. Locke, J. L., & Mather, P. L. (1989). Genetic factors in the ontogeny of spoken language: Evidence from monozygotic and dizygotic twins. Journal of Child Language, 16 (3), 553-59. 101 Martin, R. (1988). The Temperament Asesment Battery for Children. Brandon, VT: Clinical Psychology. Martin, R. P., Drew, K. D., Gaddis, L.R., & Moseley, M. (1988). Prediction of elementary school achievement from preschool temperament: Thre studies. School Psychology Review, 17 (1), 125-137. Martin, R., & Holbrook, James. (1985) Relationship of temperament characteristics to the academic achievement of first-grade children. Journal of Psychoeducational Asesment, 3, 131-140. Matheny, A. P., Jr., & Bruggemann, C. (1972). Articulation proficiency in twins and singletons from families of twins. Journal of Speech and Hearing Research, 15, 845-851. Maynor, N. (1996). The pronoun y?all: Questions and some tentative answers. Journal of English Linguistics, 24, (4), 288-294. Miler, J.F., & Chapman, R. S. (1981). The relation betwen age and mean length of utterance in morphemes. Journal of Speech and Hearing Research, 24, 154-61. Nelson, K. (1981). Individual diferences in language development: Implications for development and language. Developmental Psychology. 17 (2), 170-187. O?Grady, W., Aronoff, A.J., & Res-Miler, J. (2005). Contemporary Linguistics (5 th ed.). New York: Bedford St. Martins. Otomo, K., & Stoel-Gamon, C. (1992). The acquisition of unrounded vowels in English. Journal of Speech and Hearing Research. 35 (3), 604-16). 102 Paul, R., & Kelogg, L. (1997). Temperament in late talkers. Journal of Child Psychology, 38, 803-11. Payne, A. (1980). Factors controlling the acquisition of the Philadelphia dialect by out-of-state children. In W. Labov (Ed.), Locating language in time and space (pp. 143-178). New York: Academic Pres. Pierce, S., & Bartolucci, G. (1976). Phonological production and perception in verbal autistic, normal, and mentaly retarded subjects. Papers and Reports on Child Language Development, 12, 189-198. Plomin, R. (1994). Genetics and experience: The interplay betwen nature and nurture. Thousand Oaks, CA, Sage Publications. Pollack, K.E., and Keiser, N. J. (1990). An examination of vowel erors in phonologicaly disordered children. Clinical Linguistics and Phonetics. 4 (2), 161-178. Roberts, J. L. (1994). Acquisition of variable rules: (-t,d) deletion and (ing) production in preschool children. Doctoral disertation, University of Pennsylvania. Roberts, J. (1997). Hiting a moving target: Acquisition of sound change in progres by Philadelphia children. Language Variation and Change. 9, 249-266. Rondal, J., Ghiotto, M., Bredart, S., & Bachelet, J. (1987). Age-relation, reliability and gramatical validity of measures of utterance length. Journal of Child Language. 14 (3), 433-446. Rymer, R. (1993). Genie: An abused child?s flight from silence. New York: Harper Collins. 103 Shatrick, R. (1980). The forbidden experiment: the story of the Wild Boy of Aveyron. New York: Farar Straus Giroux. Shatz, M. (1994). A toddler?s life: Becoming a person. New York: Oxford University. Pres. Shuy, R., Wolfram, W., & Riley, W.K. (1968). A study of social dialects in Detroit. (Final Report, Project 6-1347). Washington, D.C.: Ofice of Education. Staford, L. (1987). Maternal input to twin and singleton children: Implications for language acquisition. Human Communication Research, 13, 429-462. Starks, D., & Bayard, D. (2002). Individual diferences in the acquisition of postvocalic /r/: Daycare and sibling order as potential variables. American Speech, 17 (2), 184-194. Thomas, Eric R. (1997). A rural/metropolitan split in the speech of Texas Anglos. Language Variation and Change, 9, 309-322. Tilery, J., Wikle, T., & Bailey, G. (2000). The nationalization of a Southernism. Journal of English Linguistics, 28 (3), 280-294. Tomaselo, M., Mannle, S., & Kruger, A. (1986). Linguistic environment of 1- to 2-year- old twins. Developmental Psychology, 22, 169-176. Wald, B & Shopen, T. (1985). A researcher?s guide to the sociolinguistic variable (ING). In V. Clark, P. Eschholz &, A. Rosa. (Eds) Language: Introductory readings. (pp. 515-542). New York: St. Martins Pres. 104 Wilson, R. S., & Mathey, A. P., Jr. (1986). Behavior-genetics research in infant temperament: The Louisvile twin study. In