Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Articulatory characteristics of English /l/ in speech development Oh, Sunyoung 2005

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2005-104279.pdf [ 20.02MB ]
Metadata
JSON: 831-1.0092349.json
JSON-LD: 831-1.0092349-ld.json
RDF/XML (Pretty): 831-1.0092349-rdf.xml
RDF/JSON: 831-1.0092349-rdf.json
Turtle: 831-1.0092349-turtle.txt
N-Triples: 831-1.0092349-rdf-ntriples.txt
Original Record: 831-1.0092349-source.json
Full Text
831-1.0092349-fulltext.txt
Citation
831-1.0092349.ris

Full Text

ARTICULATORY CHARACTERISTICS OF ENGLISH IV IN SPEECH DEVELOPMENT by SUNYOUNGOH B.A., Sogang University, 1990 M.A., Sogang University, 1993 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (LINGUISTICS) THE UNIVERSITY OF BRITISH COLUMBIA April 2005 © Sunyoung Oh, 2005 ABSTRACT This dissertation investigates articulatory characteristics of English IV in child speech. The study is primarily based on experimental data collected using ultrasound imaging techniques from eight English children ages 3;11 to 5;9. Replicating previous articulatory studies of syllable-based allophones of IV in adult speech production, the articulatory components of III in child speech production are analyzed for the static information and relative timing between tongue movements. Secondarily, the acoustic analysis of this data and its perception judgments by adults are presented.' One of the major findings of this study is that children at these ages produce IV using different spatial and temporal coordination than adult speech production, although some children produce IV more similar to adult IV in terms of articulatory organization. Further, the findings are addressed in relation to speech motor development, and hypotheses are tested to see which motor developmental process(es) (differentiation, integration, refinement) can describe the acquisition of IV. The ultrasound results of the tongue movements in children's IV indicate that all general motor developmental processes are active in these children, and the spatial and temporal coordination of the articulatory gestures of IV is rather simplified or modified, and needs to be further refined. I argue that the tendency toward late acquisition of IV is due directly to the articulatory complexity of its spatial and temporal characteristics. This work contributes much-needeid empirical data of the articulatory characteristics of IV to both language acquisition and speech sciences^ and constitutes a novel application of ultrasound imaging to child speech research.. ii • Organization of this dissertation is as follows. Subsequent to the overall introduction of the study in Chapter 1, Chapter 2 presents the empirical background and hypotheses for the study. It reviews speech and developmental studies in production and perception conducted by other researchers, and proposes empirical questions. Chapter 3 provides the methodology for the study. It introduces ultrasound techniques and experiment design and procedure. Chapter 4 presents the results of the spatial characteristics of the children's III in terms of number of gestures, tongue shape, constriction location, and allophonic variation with respect to different syllable positions. Chapter 5 discusses the results of the temporal characteristics of the children's IV gestures. Inter-gestural timing of allophones of IV is examined to determine whether timing distinguishes positional allophones in these children's speech. Chapter 6 provides post-experiment perception judgments made by adults, and acoustic analysis of samples of tokens used in the current study. Finally, Chapter 7 summarizes the results and discusses the implications of the dissertation. iii TABLE OF CONTENTS Page A B S T R A C T ii T A B L E OF CONTENTS iv LIST OF T A B L E S viii LIST OF FIGURES ix A C K N O W L E G M E N T S xi C H A P T E R 1 Introduction 1 1.1 Introduction — 1 1.2 Theoretical B ackground 4 1.2.1 Articulatory Characteristics of English IV 5 1.2.2 Speech Motor Development 10 1.3 Empirical Questions 13 1.4 Organization of Dissertation 15 2 Studies in Articulatory Phonetics and Speech Development 17 2.1 Introduction 17 2.2 Articulation-based Studies 17 2.2.1 Articulatory Phonology 18 2.2.2 Allophonic Variation 21 2.2.2.1 Nasals 21 2.2.2.2 Stops 22 2.2.3 Ambisyllabicity 23 2.2.4 Cross-linguistic Studies of Liquids 24 2.3 Studies in Speech Development 27 2.3.1 Physiological Development— 27 2.3.1.1 Vocal Tract 27 2.3.1.2 Vocalizations 29 2.3.1.3 Motor Development 30 iv 2.3.2 Phonological D evelopment of/1/ 35 2.3.2.1 Speech Production 35 2.3.2.2 Speech Perception 37 2.4 Hypotheses 39 3 Experiment Methods 46 3.1 Introduction 46 3.2 Ultrasound Techniques 46 3.3 Methods 50 3.3.1 Participants 51 3.3.2 Stimuli 54 3.3.3 Apparatus 5 9 3.3.4 Procedure 60 3.3.4.1 Data Collection 60 3.3.4.2 Spatial Measurement 62 3.3.4.3 Temporal Measurement 66 4 Spatial Characteristics of IV in Children's Speech ~ 68 4.1 Introduction 68 4.2 Hypotheses 69 4.2.1 Gestural Coordination 69 4.2.2 Motor Development 71 4.3 Results 75 4.3.1 Number of Gestures and Constrictions 75 4.3.2 Tongue Shape- 79 4.3.3 Tongue Dorsum Backing 91 4.3.3.1 Individual Results 93 4.3.3.2 Age and Gender 96 4.4 Discussion 97 4.5 Conclusion 100 .1 ; v 5 Temporal Characteristics of IV in Children's Speech 102 5.1 Introduction 102 5.2 Hypotheses 105 5.3 Results --- 107 5.3.1 Relative Timing 108 5.3.1.1 Isolated Words 110 5.3.1.2 Two-word Combination 111 5.3.2 Individual Results 113 5.4 Discussion '-- — 122 5.5 Conclusion 124 6 Acoustic and Perceptual Evaluation 126 6.1 Introduction 126 6.2 Acoustic Analysis— — 127 6.2.1 Methods 130 6.2.2 Results 131 6.2.2.1 Age — 134 6.2.2.2 Gender 134 6.2.2.3 Individual Results— 135 6.3 Perceptual Judgment-;— — 145 6.3.1 Methods 145 6.3.2 Results 145 6.4 Discussion 150 6.5 Conclusion 151 7 Conclusion 153 7.1 Summary 153 7.2 Limitations — 155 7.3 Implications 156 . vi BIBLIOGRAPHY 158 APPENDIX A Lists of Utterances — 176 APPENDIX B Word List for Perception Judgment 183 APPENDIX C Statistics for Spatial Measurement: TD Backing 191 APPENDIX D Statistics for Temporal Measurement: Relative Timing 192 vii LIST OF TABLES Table Page 1-1 Summary Chart: Spatial Characteristics of English IV 9 1-2 Summary Chart. Temporal Characteristics of English IV 9 1- 3 Summary Chart: Three Motor Development Processes 13 2- 1 Summary Chart: Characteristics of Motor Development 34 3- 1 Participant List 52 3-2 Stimuli List 56 3- 3 Distribution of IV by Adults' Perception Judgment 57 4- 1 Distribution of Allophones of IV 79 5- 1 Mean of Intergestural Timing 112 5- 2 Distribution Chart: Intergestural Timing in Six Syllable Positions — 115 6- 1 Formant Frequencies (Hz) of IV in Adults and Child 129 6-2 Word List 131 6-3 Mean of Formant Frequencies: FI, F2, and F3 of III 132 6-4 Perceptual Judgment of IV by Adult Listeners 147 viii LIST OF FIGURES Figure Page 1- 1 Gestural Timing of English IV 9 2- 1 Gestural Timing of Nasal Iml 22 3- 1 2D Ultrasound Images of the tongue 47 3-2 Picture Stimuli Examples — 58 3-3 Experiment Setup 59 3-4 Examples of Tongue Shape Tracing 63 3- 4 Spatial Measurement Diagram 65 4- 1 Constrictions of IV: adult — 76 4-2 Constrictions of IV: children ^ 77 4-3 Constrictions of Itl, In/, IV: children 78 4-4 a Tongue Shape (Final VL#C): F L 81 4-4 b Tongue Shape (Ambisyllabic VL#V): F L — 82 4-4 c Tongue Shape (Initial V#LV): F L 82 4-4 d Tongue Shape (Intervocalic V L V ) : F L 83 4-5 Tongue Shape: JN —- 84 4-6 Tongue Shape: JL 85 4-7 Tongue Shape: K R 86 4-8 Tongue Shape: L R 87 4-9 Tongue Shape: M C 88 4-10 Tongue Shape: R C 89 4-11 Tongue Shape: A D 90 4-12 Tongue Location at Speech Rest Position 91 4-13 Speech Rest Position: Distance from reference point 92 4-14 TD backing and height-- 93 4-15 TD backing: individual results 95 4-16 TD backing by Age — 96 4-17 TD backing by Gender 97 ix c 5-1 Intergestural Timing (TT-TD) in Three Syllable Positions 109 5-2 Intergestural Timing: Age and Gender 110 5-3 Intergestural Timing: Isolated vs. Two-word combination. — 113 5-4 Intergestural Timing: A l l Children 114 5-5 Intergestural Timing: JL and F L 117 5-6 Intergestural Timing: K R and RC ~ 119 5-7 Intergestural Timing: JN (4; 10 and 5; 1) i- 120 5-8 Intergestural Timing: L R and M C 121 5- 9 Intergestural Timing: A D 122 6- 1 Spectrograms of Initial and Final III: Adult Speaker 128 6-2 Histograms of Formant Distribution (Hz) 133 6-3 Frequencies of FI and F2 by Age 134 6-4 Frequency of F2 by Gender 135 6-5 Spectrograms of III: JL 136 6-6 Spectrograms of IV: F L 137 6-7 Spectrograms of IV: JN 138 6-8 Spectrograms of IV: L R ~ 13 9 6-9 Spectrograms of IV: M C '- 140 6-10 Spectrograms of IV: K R 141 6-11 Spectrograms of/1/: A D * — 142 6-12 Spectrograms of IV: R C — 143 6-13 [w] Substitution for Initial IV: R C 144 x ACKNOWLEDGEMENTS There are many people who have made this dissertation possible and helped me through a long journey. I would like to acknowledge my wonderful committee: Bryan Gick, Joe Stemberger, and Janet Werker. Without their support and patience, I would not have been able to stand where I am today. Joe and Janet have provided much in terms of their knowledge of phonological development and language acquisition. Their expertise and comments have inspired my enthusiasm for wanting to understand more about language acquisition. I thank my supervisor Bryan Gick for his teaching and walking with me throughout the entire time. He not only introduced me to phonetics, but also made it a great pleasure for me. I could not have imagined having a better advisor, one who has befriended and guided me both intellectually and personally. Thank you. I also would like to thank my examiners—Barbara Bernhardt, Eric Vatikiotis-Bateson, and Rena Krakow—for reading my dissertation thoroughly, and for their helpful insights, advice, and critiques. M y sincere thank-you also goes to Doug Pulleyblank, who has convinced and encouraged me to stick with my ambitions in linguistics; to Rose-Marie Dechaine, who has taught me that a doctorate is the process of overcoming personal hardships and limitations in life; to Henry Davis for his great enthusiasms and love for students; and to Pat Shaw for her genuine support and interest in my research. Thanks to Sook Whan Cho for opening a door to linguistics, and Lise Menn for giving me inspiration. I have learned a great deal about linguistics from all of them, and they always made linguistics fun. I am indebted to my great colleagues Marion Caldecott, Yunhee Chung, Suzanne Gessner, Eun-SookKim, Masaru Kiyota, Add (Sugunya) Ruangjaroon, Kayono Shiobara, xi and Linda Tamburri-Watt for providing a happy environment to learn and grow in, and for supporting and understanding without words. Thanks to lab buddies Oladiipo Ajiboye, Penelope Bacsfalvi, Sohya Bird, Fiona Campbell, Ramona McDowell, and Ian Wilson for sharing similar interests and stresses. Much thanks to Shaffiq Rahemtulla for his technical support; and Julee Botting, Jason Chang, and Jonathan Howell for their assistance during experiments. Many thanks to all the children and parents who participated in the study and had to put up with whale and ladybug. Much thanks to Edna Dharmaratne. She is a home-away mom who always has been supportive, generous, and kind to me. Thanks to all the rest of the faculty and graduate students in the Department of Linguistics at the University of British Columbia, who make a great community in which I am proud to be a part. I also am grateful to my proof readers Wayne Egers and Anna'Mushynski for their great help. My friends Sooyoung Kim, Meesook Kim, Nancy Smith, and Sunnie Song deserve a huge hug for helping me to get through difficult times, and for all the emotional support, entertainment, and caring they provided; and Marc des Jardins for sharing spirituality. Lastly, I wish to thank my father and mother, Jung Yong Oh and Soon Ock Hwang for their tremendous love and support. They have stimulated my interest in the world around me, and always inspired me to explore without boundaries. I could not have finished this journey without "Mr. Jack", who not only, never minded sleeping in the sound booth for countless nights, but took the experience as exciting camping trips. He is a constant inspiration and the one who makes life beautiful. xii ' \' CHAPTER 1 I N T R O D U C T I O N 1.1 Introduction Speech, one of the most complex motor tasks humans encounter, requires control of the articulatory system of the vocal tract. In speech, the tongue is the articulator responsible for producing the largest number of contrasts. Models of early speech motor development have shown specific changes in articulatory coordination over time (Kent, 1983; MacNeilage & Davis, 1990; Oiler, 1978). However, the development of speech motor control has not been directly mapped onto language acquisition. The goal of this dissertation is to discover how speech motor behaviours are related to speech development. In particular, I will investigate young children's articulation of III in English, and test how tongue movements are coordinated in different allophones of IV. Although children are able to produce speech-like vocalizations before the end of the first year of life (Vihman, 1996), they do not master all the sounds in their ambient language until they are in their early school years (Kent, 1992; Sanders, 1972). Until they are fully capable of producing target sounds as adults do, young children have limited control of their speech motor system (Kent, 1992). Due to limited coordination skills, the speech production of young children may be systematically different from that of adults. Moreover, 1 coordinative limitations in motor development may be linked to phonemic developmental patterns that are universally found in the early emergence of sounds (Stoel-Gammon, 1985). The motor skills required for some tongue movements are reportedly acquired late in the early school years (ages five to eight) (Kent, 1992). For instance, although it is reported that IXI can emerge sporadically as early as 16 months (Bernhardt & Stemberger, 1998), the emergence of English IXI appears around age three (Sanders, 1972 as cited in Leonard, 1995), and some children do not master it until the age of nine (Bernhardt & Stemberger, 1998; Locke, 1980, 1993). IXI may be difficult for young children, because speech motor skills for tongue movements have not fully matured, and learning how to control different parts of the tongue may be closely related to the mastery of IXI. The acquisition of IXI in speech development has a number of interesting aspects. First, for a long time, IXI in adult speech in many languages has been known to exhibit allophonic variation based on syllable position: in English, IXI in the final position (e.g., "bell") is known to be acoustically "darker" than its "light" counterpart in initial position (e.g., "led") (e.g., Giles & Moll , 1975; Ladefoged, 1993). To acquire IXI, children must know that IXI sounds different in the initial and final positions. Second, IXI involves a complex articulation using two lingual constrictions: tongue tip raising and tongue dorsum backing (Browman & Goldstein, 1995; Sproat & Fujimura, 1993) along with lateral dipping. The other liquid Irl, affricates, fricatives /J"/, and glides also are reported to have multiple lingual gestures (Gick, 2002b; Pouplier, 2003; Studdert-Kennedy & Goldstein, 2003). To produce IXI, children must learn how to use two different parts of the tongue independently, but in coordination (Gick, 2002b). Third, syllable-based allophones of IXI are distinguished by relative timing and magnitude change between the two tongue movements involved in the articulation of IXI (e.g., 2 Sproat & Fujimura, 1993): in the initial position, the tongue tip and tongue dorsum movements occur almost simultaneously; in the final position, the tongue dorsum movement occurs earlier than tongue tip movement and retracts further, leading to a reduction in the magnitude of the tongue tip movement. To master all these complex characteristics of IV, children must know how to coordinate different parts of the tongue, according to syllable position. With the characteristics of III given above, it is not surprising that it is difficult for children to learn to accommodate the two lingual gestures of IV with the appropriate position-specific timing and magnitude changes. Further, the tongue itself is a unique and highly complex motor system in the human body, formed by a three-dimensional hydrostatic muscle mass with many degrees of freedom. Extrinsic muscles enable the main volume of the tongue to move in multiple directions within the oral and pharyngeal cavity, while its shape is determined largely by intrinsic muscles (Borden, Harris, & Raphael, 1994; Kent, 1992). Gestures involving tongue movements require more complex and graded motor control than those using only lip and jaw movements (Green, Moore, Higashikawa, & Steeve, 2000; Kent, 1992; MacNeilage & Davis, 1990; Nittrouer, 1993). In relation to language acquisition or speech motor development, the articulatory characteristics of liquids have never been studied in depth. This is partly due to technical difficulties in measurement and the cooperation of young participants, and also to the fact that the widespread use of articulation-based studies is relatively new to the field in general, even for adult speech. In acoustic analysis, it has been difficult to compare speech between adults and children, male and female speakers, or between younger and older speakers, because acoustic information such as fundamental frequency and formants are not easy to 3 normalize for direct comparisons. Further, although a spectrogram can provide articulatory information, it is not easy to see how articulators are involved in speech. On the other hand, in articulatory phonetics, articulatory gestures are used as a unit of measure of speech production across speakers (Browman and Goldstein, 1992, 1995; Byrd, 1996; Gick, 2003; Krakow, 1999; Sproat & Fujimura, 1993). Using articulatory gestures as a measurement unit allows comparisons to be made between child and adult speech in a way that is more accessible than acoustic analysis, because the measurements are based on articulators that we all share. In this dissertation, 1 adopt an articulatory phonetic approach and use articulatory gestures as a unit of measure. I measure the spatial and temporal organization of spoken English IM produced by English-speaking children who are between 3; 11 and 5;9, and who already have been producing IM with adult-like qualities. I am interested in this age group, because IM often already has emerged in speech but may not be developed fully. Investigating the articulatory properties of IM at this stage of children's speech will provide the phonetic details for better understanding the role of motor development in speech acquisition. In the following sections, I will present the theoretical background for the dissertation, the articulatory characteristics of gestures of IM in adult speech, the different processes of speech motor development, the empirical questions for this study, and the organizational plan for the dissertation's chapters. 1.2 Theoretical Background 4 My primary theoretical assumptions, based on articulatory studies of III in adult speech and speech motor development, are used to explain the articulatory characteristics of III in children's speech. 1.2.1 Articulatory Characteristics of English l\l Browman and Goldstein (1989, 1992, 1995) have proposed that articulatory gestures are the fundamental units of phonological structure in speech production. Gestures are abstract characterizations of vocal tract constriction, and spatial and temporal relations between gestures are represented as coordinative structures. In particular, gestural relation in timing and magnitude has been found to be consistent with phonetic correlates of allophonic variation (e.g., initial versus final) (Browman & Goldstein, 1995; Gick, 1999, 2003; Gick, Campbell, Oh, & Tamburi-Watt, in press; Kochetov, 2003; Krakow, 1989, 1999; Oh, 2002; Sproat & Fujimura, 1993; Tuller & Kelso, 1990). English III in adults has been relatively well studied using this articulatory approach. English III comprises two lingual gestures, a tongue tip raising gesture and a tongue dorsum backing gesture (Browman & Goldstein, 1995; Gick, 1999, 2003; Gick et al., in press; Sproat & Fujimura, 1993): the tongue tip moves to the alveolar ridge for tongue tip closure; the tongue dorsum moves back, sometimes up, toward the rear wall of the upper pharynx or the uvular (Narayanan, Alwan, & Haker, 1997)', accompanied by a lateral dipping of the tongue. These lingual gestures are characteristic of III. The existence of both consonant-like (tongue ' The lingual gestures of III also include a lateral dipping of the tongue (pc with Bernhardt 2004). However, in the articulatory phonetics the tongue tip and tongue dorsum gestures have been mainly studied and compared in terms of timing and magnitude. 5 tip) and vowel-like (tongue dorsum) gestures has been interpreted as an articulatory indication of IM as a sonorant consonant (Walsh-Dickey, 1997). English IM also shows syllable-based allophonic variation: a light IM [1] appears in the syllable-initial position, as in leap, and a dark IM [f] in the syllable-final position, as in peel. Defining IM with respect to its articulatory gestures provides insight into how its light and dark allophones are distinguished from each other. In particular, the gestural characteristics of the allophones of English IM are distinct in terms of spatial and temporal organization. In terms of spatial organization, the tongue tip (TT) gesture is more tightly constricted in the syllable-initial allophone than in the syllable-final allophone, whereas as the tongue dorsum (TD) gesture produces a tighter constriction in the syllable-final allophone than in the syllable-initial allophone. The tongue body (TB) is lowered more in the syllable-final allophone than in the syllable-initial allophone (Browman & Goldstein, 1995; Gick, 1999; Sproat & Fujimura, 1993). Using high-speed lateral cinefluorography, Giles and Moll (1975) also have provided information on the articulatory differences between the initial and final allophones of IM in American English: the TD was found to retract more in the syllable-final than syllable-initial productions. While tongue tip closure was relatively consistent in the initial position, a lack of complete closure of the tongue tip was manifested in final position. Giles and Moll also proposed that a final IM is more variable with respect to the constriction location. Further, in casual and fast speech, the tongue tip-palate contact was missing for the final IM or for consonant clusters (e.g., as in "pledge" and "help") in some English dialects. More recent articulatory studies have identified further differences between syllable-initial and syllable-final IM. TT raising and TD backing gestures occur almost simultaneously 6 in the syllable-initial position, whereas the TD gesture occurs earlier than the TT gesture in the syllable-final position (Browman & Goldstein, 1995; Gick, 1999; Gick, 2003; Sproat & Fujimura, 1993). That is, the timing of the two gestures is near synchronous in the syllable-initial position, although TT closure often reaches its maximum constriction slightly before the TD backing gesture reaches its closure and constriction. In contrast, in syllable-final position, the TD backing reaches its maximum constriction well before the TT closure reaches its maximum; hence a timing "lag" occurs (Sproat & Fujimura, 1993). To explain this pattern, Sproat and Fujimura have claimed that a dorsal gesture always is closer to the nucleus of a syllable (i.e., the vowel), because it is more vocalic. In other words, the posterior gesture of l\l (TD backing) may be described as being attracted to the nucleus of a syllable, while the anterior gesture (TT raising) appears at syllable peripheries. This also is related to the speed of movement of gestures, and it has been reported that the posterior gesture (TD) is slower than the anterior gesture (TT). Since the TD gesture is slower it starts earlier than its TT counterpart post-vocalically. In articulatory studies, the timing between gestures is usually measured at their peak, not at the initiation of a movement. That is, the measurement is not taken at the onset but at maximum constriction. 7 Initial Final TB TT •narrow narrow clo clo Time Figure 1-1. Gestural Timing of English In Browman & Goldstein's (1989) terms, the TB constriction is characterized as "narrow" and the TT constriction is characterized as "closure" for III. The constriction of TB and TT occurs almost simultaneously syllable-initially, whereas TB (and TD) lowering (and retraction) occurs earlier than TT closure syllable-finally. Articulatory gestures also have been used to find correlates of ambiguously syllabified allophones (Gick, 1999; Gick, 2003; Kochetov, 2003; Krakow, 1999). While the articulatory properties of syllable-initial / syllable-final allophones of IV have been relatively well-studied (e.g., Browman & Goldstein, 1995; Sproat & Fujimura, 1993), the properties of intervocalic allophones have not been. Ambisyllabicity, a situation in which a single segment is considered attached to both a preceding and a following syllable, is interesting, because the ambiguously syllabified allophone is typically in word-final position (VC # V), yet exhibits some properties of an initial allophone. The goal of the study of ambisyllabicity is to determine whether, and how, a segment behaves as a final or an initial allophone (resyllabification). For instance, the magnitude and timing of ambisyllabic allophones of III have been studied (Gick, 2003). In data collected using an E M M A (electromagnetic midsagittal articulometer), Gick compared syllable-initial (a # la), syllable-final (al # h), and ambisyllabic (al # a) allophones and predicted an intermediate magnitude for the latter. The 8 results showed that while the reduction of the magnitude of TT in the final allophone was significant when compared to that of the initial allophone, no significant difference was observed in the magnitude of TT between the initial and ambisyllabic allophones, or between the final and ambisyllabic allophones. The TD backing gesture was present in all allophones with no significant difference in magnitude. The timing between TT and TD gestures of the ambisyllabic allophone patterned similarly to that of the final allophone. The studies discussed in § 1.2.1 are based on an articulatory model of speech. The spatial and temporal characteristics of the articulatory gestures of allophones of English IV in adult speech are summarized in tables 1-1 and 1-2. Table 1-1. Summary Chart: Spatial Characteristics of English IM TT TD Initial tighter, consistent closure less constricted Final reduction, more variable greater backing Ambisyllabic intermediate same as final Table 1-2. Summary Chart: Temporal Characteristics of English IM TT-TD Relative Timing Initial no lag or short negative lag TT and TD move simultaneously or TT slightly precedes TD Final positive lag TD precedes TT Ambisyllabic intermediate positive lag TD precedes TT Note: These are syl able-based allophones, not in isolation. 9 1.2.2 Speech Motor Development Speech development includes physiological changes as well as perceptual and cognitive change. In particular, a number of researchers have proposed that the phonological development of speech can be explained by biological changes of the motor processes (Green, Moore, Higashikawa, & Steeve, 2000; Kent, 1983, 1992; Locke, 1983; MacNeilage & Davis, 1990; Nittrouer, 1993). For instance, bilabials (e.g., Im/, Ibl) emerge earlier than coronals (e.g., /n/, /d/), because motor control for orofacial movements is acquired earlier than for tongue movements, since the tongue is more difficult to control (Green et al., 2000; Kent, 1992; MacNeilage & Davis, 1990; Nittrouer, 1993). As for the tongue, vowels that require the dorsal gesture (e.g., /a/, lol) are acquired earlier than consonants that involve the tongue tip gesture (e.g., Itl, IM) (Boysson-Bardies, 2001). Further, neuromotor studies (Smith, 1978, 1994; Smith, Goffman, & Stark, 1995; Thelen & Smith, 1994) have suggested that speech production improves as the neuromotor system matures. Thus, it is likely that speech development is closely related to the speech motor control that allows young children to produce certain sounds. To produce target sounds, children must learn how to control articulatory movements specific to speech. Recently, Green et al. (2000) linked speech motor development to the sequence of phonological acquisition based on distinct motor coordination phases: differentiation, integration, and refinement (see § 2.3.1.3 for more detailed discussion of Green et al.'s (2000) study). These distinct changes of motor coordination have been used to characterize motor development, both in general (Lazarus & Tudor, 1987; Provins, 1997; Schuster & Ashburn, 1992; Thelen, Ulrich, & Jensen, 1989; Trevarthen, 1984; etc.) and with respect to 10 speech (Green et al., 2000; Kent, 1983, 1992; Locke, 1983; MacNeilage & Davis, 1990; Piske, 1997; Tobin, 1997; etc.). To begin, motor development involves Differentiation, whereby "a pre-existing behaviour becomes more independent and specific" (Green et al., 2000). For instance, in the early motor development of grasping, infants first move the arm as a whole. Gradually, they gain independent movement of the arm, hand, and fingers (Schuster & Ashburn, 1992; Trevarthen, 1984). Limited independence and the inability to associate movements of distinct segments are common in immature motor systems (Provins, 1997). This lack of independent movement decreases as motor systems mature (Lazarus & Todor, 1987). Speech motor development may involve a similar progress, since speech motor skills require independent control of different parts of the vocal tract, and speech motor systems become more specific with maturation (Green et al., 2000; Kent, 1992). Limited by an immature speech motor system, young children may have coordination difficulties (Kelso, 1977). Thus, considering the lack of "differentiation," early speech motor development can be characterized as having: (a) gross and holistic movements, missing the details (Kent, 1992; Studdert-Kennedy, 1991) and (b) high incidences of the coupling of independent gestures (Kelso, 1977; Smith, 1978). In addition, motor development involves Integration, whereby a new behaviour is added to existing ones. Motor control may not develop uniformly across motor systems, and the movement of a certain component may be dominant in early motor development (Green et al., 2000). For instance, in the development of posture, control of the head and neck come first, and then the trunk and limbs (Schuster & Ashburn, 1992). 11 Again, speech motor development may exhibit a similar tendency. In early language development, young children do not like sounds or gestures that require motor skills beyond their ability (Locke, 1983). For instance, jaw movement is predominant in early speech development of ages one to two years (MacNeilage & Davis, 1990), while movements with other articulators tend to emerge later and be integrated with the dominant gesture (Green et al., 2000): in particular, independent lip movement emerges around age two (Green et al., 2000) and independent tongue movement emerges much later around the age of three (Kent, 1992), although they never reach full independence during these years (Folkins & Abbs, 1975). Thus, the motor developmental process may include both "differentiation" and "integration," or just one of them. Further, motor development also may involve Refinement, whereby motor skills develop "sequentially and orderly" (Herring, 1985) and undergo "constant improvement" (Green et al., 2000). For example, although chewing and speech require oromotor (lip and jaw) control, the infrastructure for chewing is established by the end of the first year of life (Green et al., 1997; Moore & Ruark, 1996), whereas coordination skills for speech are not mastered until the early school years (Sanders, 1972). The requirements for speech production are "time-specified" (Gracco, 1994) according to "refinement." Thus, i f speech motor development is viewed as "refinement," there would not be a dramatic shift in the role of each articulator, and spatial and temporal coordination would gradually improve with age (Green et al., 2000). Due to the lack of accessible methods for obtaining articulatory measurements in young children, many questions related to speech motor development have not been fully 12 answered. Nevertheless, on the basis of the findings from previous studies reviewed in this section, early speech motor development can be summarized as follows: Table 1-3. Summary Chart: Three Motor Development Processes Differentiation Integration Refinement a. gross movements b. missing details c. coupling a. preference of a certain movement to others b. adaptation of new movements c. modification a. more detailed, specific movements b. improvement in coordination c. gradual improvement Based on our knowledge of gestures of III in adult speech and early speech motor behaviours in children's speech, a number of empirical questions can be raised with respect to the acquisition of the articulatory gestures of English III. The next section will lay out the questions which this dissertation aims to investigate. 