R. Lomin and J. Dunn (Eds), The Study of temperament: Changes, continuities and challenges. Hilsdale, NJ: Lawrence Erlbaum Asociates. Wong-Filmore, L. (1979). Individual diferences in second language acquisition. In C. W. Filmore (Ed.), Individual diferences in language ability and language behavior. (pp. 203-228). New York: Academic Pres. 105 APENDIX A Computation of MLU Guidelines from Brown (1973) pg 55. Morphemes in 100 utterances are counted for each child. The total is then divided by 100. How to determine utterances: 1. Start with second page of the transcription unles that page involves a recitation of some kind. In this later case, start with the first recitation-fre stretch. Count the first 100 utterances satisfying the following rules. 2. Only fully transcribed utterances are used, none with blanks. Enter portions of utterances in () to indicate doubtful transcription is used. 3. Include al exact utterance repetitions. Stuttering is marked as repeated eforts at a single word; count the word once in the most complete form produced. In the few cases where a word is produced for emphasis or the lie, (no no no) count each occurrence. 4. Do not count such filers as m or oh, but do count no, yeah, and hi 5. Al compound words (two or more fre morphemes), proper names and ritualized reduplications count as single words. Examples: birthday, rackety-boom, choo-choo, night-night, se saw). Justification is that there is no evidence that the constituent morphemes function as such for these children, but are sen as one morpheme. 106 6. Count as one morpheme al iregular pasts of the verb (got, did went, saw) Justification is that there is no evidence that the child relates these to present forms. 7. Count as one morpheme al diminutives (doggie, mommie) because these children at least do not sem to use the suffix productively. Diminutives are the standard forms used by the child. 8. Count as separate morphemes al auxiliaries, (is, have, wil, can, must, would) Also al catenatives: gonna, wanna, hafta, These later counted as single morphemes rather than as going to or want to because evidence is that they function so for the children. 9. Count as separate morphemes al inflections; for example, possesives, third person singular -s, regular past -ed, progresive -ing, plural -s, 107 APENDIX B Developmental Sentence Scoring Rubric (DS) S c o re Ind pronoun or noun modifiers Personal pronous Main verbs Secondary Verbs Negatives Conjunctions Interrogative reversals Wh- questions 1 Is, this, that 1 st and 2 nd person A. uninflected verbs B. copula is or ?s C. is+verb+ing Copula or auxilary reversal+not Copula reversal 2 3 rd person A. Regular inflected verbs, B. irregular past C. Copula am, are, was, were D. auxiliary am, are, was were Early infinitives Leme, wana, gota Who, what, what+noun B. Where, how many 3 No, some, more, al lot(s), one(s), other, another Plurals, we, us, our, these, those Non- complementing infinitives: I stoped to play, I?m afraid to lok and 4 A .Can, wil, may+verb B. Obligatory do+verb C. Emphatic do+verb Participle, present or past Can?t don?t Reversal of auxiliary be 108 5 Reflexives: myself, himself, etc A. Infinitival complements B. obligatory deletions: make it [to] go. C. Infinitive with wh-word Isn?t won?t But, so, and so, so that, or, if When, how, how+adj How do you do it? 6 Wh- pronouns: who, which, whose, whom Could, would, should,might+ver b B. obligatory does, did+verb C. emphatic does, did+verb because Obligatory do, does, did: Do they run? Didn?t it hurt B. reversal of modal C. tag question 7 A. Any, anything B. everybody, everyone C. both, few, most, next, first etc. His own, oneself, whichever, whatever A, pasive with get B. must, shal+verb C. have+verb+en D. have got A. Pasive infinitival with get: I have to get dresed. B. with be: I wana be puled. Al other negatives Why, what if, how come+gerun d Why are you crying? 8 A, have ben+verb+ing B. modal+have+ver b +en C. other auxiliary combinations: should have ben sleping Gerund Where, when, how, while + adjective+as if,m like, that, B. obligatory deletions: I run faster than you [run] C. wh- words+infinitive: I know how to do it. Reversal of auxiliary have: has he sen you? B. Reversal with two or three auxiliaries? Could he have ben crying? Whose, which, which+noun Whose car is that?