1.3 Empirical Questions In the previous sections, we have seen that adult English IV comprises two lingual gestures — a tongue tip raising and a tongue dorsum backing (with a lateral dipping of the side[s] of the tongue) -— and that the gestural organization for IM allophones shows distinct 13 patterns with respect to timing and magnitude. We also have seen distinct processes of motor coordination (differentiation, integration, refinement) in speech motor development. Based on previous literature in articulatory studies of IM in adult English (e.g., Browman & Goldstein, 1992, 1995; Gick, 1999; 2003, Gick et al., in press; Sproat & Fujimura, 1993) and motor development (e.g., Green et al, 2000; Kelso, 1977; Kent, 1992; MacNeilage & Davis, 1990), we may expect the articulation of IM to be difficult for young children for two reasons: (a) it has multiple lingual gestures, which require more degrees of freedom of tongue movement, and (b) depending on syllable position, the articulatory components must be timed differently. Thus, English IM provides two independent dimensions of complexity, which offer an excellent context for testing the theories of speech acquisition in relation to speech motor development and syllable-based articulatory study. Thus, this dissertation is interested in answering the following empirical questions: 1) How do children produce IM? Do they coordinate the articulatory gestures of IM like adults, or do they do it differently? 2) How do(es) motor process(es) map onto the speech acquisition of IM? Which motor process(es) is/are involved in the process of learning the articulation of IM? To produce the articulation of IM, children must implement the two lingual gestures and coordinate them appropriately in timing and magnitude, according to different syllable positions. The tongue is a complex muscular hydrostat with many degrees of freedom (Kent, 1992; MacNeilage & Davis, 1990). To successfully perform the complex articulatory gestures of IM, we may assume that a child uses the least complex manipulation of the tongue, a manipulation that maximally reduces the degrees of freedom, i.e., movement of the entire volume as a uniform whole or modification of existing movements. A fundamental question 14 concerns how the acquisition of the articulation of IM is accounted for with respect to motor development. This issue will be investigated based on the spatial and temporal characteristics of the articulatory gestures of IM in children's speech production at the ages 3;11 to 5;9. More detailed hypotheses and predictions are laid out in chapter 2 (§ 2.4). An articulatory gestural approach provides a useful framework for representing articulatory behavior in children's language. It can facilitate the study of speech acquisition both theoretically and methodologically, since both child and adult utterances can be described and compared in terms of the same basic units of gestures. This dissertation is in the pursuit of the articulation-based approach. 1.4 Organization of Dissertation Chapter 2 presents an overview of previous articulatory and speech development studies and an hypotheses. The articulatory studies show how phonological structure may be represented by the organization of articulatory gestures with respect to syllable position, ambisyllabicity, coarticulation, and cross-linguistic timing patterns. The review of developmental research includes physiological development of the vocal tract and phonological development in speech production and perception. Considering the articulatory gestural properties of IM with respect to developmental patterns in speech motor theory, hypotheses and predictions are suggested. Chapter 3 introduces ultrasound imaging techniques relevant to the present study, and the experimental methods. 15 Chapters 4 and 5 present experiments of IM in child speech. In chapter 4, the spatial characteristics of IM — the number of gestures, constriction location, and allophonic variation — will be tested to determine whether distinct gestural organization is manifested in an adult-like way in different syllable positions. Chapter 5 focuses on the temporal characteristics of IM gestures. The gestural timing of initial, final, and intervocalic allophones of IM will be examined to determine whether timing is as significant in distinguishing allophonic variation in young children as it is in adults. Chapter 6 provides a post-experiment study of perception judgment by adults, and an acoustic analysis of sample tokens from the collected data. This will allow us to compare the articulatory characteristics of children's IM with acoustic and perceptual information. Finally, chapter 7 summarizes the results and discusses their phonological and developmental implications for speech acquisition. 16 CHAPTER 2 STUDIES IN ARTICULATORY PHONETICS AND SPEECH DEVELOPMENT 2.1 Introduction This chapter introduces the literature that provides the empirical and theoretical background for this dissertation. Beginning with articulation-based studies, § 2.2 presents a theory of articulatory phonology, articulatory phonetic studies in allophonic variation and ambisyllabicity, as well as cross-linguistic studies of liquids. § 2.3 introduces studies in speech development, including the physiological development of the vocal tract and speech motor control, and the phonological development of IM in speech production and perception. The hypotheses tested by this dissertation are presented in § 2.4. 2.2 Articulation-based Studies Traditionally, speech has been studied from two perspectives: phonological and phonetic. Phonetics is concerned with actual sounds we make in speech. In phonetic structure, the description of speech is based on physical properties such as articulatory, aerodynamic, acoustic, and auditory information in real time (Gafos, 1999; Ladefoged, 1993). Phonology is concerned with the way these sounds are organized into a salient language. In phonological structure, speech is described as sequences of abstract symbols belonging to the phonemic or 17 featural inventory. Phonetics and phonology may not have one-to-one mapping relations, because the sounds that occur in human languages, and people's actual production and perception of those sounds, may not all be represented in phonological rules. Recently, a growing interest in the interface between phonology and phonetics has led researchers to recognize that phonetics is a ground for phonological patterns, and a source for understanding phonological rules (e.g., Browman & Goldstein, 1995; Keating, 1988; Kelso, Saltzman, & Tuller, 1986; Krakow, 1989; Zsiga, 1995). In this regard, as a way of finding phonological correlates in speech, experimental phonology has been concerned with speech processes such as the measurement of articulatory movements, aerodynamics, acoustic analyses, and speech perception. For example, Articulatory Phonology (Browman & Goldstein, 1989, 1992, 1995) has provided a base for articulatory studies to find correlates between articulatory gestures and phonological structure, especially in terms of syllable position. The remainder of this section introduces articulatory studies developed by Articulatory Phonology. 2.2.1 Articulatory Phonology A gestural approach (Browman & Goldstein, 1995; Byrd, 1996a, 1996b; Gick, 1999; Gick, 2003; Kochetov, 2003; Krakow, 1989; Sproat & Fujimura, 1993; Tuller & Kelso, 1990) has been introduced as an alternative to a segmental representation of phonology. In ai attempt to integrate the phonetic and phonological structures of speech, Browman and Goldstein (1989, 1992, 1995) have proposed Articulatory Phonology. In this approach, 18 articulatory gestures are the fundamental units of phonology. The main assumptions in Browman and Goldstein (1989, 1992) are summarized below: 1. Gestures are primitive phonological units, which do not correspond to either features or segments. They are the basic units of phonological contrast, and represent an abstract characterization of coordinated movements of articulators within the vocal tract. 2. The gestures of a given utterance are organized into a larger structure or constellation which can be represented as a gestural score. The score specifies a set of parameters for each gesture: each gesture is specified in terms of a vocal tract variable; each variable is associated with a set of articulators; movements of specific sets of articulators determine the parametric value of that variable. For example, oral gestures involve tract variables that define the constriction degree — lip aperture (LA), tongue tip constriction degree (TTCD), and tongue body constriction degree (TBCD) — and constriction location — lip protrusion (LP), tongue tip constriction location (TTCL), and tongue body constriction location (TBCL). Velic and glottal gestures involve a single tract variable of aperture size — velum aperture (VEL) and glottal aperture (GLO). The lip gesture includes LP and L A ; the tongue tip gesture contains TTCL and TTCD; and the tongue body gesture contains T B C L and TBCD. Temporal relationships between gestures are represented in a gestural score, and hence, constellations of gestural units potentially overlap in time and in space. 3. Patterns of overlapping organization can be used to specify important aspects of the phonological structure of particular languages and account for a variety of different types of phonological variation. Such variation includes allophonic variation and speech alternations, as well as coarticulation and speech errors.1 ' Further, connected speech such as assimilation, deletion, and reduction can be explained by an increase in gestural overlap and a decrease in gestural amplitude (Farnetani, 1999). In casual fast speech, 19 Browman and Goldstein present a number of examples of gestural analyses of coarticulation of consonants and vowels, nasals, and lateral IM. Using these examples, and without positing implementation rules for converting phonological units into variable physical parameters, they show how gestures give rise to context-dependent articulatory and acoustic trajectories. That is, Articulatory Phonology unifies the seemingly unrelated speech processes such as coarticulation, allophonic variations, and alternations, which, in Feature Geometry, might require a number of separate phonological rules. One example is coarticulation of "perfect memory." In Browman and Goldstein (1990b), two productions of the sequence "perfect # memory" were analyzed in isolation, and in continuous speech. While [t] of "perfect" was not audible in continuous speech, X-ray trajectories revealed the presence of the alveolar closure gesture of III, and that in some tokens, its magnitude was the same as [t] in isolation. The difference between the two productions was that in continuous speech the alveolar closure gesture of [t] was completely overlapped by the following [m] in "memory." This example is an indication of significant information on articulatory gestures: although hidden gestures may be reduced in magnitude and not be audible in fast speech, the gesture itself is present and remains unchanged (Browman & Goldstein 1990, p. 336). The following sections discuss how Articulatory Phonology is applied in articulatory studies in phonetics. Studies included in these sections are concerned mainly with how articulatory gestures behave according to syllable position. consonantal gestures can be overlapped or blended. For instance, Hardcastle (1985) found in his EPG study that both speech rate and prosodic boundaries influence gestural overlap: (a) the amount of overlap was greater at the fast speech rate; (b) the boundary effect was largest at the clausal and sentential boundaries; and (c) while slow rate showed long separation intervals between gestures, the fast rate showed consistent overlap. 20 2 . 2 . 2 Allophonic Variation Many studies have shown that gestural behaviors vary according to syllable position. For instance, syllable-initial and syllable-final allophones are associated with different gestural patterns (Brcwman & Goldstein, 1995; Byrd, 1996; Gick, 1999, 2003; Kochetov, 2003; Krakow, 1999): in general, syllable-initial allophones show tighter constriction and more stability than syllable-final allophones; the timing relation between gestures also shows different coordinative patterns in different syllable positions. The following reviews some studies of gestural patterns of syllable-based allophones of nasals and stops (see § 1.2.1 for laterals). 2 . 2 . 2 . 1 Nasals In English, nasals have two gestures: oral constriction and velic lowering. Krakow (1989) has demonstrated a difference in coordination between syllable-initial and syllable-final nasals by manipulating the location of word boundaries across matched phonetic sequences involving initial and final /m/, and using stimuli such as see # more versus seem # ore, and pa # made versus palm # aid. Comparison of initial and final allophones showed that for Iml, syllable position was more consistently distinguished in the velum than in labial articulation: syllable-final nasals produced (a) lower velic positions, (b) longer low velic plateaus, and (c) larger lowering and raising movements. In the syllable-initial position, the end of the velum lowering movement is roughly synchronous with the end of the lip-closing movement. However, in the syllable-final position, the velum lowering ends substantially 21 earlier (100-350 ms) than lip closure, leading to the fact, and perception, of vowel nasalization. Nasals have been considerably well-studied in phonetics, and the findings of this study of lip-velum coordination (Krakow, 1989) also correspond to traditional ear-based descriptions of nasals. In addition to the acoustic information of nasals, this articulatory study has provided a finding that distinct articulatory organization can correspond to syllable position. Initial Final V E L LIP wide wide clo clo Figure 2-1. Gestural Timing of Nasal /ml. This schematic diagram shows the gestural timing difference in initial and final allophones. In final Iml, the velum lowering occurs earlier than the lip closure. Evidence that the organization of articulatory gestures of sounds is sensitive to phonological structure, especially to syllable position, has been found consistently in articulatory studies of sounds other than nasals. 2.2.2.2 Stops Browman and Goldstein (1995) have measured the degree of lip constriction for syllable initial and final Ipl with respect to the vertical difference between upper and lower 22 lip positions. Although full lip closure was found in both syllable positions, tighter lip constrictions were observed in the syllable-initial position. Likewise, Keating (1995) showed that the alveolar stops III and /d/ showed greater peak constrictions in the initial position in both stressed and unstressed syllables as compared to final position. Fougeron and Keating (1997) provided further evidence that strengthened articulatory constrictions are manifested in the syllable-initial position. As for temporal coordination, Tuller and Kelso (1990) looked at the relative timing of labial and laryngeal movements for syllable-initial and syllable-final /p/, and found that lip and larynx coordination for the syllable-initial position was more stable than for the syllable-final position. Recent work by Kochetov (2003) also supports gestural organization as a function of syllable position. In his E M M A (Electromagnetic Midsagittal Articulography) study of the Russian labial stop /pV, Kochetov found tighter and more stable articulatory constriction of the lip aperture (LA) and tongue body (TB) fronting and raising gestures for the initial allophone. For the syllable-final allophone, a lower and less front TB, longer constriction duration, and distinct timing pattern were observed: while TB is timed with the release of L A syllable-initially, TB is achieved earlier than L A syllable-finally. Kochetov found that ambisyllabic allophones (VC # V) are less consistent across subjects and across measurements, in that they show a similarity with either initial or final allophones. In sum, the gestural organization for syllable-initial stops is less variable than for syllable-final stops, and intermediate for intervocalic ones. 2.2.3 Ambisyllabicity 23 In articulation-based studies, the properties of allophones are not specified directly, but are determined by changing the component gestures according to their adjacent gestures. Gick (2003) investigated the articulatory properties of English liquids (IM and lrf) and glides (/w/ and /j/) in various syllable positions. He compared ambiguously syllabified allophones (e.g., aw # a) to syllable initial (e.g., a # wa) and final (e.g., aw # ha) allophones and predicted that syllable-based allophones are distinguished in magnitude. That is, gestures involved in ambisyllabic allophones would display an intermediate magnitude between that of an initial and final allophone. The results showed that in the intervocalic allophones, a reduction occurred (compared to the initial allophones) in the magnitude of TT with a delay of TT . However, although the intervocalic allophones were sensitive to the context following the allophone, neither the reduction in magnitude nor the TT delay was significantly different from that found in initial allophones. Like the Russian ambisyllabic allophones of labial stops ( § 2.2.2.2), Korean liquids (Oh & Gick, 2002) also showed inconsistent patterns of ambisyllabic allophones (VC # h)2: they patterned with initial allophones or final allophones depending on individual differences and real versus novel word distinctions (see § 1.2.1 for lateral III). 2.2.4 Cross-linguistic Studies of Liquids , In Gick, Campbell, Oh, and Tamburri-Watt (in press), inter-articulator lingual timing between the component gestures of III and Ixl in six previously unstudied dialects and languages (Canadian English III, Quebecois French III, Serbo-Croatian l\l, Korean IM, Beijing 2 In Korean /h/ behaves more like a vowel and is only recognized by devoicing/aspiration of the preceding sound. 24 Mandarin hi, and Squamish Salish III) were measured using ultrasound imaging. Each language has at least one liquid in both the initial and final position, and the degree of lightness or darkness of a liquid varies. In addition, the place of the articulators varies across languages. Thus, it was important to test whether lightness degree is reflected in terms of gestural timing, or whether an effect of anteriority of a vocalic gesture effects timing. Results for Canadian English IM suggested a similar but not identical pattern to that previously reported for American English III (Browman & Goldstein, 1995, Gick, 2003; Sproat & Fujimura, 1993). As in previous studies, the TT raising and fronting gesture and a TD backing gesture were present in all allophones. While American English has a short negative lag (often nearly simultaneous) in the initial allophone, a greater negative lag (TT precedes TD) in initial position was observed in Canadian English. Both studies found a relatively long positive lag (TD precedes TT) in the final allophone. The general findings were: (a) English, French, Mandarin, and Squamish exhibited a positive lag (the posterior gesture is achieved earlier) in final position, as previously observed for American English; (b) French, Korean, and Mandarin showed no posterior gesture in the initial position; (c) Of the other three languages that retained a posterior gesture in the initial position, two (Serbo-Croatian and Squamish) showed a near-zero lag (i.e., simultaneity), similar to that observed in previous studies of American English; (d) (i) In all but one language (Serbo-Croatian), the intervocalic allophone was categorically different from the final allophone, (ii) Except for one speaker (Quebec French subject 2, whose pattern was generally anomalous), gestures were simultaneous in every case where two gestures were present in the intervocalic allophone. The results in Gick et al. (in press) support the claim 25 that syllable positions can be distinguished and characterized with respect to gestural organization (Krakow, 1999). An articulation-based model provides a useful framework for representing articulatory behavior in children's language, because child and adult utterances can be described using the same physical units called "gestures." Most developmental studies in speech production have analyzed children's speech in terms of segments or features, which often have been found to be insufficient to describe what we hear, and what children actually produce in terms of physical speech production. Some researchers have noted that in early development (especially for the first 50 words), children may not produce sounds based on segments (Peters, 1977), and children's speech could be more approachable if described in terms of "articulatory routines" (Menn, 1983). This arbitrary boundary for phonemes is one problematic aspect of using segmentation to understand children's speech, because what they produce may not be transcribed adequately as sounds or phonemes to which we are familiar. An acoustic analysis such as a spectrogram also provides limited information with respect to children's speech, since it provides too much information that cannot be interpreted. It would be difficult to detect the onset of a following sound from the offset of a preceding sound (i.e., a vowel followed by a nasal or glide), because an overlapping section occurs in a spectrogram. If children's spectrograms do not resemble what we would expect from adult speakers, it would be difficult to describe whether they produce a correct sound or a different sound. In contrast, the articulation-based model is founded on the physical alignment of gestures, and some gestures can be longer/shorter than others. How gestures are associated with one another in time during the course of speech production is a focus of the articulatory 26 model. The articulation-based model allows us to separate more easily the complex articulatory properties in children's speech, because basic units are gestures, not segments, and the speech of children can be described and compared with adults' speech by using the same gestures as basic analyzing units. This would be particularly useful in study of IM, since it comprises multiple gestures, and the gestures are organized differently in magnitude and timing in different syllable positions. 2.3 Studies in Speech Development This section presents studies in speech development: how the physiological properties of the vocal tract are associated with speech production, vocal tract development, vocalizations, and motor development with respect to age effect; and studies related to liquids in children's speech production and perception. 2.3.1 Physiological Development 2.3.1.1 Vocal Tract The infant's vocal tract is not a miniature version of the adult's. In fact, in some ways it resembles that of the nonhuman primate more than that of the human adult (Kent, 1983). Characteristics of the infant's vocal tract include: (a) a high placement of the larynx and a much shorter vocal tract; (b) a relatively short pharyngeal cavity, giving little room for the posterior portion of the tongue to move; (c) a large tongue in relation to the size of the oral cavity, leaving little room for distinct vertical movements of the tongue tip or blade; and (d) a 2 7 gradual bend, not a right-angle curve, in the oropharyngeal channel, and a closely placed soft palate and epiglottis, forcing infants to breathe either through the nose or with a wide mouth opening (Boysson-Bardies, 2001; Kent, 1983; Stark, 1980). Due to the configuration of their vocal tracts, infants have restricted movements that prevent them from articulating sounds in an adult-like way. At three months, the palate is lowered and moves forward. The tongue becomes lengthened, and its muscles become more developed; the opening of the pharynx allows it to move from front to back. At five months, babies are capable of breathing and using their larynx (roughly) as adults do (Boysson-Bardies, 2001). Once the facial skeleton has grown downward and forward, the size of the oral cavity increases relative to the tongue, and the child's ability to produce a diversity of vowel types increases within the first six months (Netsell, 1981, cited in Vihman, 1996). Although the child's vocal tract begins to resemble an adult's between 6 to 12 months, its transformation is not complete until age five to six, at which point all the articulators can be controlled (Kent, 1976). The maturation of the vocal tract begins with the most central organs and then extends to the peripheral ones, following the common pattern of motor development where gross movements are usually mastered before specific ones. For instance, control of the tongue tip, and the lips is mastered shortly before five or six (Boysson-Bardies, 2001).3 Thus, the remodeling of the vocal tract affects young children's ability to produce certain sounds. 3 Although bilabials emerge early in production, the mastery of lip-jaw movement, upper and lower-lip coordination may take longer. See Green, Moore, Higashikawa, and Steeve (2000) for orofacial motor development of bilabials. 28 2.3.1.2 Vocalizations Infants begin to develop their vocalizations with oral articulations — babbling — at around seven months, which coincides with the elevation of the mandible (Locke, 1993). Babbling is a unique character of human speech. Infants are aware of auditory-motor correspondences well before the babbling stage (Kuhl & Meltzoff, 1982), and the upgraded vocal system enables them to imitate the speech articulations they see and hear. Traditionally, babbling was considered as prelinguistic sounds (vocalizations prior to the first words) not related to phonological development (Jakobson, 1968). In Jakobson (1968), a transition from babbling to speech was seen as a developmental discontinuity. Contrary to the discontinuity hypothesis, Oiler (1978) has shown that a close resemblance exists between the sounds and syllable structures of the canonical babbling period and those of the first words. This continuity hypothesis has been supported by many studies (Locke, 1983; Menn & Mattel, 1992; Menyuk, Liebergott, & Schultz, 1986; Piske, 1997; Stoel-Gammon, 1985; Vihman, 1986, etc.) which have shown positive correlations between babbling and later speech production. Due to predispositions in vocal development, it has been suggested that infants prefer certain articulatory dynamics to others and are functionally incapable of producing later-developing sounds (Locke, 1983; Piske, 1997). The transition from prelinguistic vocalizations to adult-like speech implies that children have mastered the coordination of complex articulatory systems, a task which will have been completed by age five to six. 29 2.3.2 Speech Motor Development In general, motor development shows a trend in which large movements are acquired long before fine movements (e.g., arm vs. finger) and central movements are acquired long before peripheral ones (e.g., torso vs. elbow) (Kalverboer, 1993; Kelso 1977; Thelen, Ulrich, & Jensen, 1989). The maturation of the vocal tract also begins with the most central organs and then extends to the peripheral ones (Kent, 1992), and gross or "ballistic" movements are usually mastered first and then extended to specific movements (Kent, 1992; Locke, 1983). In an attempt to find a correlation between motor development and speech acquisition, some research has been done on early speech motor development. However, since the tongue cannot be measured using external techniques such as video, most studies are limited to oromotor coordination such as upper and lower lip movements, jaw movements, lip-jaw coordination, and upper-lower lip coordination (Green et al., 1997, 2000; Kent, 1992; MacNeilage & Davis, 1990; Nittrouer, 1993; Oiler, 1978, 1980; etc.). In one study of lip and jaw coordination in bilabials, Green, Moore, Higashikawa, and Steeve (2000) compared the speech of young children aged 1, 2, and 6 to that of adults. They observed age-related changes: frequent jaw movements and large displacements were apparent in one-year-old children, whereas jaw displacements decreased as upper and lower lip displacements increased in two-year-old children. Lower lip movements increased significantly between ages two and six; the contribution of lower lip and jaw movements to oral closure was similar to that of adults in six-year-old children, although upper lip movements were less refined. That is, the coordinative organization of gestures changes 30 dramatically during the first two years (differentiation and integration) and continues to improve even after the age of six (refinement) (See § 1.2.2). Green et al. (2000) suggest that physiological coordinative constraints may limit sound-producing capability during the early years, and that the specific changes in articulatory coordination over time can predict the sequence of phonological development. For example, while the bilabial stop Ibl can be produced using a rapid jaw movement, labiodental 111 requires more graded and independent movements of the lower lip and jaw control to produce a constriction (MacNeilage & Davis, 1990). In general, sounds with complete oral closure require less control than sounds without complete oral closure. This tendency reflects the universal sequence of phonemic acquisition (Stoel-Gammon, 1985). The findings in Green et al. (2000) support previous claims and assumptions: (a) that the jaw is the predominant articulator in early speech production, and its contribution to oral closure is greater than that of the lips (MacNeilage & Davis, 1990); (b) that during the first one to two years, a dramatic shift appears in the role of each articulator in oral closure, after which neuromotor control gradually improves, with spatial and temporal coordination maturing sometime after age six (Smith, 1978, 1994; Smith et al., 1995); (c) that as motor skills develop, children are able to separate movements independently as adults do, and adult-like articulatory fluency begins during the early school years (Kent, 1983); (d) that children have a preference for sounds with certain gestures due to coordinative constraints: observations of poor lip-lip and lip-jaw coupling in early speech production explain why bilabial stops are more predominant than labiodental fricatives4 (Bernhardt & Stemberger, 1998; Locke, 1983; Piske, 1997; Tobin, 1997); and (e) that later-developing sounds such as 4 Bernhardt & Stemberger (1998) found that children often first produce /f/ as a bilabial fricative rather than a bilabiodental fricative, an articulation which requires less differentiation of gestures. 31 liquids, fricatives, and affricates require more exertion of graded muscle force and sustained effort over time (Kent, 1992; Tobin, 1997). On the other hand, some studies have argued that speech motor development is not effected by age. For example, Sharkey and Falkins (1985) studied the variance of lip and jaw movements in the duration and displacement of two syllables in children aged 4, 7, and 10. No age differences in duration and displacement among these three groups were found, although the results were more variable than adults on all measures. Smith and McLean-Muse's (1986) study of upper lip, lower lip, and jaw movement in three groups of children (ages 5, 8, and 11) and adults also showed that mean displacement and peak velocity of the individual articulators did not dramatically increase with age. While lower lip displacement of all three groups of children was greater than that of adults, the oldest group showed the highest lower lip displacement. In addition, labial consonant productions by the youngest group were not significantly different from those of adults and older groups. Further, mixed results regarding age differences were found in Smith and Goffman (1996). To test whether speech motor performance improves as children mature, they examined bilabial productions of two groups of children (ages 4 and 7) as well as young adults. While age had a significant effect on duration and variability, no age difference was found in displacement and peak velocity. Furthermore, relative timing showed no significant difference between younger children and adults, although some difference was found between older children and adults. On the basis of these findings, Smith and Goffman suggest that speech motor performance be considered as nonlinear maturation that does not necessarily develop in parallel with age. 32 Moreover, in an acoustic analysis, Nittrouer (1993) found different degrees of co-production in disyllables in the production of children ages 3, 5, and 7 and that of adults (e.g., a # Ca [a bag], o # C i [a tea], and 9 # Cu [a two]). Articulatory gestures involving tongue movement (e.g., IXvJ and /ti/) showed more marked vowel effects in children than in adults, whereas those involving jaw movements only (e.g., /ta/ and /da/) showed little age effect. These results reflect the position of MacNeilage and Davis (1990) that children acquire adult-like skill for jaw movements sooner than they do for tongue movements (p. 970). Nittrouer concluded that children produced gestures similar in shape to adults, but movements were produced more slowly with greater temporal variation. She noted that although children were able to produce roughly the required adult gestures by ages three to four, maturing patterns varied across articulators (e.g., jaw vs. tongue). Due to the lack of accessible methods for obtaining direct articulatory measurements in young children, many questions related to speech motor development have not been fully answered. Nevertheless, on the basis of the findings from previous studies reviewed in this section as well as in chapter 1, early speech motor development can be summarized as in table 2-1: 33 Table 2-1. Summary Chart: Characteristics of Motor Development Preference . Children use some movements more than others due to motor constraints (Locke, 1983; Piske, 1997; Tobin, 1997). Modification . Children may use different articulatory movements or adopt new behaviour, even if the target sound is similar to what adults produce (Kent, 1992; Nittrouer, 1993). Gross to fine movement . Young children start with holistic and ballistic movements (Kent, 1992; Studdert-Kennedy, 1991) then gradually extend to specific movements (Kelso, 1977; Kelso et al., 1986). . Jaw movement is acquired sooner than lip and tongue movements (Green et al., 2000; Kent, 1992; MacNeilage & Davis, 1990). Time-specified maturation . Temporal coordination improves from rapid and holistic movements to appropriate timing sequences (Green et al., 2000; Kent, 1992). . High incidence of the coupling of independent movements decreases as speech motor control develops (Smith, 1978; Smith et al., 1995). . The requirements for speech coordination are highly time-specified (Gracco, 1994). Non-linear maturation . Speech motor performance does not necessarily develop in parallel with age, or has no age effect (Sharkey & Falkins, 1985; Smith & Goffman, 1996; Smith & McLean Muse, 1986). . Maturing patterns vary across articulators (Nittrouer, 1993). Variability . Children produce articulatory gestures similar in shape to those of adults, but more slowly and with greater variability (Nittrouer, 1993). 34 2.3.3 PhonologicaS Development 2.3.3.1 IM in Speech Production Laterals and nasals, both of which are sonorants, are very common sounds across languages. However, they have different patterns in language development. Nasals, which appear from the beginning of vocalizations and babbling stages (0;9-l ;6), are among the most frequently occurring sounds (along with stops) throughout all developmental stages (0;9-4;6) in child speech production (Grunwell, 1982; Kent & Bauer, 1985; Locke, 1983; Oiler, 1980; Vihman, 1996). On the other hand, IM is rare in early speech production and emerges much later (around 3;6) (Grunwell, 1982). The age at which IM and Ixl are acquired, and the order in which they are acquired, vary from child to child. Nevertheless, general tendencies in phonemic acquisition in English have been acknowledged: (a) sounds that are dependent on the precise timing of glottal and supralaryngeal processes (e.g., aspirated or glottalised stops) and (b) sounds that require more precise positioning (e.g., fricatives, liquids) are mastered late (Menn & Stoel-Gammon, 1995). In fact, laterals (and r-sounds) and fricatives have a low incidence in child speech production, and some children do not master these sounds until they reach the age of six to nine (Bernhardt & Stemberger, 1998; Boysson-Bardies, Vihman, Roug-Hellichius, Durand, Landberg, & Arao, 1992; Locke, 1980, 1993). Bernhardt and Stemberger (1998) point out that IM has a unique (lateral) tongue-shape, as well as an unusual zero in the acoustic spectrum. They note that a child has no basis for generalization from sounds present in babbling or early words, and argue that this could reasonably delay acquisition. 35 Many studies in phonological development have focused on questions involving phonological representations (e.g., phonemic contrast or opposition, and consonant harmony) (Jakobson, 1968; Smith, 1973; Stampe, 1969), developmental stages and corresponding patterns (Ingram, 1974, 1986; Kent, 1983; Macken, 1980; Oiler, 1975, 1980; Vihman, 1993), and a constraint-based phonological system (Bernhardt & Stemberger, 1998; Stoel-Gammon, 1992, 1996). These studies have provided not only a general timetable for vocalizations and syllable and word formation, but also a basis for structural similarities and differences between adult and child phonology. For their first three to four years, most children go through stages where IV is frequently replaced by glides [w] or [j] or other consonants, or deleted (Bernhardt & Stemberger, 1998; Goad, 1996; Grunwell, 1982; Kent, 1992; Menn & Stoel-Gammon, 1995; Vihman, 1996, etc.). For instance, English children often substitute IV with glides, stops, or fricatives in onset (e.g., [wept] "left," [daft] "light," [gok] "lock," [baek] "black," [bwak] "block"); in coda IV is often deleted (e.g., [mik] "milk," [da] "doll") or replaced with vowels (e.g., [tiko] "tickle," [fau] "fall") (Goad, 1996; Kent, 1992, pc with Stemberger, 2004). The substitution and deletion of IV has been one of the main issues in child phonology. A series of studies on consonant harmony as a factor in IV substitution have been produced (see Bernhardt & Stemberger, 1998; Goad, 1996; Pater & Paradis, 1997; Stemberger, 1988; Stemberger & Stoel-Gammon, 1991; Stoel-Gammon, 1992, 1996). The substitution and deletion of IV is also found in the speech production of children with phonological impairment. Leonard (1995) reported that both normal and language-impaired groups of children speaking English, Italian, and Swedish substitute other consonants for III, although substituting categories may vary due to a language-specific phonemic inventory and 36 phonological details of their ambient languages. Similar tendencies observed from both normal and language-impaired children indicate that the IV sound is universally difficult. 2.3.3.2 l\l in Speech Perception In perceptual development, it has been found that from as early as one to two months, most of the basic auditory capacities that are needed to determine speech sounds are present (Eimas, Siqueland, Jusczyk, & Vigorito, 1971). A series of studies in infant speech perception has shown that young infants (younger than 6 to 8 months) can discriminate categorically native phonetic contrasts (e.g., /ba/ vs /da/, /ra/ vs. /la/) and phonetic contrasts between variants (Aslin, Pisoni, Hennesey, & Perey, 1981; Jusczyk, Friederici, Wessels, Svenkerud, & Jusczyk, 1993; Werker & Tees, 1984; Werker, Gilbert, Humphrey, & Tees, 1981, etc.). Nazzi, Bertoncini, and Mehler (1998) found that by 4 months, infants can distinguish their native language from a very similar yet unfamiliar language. However, 4-month-old infants who were raised where two languages are similar (e.g., Spanish-Catalan) failed to discriminate one language from the other (Bosch & Sebastian-Galles, 1997). Unlike infants, adults have more difficulty in discriminating a phoneme which is not used in their native language. For instance, the distinction between /ra/ and Hal was difficult for Japanese adult speakers (because Japanese does not have IV). Also, stops in Hindi and Czech (whose voice onset time are different from English) were difficult for English speakers to distinguish (see Strange & Jenkins, 1978; Trehub, 1976; Werker & Tees, 1984, Werker etal., 1981). 37 Understanding how children perceive sounds is an important step to understanding speech production, as speech perception and production closely interact with each other (Rizzolatti & Arbib, 1998). Many studies have been produced on infants and L2 learners' speech perception, yet few have been done with young children ages three and four, especially to look at IV. One of these is a study by Strange and Broen (1980), who conducted production and perception tests of word-initial Iwl, IV, and Irl by children aged 2;11 to 3;5. In production, some of the children had not yet mastered III and Irl, whereas, in perception, they recognized them with high accuracy (over 90 percent). Although a great deal of variability in performance among the subjects existed, all showed better than chance identification of all three contrasts, and identification accuracy was clearly related to the ability to produce Irl and III correctly. Strange and Broen concluded that both the production and perception of phoneme contrasts develop gradually, and the perception of a contrast normally precedes its production. To understand the interrelation between production and perception, Locke (1980) developed a speech production-perception task in which perception stimuli were chosen to reflect a production test. He tested 131 children ages 3 to 9 years, most of whom had speech substitution. The results showed that about one-third of the contrasts produced incorrectly were also misperceived; however, misperceived contrasts were not likely to be involved in all production errors. It is important to find out how they are related — i f production errors are based on perception errors. Likewise, it also is vital to distinguish production errors that are due to perceptual difficulties from those due to articulatory difficulty. According to Barton (1976), before the age of three, children can discriminate phonemic contrasts such as Irl versus IV and 38 /w/ versus Ixl with 80 percent accuracy, supporting the view that full perception is achieved very early for some children at least (Smith, 1973; Stampe, 1969; Stemberger, 1992). Nevertheless, in speech production, some phonemic contrasts (e.g., Iff vs. IQI, Ixl vs. /w/, IV vs. Ixl) remain difficult to discriminate even at the age of three or older. Due to a lack of attention to the perceptual contrast, this may result in production errors in the preschool period (Vihman, 1996). Although it has received less attention, Shvachkin (1973, cited in Vihman, 1996) noted that articulation evidently influences phonemic development. Having rejected the view that speakers only hear the sounds that they are able to pronounce, Shvachkin suggested that an intimate connection exists between articulation and auditory processing (see also Studdert-Kennedy, 1991). Using modified or reduced articulatory gestures, children's production may lead to misperceived internal representations of phonemes, which then in turn could affect their speech (Straight, 1980). While speech perception and speech production are closely related to each other, this dissertation will focus only on the phonetic aspect of children's speech production. The relation between speech production and perception will be considered in future research. Hypotheses and predictions of the current study are laid out in the following section. 2.4 Hypotheses This chapter has reviewed studies in articulation-based phonetics and speech motor development, and briefly has introduced studies in phonological development. The acquisition of IV raises fundamental questions that will interest investigators of both 39 articulatory studies and speech motor development. The question in both areas is the same: How do children coordinate the lingual gestures in their speech production of /l/? Concerning articulation-based phonetics, the speech production of IM is interesting, because two lingual gestures — different in timing and magnitude sensitive depending on the syllable position in which IM occurs — are coordinated. This can be challenging for children, because two different parts of the tongue need to be differently coordinated when IM appears syllable-initially and syllable-finally. First, the timing between the two gestures is sensitive to syllable position: they are likely to be synchronized in the initial allophone, but have a lag in the final allophone as the tongue dorsum moves earlier than the tongue tip. Second, the magnitude is also sensitive to syllable position: it is reduced in the final allophone as the tongue dorsum is stretched more and pulled back farther. Whether children use a gestural coordination for the articulation of IM similar to that observed in adults' speech production of IM will be tested. Concerning speech motor development, movement of the tongue associated with the articulation of IM is interesting, because the anterior and posterior of the tongue are associated with movement in opposite directions. The tongue tip is moving upward and forward, whereas the tongue dorsum is moving backward. In addition to a directional contrast, motor skills such as using the apex of the tongue tip while pressing down the side(s) of the tongue, and lowering the tongue body while retracting it also are required to speak IM appropriately. In speech motor development, gestures involving tongue movements are acquired later than lip and jaw movements, presumably because the tongue is a complex muscular hydrostat with many degrees of freedom (Kent, 1992; MacNeilage & Davis, 1990). Again, it can be 40 challenging for children whose motor skills are not yet as mature as adults, because they may have to modify tongue movements and thus reduce the tongue's degrees of freedom. In terms of speech motor development, this dissertation tests these two different dimensions of complexity in the articulation of IM. Most researchers of speech motor development have generally espoused a view of motor development processes as occurring serially: they have argued that these processes are associated with particular ages (or "phases" or "stages" of development) (Gracco, 1994; Green et al., 2000; Kelso, 1977; Kelso et al., 1986; Kent, 1992; Locke, 1993; MacNeilage & Davis, 2000; Nittrouer, 1993; Piske, 1997; Smith, 1978; Smith et al. 1995; Studdert-Kenndey, 1991, etc.). I will call this argument a "Serial Motor Development (SMD)" hypothesis. Much literature in speech motor development has put forward the SMD view that control of oral structures emerges sequentially (Kent, 1992; Green et al., 2000) and that some articulations or gestures are mastered earlier than others (Locke, 1993; MacNeilage & Davis, 2000). Green et al. (2000), in particular, have suggested that the motor development processes of "differentiation," "integration," and "refinement" occur as developmental stages. They have observed that: a lack of differentiation of individual movements (coupling) or simplified movements is characteristic of motor development at age 1 (supporting Kent, 1992; Locke, 1983); modification to ease the articulatory complexity and/or dominance of a certain gesture occurs at age 2 (supporting Kent, 1992; Locke, 1983; MacNeilage & Davis, 1990); and adult-like coordination with greater variation occurs at age 6 (supporting Nittrouer, 1993). (See also § 2.3.2.) In contrast to the SMD hypothesis, some researchers have argued that an age effect does not exist, or that a crossover between sequential processes in speech motor development 41 occurs (Nittrouer, 1993; Sharkey & Falkins, 1985; Smith & Goffman, 1996; Smith & McLean-Muse, 1986). I will call this argument a "Parallel Motor Development (PMD)" hypothesis. In terms of the PMD view, although control of oral structures is less stable than that of adults, the literature has shown no age effect in the displacement and duration of oral movements (lip and jaw) (Sharkey & Falkin, 1985), in the displacement and peak velocity of individual articulators (Smith & Goffman, 1996; Smith & McLean-Muse, 1986), or in jaw movement (Nittrouer, 1993). Mixed results regarding age also are found in the literature. While production of labials by the youngest group was as-good as those of adults or older groups (Smith & McLean-Muse, 1986), or had an age effect (Smith & Goffman, 1996), the highest lower lip displacement (Smith & McLean-Muse, 1986) and significant difference in relative timing (Smith & Goffman, 1996) were observed in the oldest group. Thus, these studies suggest that speech motor development may not coincide with age. Moreover, a tendency of crossover between stages also is found within the same age group. That is, a maturing of motor control skills may vary across individual articulators (e.g., jaw vs. lip vs. tongue) even at the same age (Nittrouer, 1993). In this dissertation, I will test the S M D and P M D hypotheses for the articulation of IV in the speech of children ages 4 to 5 (3;11-5;9) in relation to motor processes such as differentiation, integration, and refinement. In the SMD hypothesis, motor development is sequential and highly age-specific: according to Green et al. (2000), the differentiation stage occurs at age 1, the integration stage at age 2, and the refinement stage by age 6. However, very little list know about stages between age 2 and 6. Thus, we would like discover which 42 motor stage(s) is/are associated with children between age 2 and 6, and whether these findings will support the serial maturing process of the SMD hypothesis. On the other hand, the PMD hypothesis suggests that improvement of motor performance is not necessarily associated with age, hence maturation of motor development is non-linear. Previous studies (focusing on children aged 4 to 11) have shown no significant age effect among different age groups, or more variation in the oldest group. Although the current study does not include a wide range of age groups, the groups aged 4 to 5 are considered relatively young compared to the age groups studied in the literature. It would be interesting to test the PMD hypothesis on this age group to see whether this group mapped onto to the older or younger groups studied in the literature. In addition to age, this dissertation focuses on another interesting aspect of speech motor development (gestures). Much of the literature on speech motor development has investigated the oral articulators such as the jaw, and the upper and lower lips, but no study has been conducted on the tongue articulator. For instance, the literature concerned with developing the SMD and PMD hypotheses describe motor developmental patterns only in terms of oral gestures, not lingual gestures, suggesting that control of oral gestures is mastered earlier than that of lingual gestures. Therefore, a great opportunity exists to test these hypotheses to see if there is a difference between oral and lingual gestures in motor development. If the tongue gesture is acquired later than oral gestures, which motor stage(s) would be associated with the acquisition of l\l at ages 4 to 5? Would motor development of the tongue gesture patterns be similar to that of oral gestures? Would a crossover exist among motor processes (stages), implying a non-age-specific development? 43 Based on the three types of processes in motor development (Differentiation, Integration, and Refinement), I lay out specific predictions deriving from each of these processes in reference to English IV articulation at ages around 4 to 5. Differentiation: Young children find it difficult to differentiate independent movements (Green et al., 2000; Kelso, 1977; Kent, 1992). Independent manipulation of different functional parts of the tongue should be difficult for children at any stage in their motor development. If lingual differentiation is lacking in the spatial realm, it is predicted that the two gestures will be coupled, and that the tongue will move as a whole (Gick, 2002b). In the realm of temporal coordination, it also should be challenging to achieve the two gestures of IV. With a lack of differentiation, the performance of two gestures of different timing would be difficult, and simultaneity of these gestures is expected. Integration: Even though a child may be able to use multiple articulators or complex temporal coordination, they may simplify their task by using one gesture or pattern more dominantly than others (MacNeilage & Davis, 1990). Integration suggests using an appropriate distribution of the functional load of a complex goal across degrees of freedom. The current hypothesis predicts that, for example, in order to reduce the complexity of the IV articulation, children may use either the TT or TD gesture more dominantly, even across syllable positions (i.e., a final IV may be produced as light [1], an initial III as dark [1], or [w] for III using dorsum and lip gestures) despite the fact that no exemplary evidence exists for such a pattern in the speech of adults around them. Likewise, a familiar temporal pattern of coordination that is learned correctly for one allophone may be inappropriately extended to another allophone as a means of simplifying the task of producing III. 44 Refinement: At any stage in speech motor development, a child has achieved some degree of mastery over a large number of relevant tasks. Refining this mastery to a normal adult-like level continues throughout development. Thus, while a child sometimes may exhibit an adult-like pattern, this will be an unrefined version of that pattern produced with less accuracy and consistency than would be typical of adult speech. For example, although appropriate gestures may be used for IM in a particular context, a TD backing gesture may be smaller in relative magnitude than that of an adult, a TT gesture may occur without adequate tip closure, or quantitative variability may simply be greater than that of an adult. Similarly, a qualitatively context-appropriate pattern of temporal coordination may be used, although the magnitude of the timing difference may be too large or small, or the accuracy less consistent than that of adults. If the SMD is correct, either Refinement or Integration should occur in the speech development of IM at these ages groups; and if the P M D is correct, any process(es) can occur at these ages, because speech motor development is neither age-specific nor linear. Specific predictions regarding the aspects of spatial and temporal characteristics of IM are spelled out in more detail in chapters 4 and 5. The following chapter introduces the methodologies used for the experiments conducted for this dissertation. 45 CHAPTER 3 EXPERIMENT METHODS 3.1 Introduction This chapter lays out the methodologies adopted for this dissertation's experiment. The experiment explores the spatial and temporal characteristics of syllable-based allophones of English children's III by using the ultrasound facilities developed for speech research at the Interdisciplinary Speech Research Lab (ISRL) of the University of British Columbia (UBC). The ultrasound imaging technique is introduced as an experimental methodology for speech research in general, and the methods for the present study including participants, apparatus, stimuli, and procedure are described. 3.2 Ultrasound Techniques Ultrasound is an image produced by ultra-high frequency sound waves that pass, in this particular experiment, through the tissue of the tongue. When the sound reaches the air above the tongue, it stops and reflects back to the transducer, so the outline of the tongue is reconstructed from the reflected waves and recorded (Stone, 1997, p. 21). In a transducer, multiple crystals rotate to emit a sound wave and receive 46 the reflected echo. The returning echoes are reproduced by a computer and displayed as a video image of up to 140 degrees (Stone, 1997, p.21). Ultrasound imaging for clinical use was first applied to speech in the 1960-1970s to measure dynamic tongue movements (MacKay, 1977; Stone, 1990, 1997). One of the main advantages of ultrasound over other experimental techniques used in speech research is that "it provides excellent studies of the tongue" (Stone, 1999, p. 252): it allows one to compare the movements of different parts of the tongue in real speech time.1 A transducer placed under the chin can capture images of the whole tongue in dynamic movement, either in coronal (crosswise) or mid-sagittal (lengthwise) views. a. Mid-sagittal b. Coronal /palam/ "wind" (Korean) "a leg" (English) Figure 3-1. 2D Ultrasound images of the tongue. Images of the mid-sagittal and coronal sections of the tongue during production of adults' IM are shown (files from ISRL, UBC). I A number of studies have used real time ultrasound to study tongue movements in speech (Stone, 1990; Stone, Shawker, Talbot & Rich, 1988, etc.). However, the early research focused on laboratory-based and clinical-based applications, and was not used in linguistic phonetic studies (Gick, 2002). 47 Early ultrasound studies of speech used 1 -dimensional measurements in the pharyngeal area. More recently, 2-dimensional (2D) and 3D techniques have been developed, along with improvements in image processing. While the 3D technique provides clearer images of the static tongue shape, time-varying movements are captured better with the 2D technique, because the temporal resolution is higher (Gick, 2002; Stone, 1999). Thus, 2D ultrasound imaging would be more ideal for measuring movements of the anterior and posterior areas of the tongue in the speech of IM, or for any natural speech. In addition to the capability to capture real-time tongue movement, ultrasound imaging (using only sound waves) is a safe and non-invasive means of measuring the tongue (Gick, 2002; Stone, 1997, 1999). While other techniques such as X-ray, magnetic resonance imaging (MRI), and electromagnetic midsagittal articulometer (EMMA) can be used for tongue measurement, none of these are suitable for use with children. First, excessive exposure to X-ray is hazardous; hence, it is not suitable for data collection. Further, since X -ray shows not only the tongue but also the jaw, palate, and teeth, it is difficult to measure the tongue only (Stone, 1999, p.247). Second, MRI is good for a static view but not for temporal measurement. Even the fastest MRI only takes 7-10 frames per second, whereas the ultrasound scan rate is as fast as 60 frames per second. Further, many people experience claustrophobia and cannot tolerate the procedure (Stone, 1997, p.20). Moreover, since a subject has to lie down without moving during MRI scanning, this might change the vocal tract shape (Stone, 1997, p. 20), and it would be difficult to use for real-time speech experiments, especially with children. Finally, the potentially harmful effects produced by the M RF s very strong magnetic fields are not well understood. Third, E M M A is a measurement that "tracks flesh point movement by measuring the movement of small 48 receiver coils through alternating magnetic fields" (p. 24), although it only is able to measure points on the anterior parts of the tongue (Stone, 1997). E M M A is quite invasive and requires a long setup time to attach coils to the articulators. Also, as with the MRI, the potentially harmful effects of an E M M A ' s magnetic fields are not well understood. Electropalatography (EPG) is another technique used in speech research of the tongue. EPG measures tongue-palate contact in real time and shows the tongue tip's contact area on the palate. However, since it only shows images when the tongue actually makes contact with the palate, no image is generated without contact (Stone, 1997). Since it can take 2 months to make an acrylic pseudo-palate that fits a palatal cast of each speaker, children might need another one to fit their growning palate, by the time it is ready to use. One limitation of ultrasound is that it only shows the tongue, and it is difficult to capture the tongue tip and epiglottis. The air at the tongue's surface reflects the sound wave, so that the structures on the far side of the vocal tract cannot be viewed: the palate and pharyngeal wall cannot be seen, and the jaw and hyoid bone in an ultrasound image appear as shadows (Stone, 1999, p. 252); due to the air under the tip, the tongue tip often cannot be seen clearly (Gick, 2002; Stone, 1997, 1999). Head and transducer movement correction is another issue, and researchers have continued to experiment with techniques to reduce head movement and control the position of the transducer relative to the head (Campbell & Gick, 2003; Gick, Bird & Wilson, submitted). Despite these limitations, ultrasound is in many ways ideal for speech research with children. It is safe and non-invasive, and it is relatively easy to set up an experiment. Moerover, ultrasound imaging can provide more complete information about overall tongue movement than other instruments such as E M M A (see Perkell et al., 1988, 1992; Stone, 49 1997) and EPG (see Gibbon, 1990; Hardcastle, 1972, 1974; Hardcastle & Hewlett, 1999). Recently, portable ultrasound machines with PC-accessible equipment have made fieldwork in speech research more accessible. Researchers at the UBC ISRL and elsewhere have been developing software for analyzing ultrasound data and have done ultrasound research on the timing and shape of articulatory gestures (e.g., Campbell & Gick, 2003; Gick & Wilson, 2001; Gick, Campbell, Oh, & Tamburri-Watt, in press; Oh, 2002; Oh & Gick 2002). 3.3 Methods In the present study, an experiment was conducted to explore the lingual gestures involved in the articulation of English IM in children's speech. Movement of the midsagittal section of the tongue was captured.2 To investigate spatial organization, the movement direction, tongue shape, and constrictions for the tongue tip and tongue dorsum were measured in syllable-initial, syllable -final, and intervocalic positions. To investigate temporal coordination, the movements of the front part of the tongue (i.e., tongue tip raising and tongue tip fronting) and of the back of the tongue (e.g., tongue dorsum backing) were measured for intergestural timing. Findings from the static analyses for the spatial characteristics of IM will be discussed in chapter 4; findings of the relative timing between gestures will be discussed in chapter 5. 2 Using ultrasound, the coronal section of the tongue also can be captured and measured. This is not included in the current study for three reasons. First, no study of coronal tongue movement in adult speech production is available for comparison. Second, technically switching a probe angle from a view of the midsagittal to the coronal section would delay an experimental session. This would mean that children would have to repeat the same words twice for both the coronal and midsagittal views. Children have relatively a short attention span, and collecting data for more than 20 minutes would make them lose interest. Third, if a transducer is fixed to a stand, switching its angle takes time to set up; whereas, if a transducer is handheld by an adult, the measurements will not be as reliable as those made from a fixed transducer. 50 3.3.1 Participants Eight monolingual North American English-speaking children (six girls and two boys) ages 3;11 to 5;9 (mean age 4;9) at the time of the experiments participated in the study. Seven out of the eight children were Canadian, and one child (FL) was an American, who had been living in Canada for three years. A set of identical twin sisters (LR and MC) also were among the participants. Children were selected carefully based on their ability to produce IM words. A l l but one already had passed the glide substitution stage and had been producing IM (perception by ear) in their utterances (see § 3.3.4.3 and 3.3.4.4 for detailed impressionistic and acoustic descriptions of these children's IM productions). The youngest subject RC (3;11, F) had [w] substitution for some /l/-initial words. Those tokens were not included in the analysis. Initially, JL was excluded from the study, because only her father is a Canadian English speaker — her mother is from New Zealand. However, JL had been spending a lot of time with her father at home and had been attending preschool, and so was impressionistically likely to use North American rather than New Zealand-influenced English.3 Four native English adult speakers perceived her speech to be North American English. Perceptual judgments of IM by adults will be discussed in chapter 6. JN participated in two sessions: a pilot study (4; 11) and this experiment (5;1). In chapter 5, both sessions are analyzed for their temporal characteristics to see i f a change exists in the temporal relation between gestures over a 3-month period. JN (4;11) was not included in the spatial measurements, because a different probe and different recording methods were used. More information on the participants is given in table 3-1. A l l participants were contacted under 3 All data collected from these children were reviewed by four adult Canadian English speakers and considered to be North American English /I/. Later, three other adults, who are professionally trained transcribers, were asked for their perceptual judgment of the small sets of tokens used by each child. 51 UBC Ethical Approval (B032) and voluntarily participated with their parents' written consent. Table 3-1. Participants List Initial/Gender Age Parents RC (F) 3;11 British Columbia (BC), Canada A D (M) 4;4 Alberta, Canada JL(F) 4;7 BC, Canada / New Zealand FL (M) 4;9 Illinois / New York, US JN(F) 4;11,5;1 BC, Canada LR(F) 5;6 BC, Canada M C (F) 5;6 BC, Canada K R (F) 5;9 BC, Canada Average 4;9 This age group was chosen for several reasons. First, IM production during the early stage of phonological development does not suffice to generalize the articulation of IM in acquisition. According to the literature in phonological acquisition, IM is often produced at around two-three but appears only in a limited context (Vihman, 1996). Working with two-to three-year-old children would be difficult, because they may know only one or two IM words. On the other hand, children at four and five already know a number of IM words and are old enough to be cooperative for an experimental session. Second, glide substitution is another problem for getting data from younger children. Children around ages two-four often substitute glides for the liquids IM and Ixl (Menn & Stoel-Gammnon, 1995; Stemberger & Bernhardt, 1998; Vihman, 1996). Whether they 52 perceive liquids as glides (and vice versa), or perceive them differently but are unable to produce them correctly, is not clear. In addition, whether sounds for which a glide is substituted should be considered as glides or something different from both glides and liquids is not clear, since no previous study regarding these matters has been done. Other issues such as consonant harmony and coarticulation in the early period of acquisition also make describing IM acquisition complicated. Although coarticulation is likely to occur in fast and fluent speech, and children produce less extensive coarticulation than adults (Kent, 1983, p.72), this study tries to avoid possible coarticulation effects in the speech of IM by children. With a baseline of the articulatory prototype of IM, such issues that may occur in earlier periods can be explained adequately . Third, although many children in the age group of four to six have adult-like IM, their IM may not be as refined as that of adults'. It has been reported that most children have mastered IM at around six to seven, and some children master it much later around age nine (Bernhardt & Stemberger, 1998; Vihman, 1996). In other words, children age four to six are in the process of mastering the production of IM: they produce IM more accurately and productively than younger children, but their IM may not be the same as that produced by adults. Thus, it will be interesting to compare IM produced by this age group with that of adult speakers. Finally, no studies have been done that refer to this age group with respect to direct measurement of articulatory speech development. One of the most in-depth studies investigates oromotor (lip and jaw) coordination of bilabials in children age one, two and six (Green et al., 2000), showing that lip and jaw movements change dramatically during the first one to two years, while more refined adult-like movements appear at age six. Although 53 the Green et al. study provides the developmental patterns of motor coordination at early and later stages, we do not know how it works for children between these two age groups. Coordination skills at age four to six might be more variable, more similar to the younger group; or they might be similar to what is found in the older group and/or adults; or they might not follow any pattern found in previous research and have their own distinct pattern. Thus, to provide a better understanding of speech development in general and to bridge the gap between younger and older age groups, it is important to investigate how children at age four to six use their coordination skills in speech. Development studies of lingual gestures (no matter what the age group) have not been done. Previous studies such as Green et al. (2000) only looked at lip and jaw coordination and not the lingual gestural coordination examined by the present study. Nevertheless, based on the findings in Green et al. (2000), we can predict that IV production at age four to six will be more stable than that produced by children at two and three, but less stable than that of older children or adults. 3.3.2 Stimuli Structurally, stimuli were designed to be as similar as possible across allophones of IV as well as to a previous study of adult III (Gick et al., in press). The target segment IV was collected in two-word combinations and in isolated words using a picture identification task. Two-word combinations included a word-initial IV preceded by a vowel (e.g., A # ladybug, bee # leaf), and a word-final IV followed by a consonant (e.g., tail # baby) and by a vowel 54 (e.g., tail # angel). Isolated words included a word-initial (e.g., ladybug), word -final (e.g., tail), and intervocalic (e.g., jellybeans) III. While IM in final position was measured in two conditions (VL#C and VL#V), initial IM was measured when preceded by a vowel (V#LV), since resyllabification normally affects only post-vocalic consonants in English (Kahn, 1995). Further, the initial IM preceded by a consonant (C#LV) is avoided because of possible coarticulation with the preceding coronal (e.g., and # leg, bucket # ladybug). Having the tongue tip already contact the alveolar ridge, it would be difficult to measure the complete movement of the tongue tip, as well as onset timing. Thus, the relevant question for ambisyllabicity concerns whether the final III followed by a vowel (VL#V) patterns similarly to the final IM followed by a consonant (VL#C), or to the initial allophone (V#LV). Isolated intervocalic III words (VLV), such as yellow, jellybeans, and helicopter were also measured for comparison. For stimuli, adjacent segments, especially vowels, were chosen carefully to maximize movement trajectories of the IM gestures and to minimize interference of the adjacent segments. That is, back vowels such as [u] and [o] were avoided since they overlap with, and potentially obscure, the dorsal gesture of III. Instead, front vowels such as [i], [ei], and [e] were used mainly so that the entire process of the tongue dorsum retraction was easily captured.4 In parallel, the labial consonants were selected with the same reason in mind. 4 Again, the target sound may have interference from the front vowels, yet the interference of an adjacent vowel is less crucial than that of a back vowel. Actually, it can maximize the movement trajectories of the tongue dorsum backing. With this in mind, Id was used, replicating a previous study of adult's /!/ (Gick et al., in press). 55 Table 3-2. Stimuli List Initial (V#LV, #L) Final (VL#C, VL#V, L#) Intervocalic (VLV) Ladybug Whale Yellow Leg Snail Baby Elephant A Leaf Jail Bee Teletubbies B Lips Tail Bed Helicopter J (Leash) Wheel A Jellybeans Bee (Lamp) Seal Angel Pillow (Lamb) Shell Eight Ear (Silly) (Lake) (Nail) (Willy Wonka) (Ladder) (Pail) (Jelly) (Words in parentheses are less familiar words) The vowel and word shape in both target and adjacent words were controlled as much as possible within the range of the children's vocabulary at these stages. However, making a stimuli list that is consistent for all participants was impossible, since each child has a different-sized lexicon based on their individual experiences and degrees of language input. For instance, for the vowel Id, although words such as lady, ladybug, lace, lake, late, and lame are available, it seemed that ladybug was the only one that every child could easily recognize. Adjectives such as late and lame were not suitable for the picture identification task, particularly when combined with another picture. Lady, lace, and lake were found to be rather difficult for this age group, since children often used "girl/doll" for lady, "string" for lace and "ocean" for lake. Moreover, some children did not know the word angel that was 56 used for the Id vowel following the final IM as in tail # angel.5 If angel were not to be used, only the letter A (not ace, ape, or ate) was a plausible alternative. The same stimulus constraints also limited available words across allophonic categories. For instance, vowels [ei] and [i] produce a good mirror image for both initial and final allophones, as in A # ladybug vs. tail # A vs. tail # baby for [ei] and bee # leaf vs. seal # ear vs. seal # leaf 'for [i]. However, no word exists that has a vowel [e] that could be used for a preceding vowel of the initial IM word such as leg. Hence, words with [ei] were used instead, as in A # leg. Typically for previous studies, adult participants have been asked to read from a stimuli list, whereas generally, children were not able to read from a list. Instead, two pictures of objects were shown at the same time, as in two pages of a book, and children were asked to name them in sequence (see figure 3-2 below). This strategy provided a similar effect to reading stimuli, since the pictures were arranged with an intent to produce the desired combinations.6 5 One child's (FL's) religious belief discouraged the use of the word "angel;" and another child (RC) simply had not acquired the word yet and used "fairy" instead. 6 In some cases, the combination of the two pictures seemed to be confusing to the children, as some of them switched the order in an effort to make sense of them. For instance, some children hesitated in saying anything when "jail" and "baby," or "nail" and "baby" were combined. 57 Initial (V#LV) ' A # ladybug' Figure 3-2. Picture Stimuli Examples Final (VL#(C)V) 'seal # bee' To begin, children were asked to identify the pictures. Once they were familiar with the pictures, the experimenter made a stimuli list for each individual child by choosing only those pictures recognizable to the child. Thus, some pictures were used for all children and some were not, leading to a slightly different combination of words for each child based on his/her vocabulary. Once the stimuli list was set for each child, two pictures were presented at a time, where IM was deliberately placed like it would appear in a word sequence of V # L V , V L # V , and V L # C. Then the child was asked to say the names of the objects from left to right. For isolated tokens, a single picture was used. An example of a typical experiment procedure is as follows: Experimenter: Look at this picture (pointing to one of the two pictures). Do you know what it is? Child: Yes (nodding). Experimenter: Can you tell me what it is? Child: (A) whale. 58 Experimenter: Good. How about this one (pointing to the other picture)? Child: (A) baby. Experimenter: That's right. Now can you say them together (pointing to the picture of a whale and a baby)? Child: Whale baby. Experiment: Yes, very good. 3.3.3 Apparatus Articulatory data were recorded to digital video (30 frames/sec) from an Aloka SSD-5000 ultrasound machine, using a 6.0 MHz 120-degree convex endo-cavity transducer. The transducer was attached to the equipment arm of an ophthalmic exam chair (see figure 3-3). Acoustic data were collected using a Pro-Sound YU34 unidirectional dynamic microphone amplified through a built-in mixer in a JVC D V D player and simultaneously recorded to the same DV tape as the video signal. Figure 3-3. Experiment Setup 59 3.3.4 Procedure Recording sessions were conducted in the Interdisciplinary Speech Research Lab (ISRL) at the University of British Columbia. The location and angle of the transducer was adjusted for each child to record the best image of tongue movements. The transducer was placed under each child's chin to view the midsagittal section of the tongue from the root to the blade. A microphone stand was placed beside the chair. Head movement was stabilized somewhat by having children sit leaning slightly forward while still resting their head against a two-point headrest attached to the back of the chair. Stimuli pictures approximately 2.5 metres away were presented at eye level in front of the children. Digitized ultrasound images of the midsagittal areas of the tongue were transferred to a PC (a Mac G4) using Adobe Premiere 6.0 software. Video images of the target sounds were edited into individual movie files. 3.3.4.1 Data Collection Each recording session took about 20 to 30 minutes. During the session, the parent(s) of a child usually stood behind the chair and gave assistance by holding the child's head. This procedure helped to stabilize the head position relative to the tongue for the ultrasound movies. However, during the experiment, parents were not allowed to speak to their children or correct their output. While an attempt was made to control the stimuli in a way similar to the previous study of adults' IM (Gick et al., in press) and to make it as consistent as possible across the 60 children who participated in the experiment, the combination of two words seemed to be unnatural and confusing to some children. Some children switched the order of the pictures, so their identifying words would sound more natural or make sense to them. For instance, an expected stimulus "snail # baby" was produced often as "baby # snail" instead, changing the syllable-final target (VL # C) to word-final. Sequences such as "jail # baby" and "nail # baby" were not favoured by some children. However, this unexpected word swapping was not always due to semantic reasoning, i.e., "leaf bee" instead of "bee leaf." Another tendency found in these children was that some of them liked to add "and" between the two words. The stimuli for the final allophone followed by a consonant (VL#C) caused frequent errors. For instance, they produced "seal and bee" instead of "seal bee" and "whale and baby" instead of "whale baby." This resulted in collecting more VL#V than the target sequence of VL#C, and these tokens were analyzed as a final allophone followed by a vowel (VL#V) instead. Future research needs to consider these kinds of issues more carefully when designing stimuli. A complete list of utterances collected from each child is provided in Appendix A. The number of tokens collected from each child was not identical. The number of tokens for each syllable position collected from each child was not same either. Due to individual differences in combining words and capability to understand the task, children produced different numbers of tokens and different combinations of stimuli. In addition, some children did not complete the session using all syllable positions, and some ultrasound images were not clear. Utterances with a long pause between two words, extremely long vowel articulation, or poor ultrasound images were discarded and not included for the measurement. 61 3.3.4.2 Spatial Measurement Two spatial measurements were made: one for tongue shape and one for the magnitude change of the tongue dorsum gesture. For tongue shape, the relevant frames showing the maximum constriction of the tongue tip and tongue dorsum gestures were extracted as JPEG picture fdes. For the tongue tip, the frame showing the tongue tip raised maximally towards the alveolar ridge or stretched maximally forward was selected. For the tongue dorsum, the frame showing the tongue dorsum backing reaching its maximum was selected. When the maximum constriction lasted for more than one frame without significant change of tongue movement (especially for the tongue dorsum backing), the first frame of maximal constriction usually was selected. Then, a line was drawn along the edge of the mid-sagittal view of the tongue, using a picture editing program of Microsoft Word. Tracings of the tongue (for different syllable positions of IM) were grouped together for each subject, so that general tongue shapes for each syllable position could be compared within each subject as well as across subjects. These images are presented in chapter 4. 62 Figure 3-4. Example of Tongue Shape Tracing: IM in "bell # bed" (KR 5;9) Magnitude change of the tongue dorsum has been an important cue to distinguish allophones of IM in adult's speech (Gick, 1999, 2003; Sproat & Fujimura, 1993). For this study, magnitude change of the tongue dorsum was measured only. In ultrasound imaging, it often is the case that the tongue tip is not clearly seen due to an air pocket below the tongue tip that hinders the view (see § 3.2). While it still was possible to see the movement of part of the tongue tip and the direction of the movement (i.e., upward, forward, or downward), it was not possible to capture the entire tongue tip for a valid measurement of magnitude. For magnitude change, the difference was measured between tongue dorsum backing at its maximum constriction and the tongue at speech rest position. First, speech rest positions of the tongue were taken between utterances to verify a consistent tongue position for each subject. Within a speaker, speech rest position has been argued to be as stable as vowel targets (Gick, Wilson, Koch, & Cook, submitted). Since ultrasound imaging techniques are relatively new in speech research, standardized spatial measures such as magnitude, constriction location, or tongue size comparison are not yet fully established. For instance, although images of the tongue would show its direction of movement when 63 constricted, specifying the location of the constriction relative to the whole tongue is not easy, since the angle of tongue's movement may not reflect actual movement direction and/or location. Besides, each speaker has a tongue of a different size, and the angle of the captured images of the tongue may vary across speakers due to where the transducer is positioned to capture the best image for each indiviudal speaker. In addition, the comparative sizes of children's and adults' tongues has not been studied yet. Consequently, this study uses speech rest position as a reference instead. This means that the magnitude change of the tongue dorsum was not measured from its initial movement to its maximum constriction, but from the speech rest position (averaged for all tokens per speaker) to its maximum constriction (see figure 3-5). This indirect measurement is useful for two reasons: first, a consistent rest position provides a reliability that images of the tongue are stable relative to the head position; and second, it provides a means to verify magnitude change across syllable position, since the magnitude change of all tokens can be compared, using the same reference point (speech rest position). 64 a. the tongue at rest position b. the tongue at maximum constriction TD backing: (x'- x) X c. measurement for TD backing Figure 3-5. Spatial Measurement Diagram. The tongue at rest position is marked in a bold line, and the tongue dorsum at the maximum constriction is marked in a dotted line. Tongue positions in the edited video images were measured using ImageJ version 1.6 software (see www.rsb.info.nih.gov/ii/ for more information). First, an arbitrary reference point was fixed at the bottom of the video images, at the approximate centre of the tongue arc. From that point, a line was drawn to the highest point of the arc of the tongue surface, while the tongue was at the rest position. The coordinate (x, y) values of that point were recorded 65 (figure 3-5a). This procedure was repeated for all rest positions, and then the numbers were averaged for each child. Second, the point on the tongue surface was identified at the extremum (maximum constriction) of the tongue dorsum movement (the point on the arc intersecting a tangent line drawn perpendicular to the primary direction of the TD movement, at a time when the tongue dorsum was at its greatest distance from the center reference point), and this point (x', y') was recorded (figure 3-4b). This was done for all allophones. Third, the difference between the tongue dorsum at maximum constriction and the tongue at rest position was calculated: (JC ', y') at maximum constriction minus (x, y) at rest position was converted into millimetres then averaged using Statview 5.01 statistics software on a Sony PC. Differences between values along the x axis (horizontal movement of the tongue) were calculated for tongue backing; and differences between values along the y axis (vertical movement of the tongue) were calculated for changes in tongue height (figure 3-5c). 3.3.4.3 Temporal Measurement Relative timing between TT and TD at full constriction was measured by counting the number of frames during the articulation of IV. Counting started when either a tongue tip or a dorsum gesture began for IV and continued until both gestures reached their maximum constriction. For maximum TT constriction, the frame was chosen where the tongue tip raising toward the palate was achieved maximally; and for maximum TD constriction, the frame was chosen where the tongue dorsum is pulled back maximally. When a maximum constriction lasted for more than one frame without visible changes, usually the first frame showing the maximum constriction was chosen. 66 Timing between TT and TD for each token was measured separately then averaged for each syllable position. The relative timing between TT and TD gestures was then calculated by subtracting a frame number of the maximum constriction of TD from that of TT (TT-TD). Counting the number of elapsed frames between TD and TT at their maximum constriction can result in either a positive or negative lag. If TT-TD is positive (TT being associated with a higher frame number than TD), a positive temporal lag is implied in which TD backing precedes TT raising. If TT-TD is negative (TD being associated with a higher frame number), a negative temporal lag is implied, because TT raising precedes TD backing. Note that this temporal measurement only shows the relative timing between TT and TD at their peak, and does not indicate the actual duration of each gesture. The number of frames (TT-TD) was then converted to milliseconds by multiplying by 33.333 (from 30 video frames per second). One frame is 33.3 ms long. Results of the experiments are discussed in the next two chapters. The spatial characteristics and static information of gestures of IM are discussed in chapter 4; and the temporal characteristics and relative timing of gestures of IM are discussed in chapter 5. 67 CHAPTER 4 SPATIAL CHARACTERISTICS OF IV IN CHILDREN'S SPEECH 4.1 Introduction This chapter investigates the spatial components of gestures of English IV in children ages 3;11 to 5;9. Articulatory studies of IV in adult language have shown that the gestural components of IV have distinct patterns with respect to spatial organization based on syllable position (see § 1.2.1). The question addressed in this chapter is whether the articulatory components of IM are all present and spatially organized in children's speech in the same way as in adult speech. IM comprises multiple lingual gestures, and the coordination of gestures may be difficult for children. Although IM produced by children may sound similar to an adult's IM, it may not be articulated using the same gestural components and/or organization. In addition to children's problems with coordinating multiple gestures, it also has been suggested that due to underdeveloped motor skills, they are likely to use a somewhat different gestural organization (Kent 1992; Nittrouer, 1993; Gick, 2002). I predict that the spatial characteristics of IM in children's speech will be different from that in adult speech, due to complexity of the gestural coordination and limited motor flexibility. In addition to describing the spatial behaviours of gestures of IM, two motor development hypotheses (see chapter 2 § 2.3.2) — the Serial Motor Development (SMD) and Parallel Motor Development (PMD) — are examined to determine which motor process(es) (i.e., Differentiation, 68 Integration, Refinement) can best accommodate the speech development of IM at this age group. In the next section, I elaborate on my predictions. The organization of this chapter is as follows: first, hypotheses and predictions about the spatial properties of IM are presented, then experiment results and discussion, and finally a summary. See chapter 3 for the methodologies. 4.2 Hypotheses 4.2.1 Gestural Coordination Based on the articulatory components of English IM in adult speech, the questions to be tested are as follows: (a) how many lingual gestures are used and what are their spatial organization (e.g., tongue shape and constriction)? and (b) how, i f at all, does the organization depend on syllable position (e.g., magnitude change) and how variable are the gestures (e.g., within vs. across subject/syllable)? I predict that spatial organization of English IM will be different in children. First, the degree of similarity of tongue shapes and constrictions for IM in children's and adults' speech will be tested. In adult speech, IM comprises two lingual gestures: tongue tip raising and tongue dorsum backing. In general, tongue tip closure at the alveolar ridge and tongue dorsum backing are present, although tongue tip closure can be less consistent in the final allophone in certain contexts (e.g., fast speech and some dialects of English) (Giles & Mol l , 1975; Hardcastle, 1985; Hardcastle & Barry, 1989; pc with Scobbie, 2004). If children have a spatial organization of IM similar to that in adult speech, both tongue tip raising and tongue dorsum backing gestures should be present in their speech production of IM. 69 Second, the sensitivity and variability of gestural organization to syllable position and subject will be tested. One distinction between the initial and final allophones of IM in adult speech is that a greater tongue dorsum backing gesture occurs in the final position. If children have a TD gesture similar to that found in adult speech, the magnitude change should be greater in the final allophone than in the initial one, and the tongue dorsum should be stretched more in the final than in the initial allophone (see § 1.2.1 and § 2.1). Moreover, for TT, I predict a tighter constriction (more closure and raising) in the initial allophone than in the final, and a greater reduction in the final allophone than in the initial (see § 1.2.1 and 2.1). An ambisyllabic allophone (VL#V) should show an intermediate degree of constriction of TT and a constriction of TD similar to the final allophone (Gick, 2003, see also § 1.2.1 and 2.2.4). However, children's production of sounds with multiple gestures is acquired late (e.g., [I ], [z], [3 ] , [r]) and would be difficult for them, i f they do not have the flexibility to use independent articulators. Further, i f multiple parts of one articulator (i.e., tongue) are to be used (i.e., apical and dorsal), they not only should be able to be moved independently, but also be coordinated. Thus, although children use multiple gestures of IM, their coordination can be challenging, and so they may coordinate them in a different way. In fact, a tongue tip gesture is commonly shared across alveolar coronals (e.g., IM, IM, Idl, Inl), although the manner of articulations are different (e.g., apical vs. laminal vs. nasal). Except for IM, alveolar coronals emerge as early as bilabials and vastly outnumber them in terms of frequency in both babbling and early word production. For instance, a subset of consonants [p, t, k, d, b, g, m, n, w, h, s] make up 92-95 percent of the consonants produced at 11-12 months (Locke, 1983) and predominate in early word production in English (Stoel-70 Gammon, 1985). It seems that children can stabilize their tongue on the alveolar ridge and easily make sounds such as [t], [d], or [n]. Thus, the difficulty with the articulation of IV may be due to problems with the coordination of tongue tip and tongue dorsum gestures, and/or complications of laterality (i.e., airflow, apicality, tongue dipping). Again, the tongue dorsum gesture also is shared with back vowels, since the dorsum is well developed for sonorants. In language development, children also commonly use big dorsal gestures to make vowel substitutions for liquids. Thus, the articulation of IV is difficult due to the need to coordinate multiple gestures. Consequently, I predict that the spatial coordination of IV in children's speech production will be different from that found in adults' speech, due to the complexity of control needed for the multiple tongue movements of IV. To ease performance loads, children may produce IV with less constricted gestures: tongue tip raising should be less than that in adult speech; and the tongue dorsum movement should not be as large as that found in adult speech. 4.2.2 Motor Development In this study, the other focus of investigation is the spatial articulatory characteristics of IV in children's speech in relation to motor development. The tongue has many degrees of freedom to be controlled. If children have less flexible tongue movements and less control skills (Kelso, 1977; Kent, 1992) than adults, the coordination of multiple movements for III will be different from that in adult speech. Based on this assumption, the spatial coordination of tongue movements in children's III will be described in terms of the motor developmental 71 process. In chapter 2 (§ 2.3.2), three general processes of motor development (Differentiation, Integration, Refinement) were introduced. Two hypotheses of motor development will be tested — Serial Motor Development (SMD) and Parallel Motor Development (PMD) — to determine which motor process(es) is/are present in the speech acquisition of IM in children ages 4 to 5. The SMD hypothesis predicts that motor development is age-related, and that one process will represent the speech development of IM at ages 4 to 5 accordingly. Green et al. (2000) suggest that the oral movements of labials show the characteristics of Differentiation at age 1, Integration at 2, and Refinement at 6. While the age-related changes suggest general developmental patterns, questions still remain as to what would happen between ages 2 and 6, and whether similar developmental patterns also would reflect the motor development of tongue movements, i.e., IM. Given the fact that the Differentiation and Integration processes are associated with earlier developmental stages, speech development at ages 4 to 5 is likely to be characterized as earlier periods in the Refinement process. Refinement is a process in which physiological coordination is similar to that of adults, but movements are made with more variation and need to be refined over time. However, a good deal of literature has reported that motor skills associated with the tongue are mastered later than those associated with oral gestures (Kelso, 1977; Kent, 1992; MacNeilage & Davis, 1990). If tongue movements associated with speech develop later than lip and jaw movements, the Integration process may be present at ages 4 and 5 instead. To summarize, the SMD hypothesis puts forward an argument that changes in articulatory coordination are age-related, and that each process approximately represents a certain age, although two processes may overlap in the course of a developmental sequence. 72 In contrast to the SMD argument, the PMD hypothesis predicts that motor development will not be age-related, and that all processes of motor development will be active (e.g., Sharkey & Falkins, 1985; Smith & Goffman, 1996). A series of studies of oromotor (upper/lower lips, jaw) coordination have found no significant difference among different age groups, and/or mixed results that younger children perform better than older children or similarly to adults in some aspects (e.g., Sharkey & Falkins, 1985; Smith & Goffman, 1996; Smiith & McLean-Muse, 1986) (see § 2.3.2). That is, speech motor performance of older children can be less accurate than that of younger children; and speech production of younger children may not be different from older children. In other words, general motor developmental processes can be associated with any age: hence, according to the PMD hypothesis, non-linear motor maturation is expected. Thus, PMD predicts that speech development of IM at ages 4 to 5 will be present in any of the motor development processes or all of them, assuming that the motor abilities of children at these ages are not as mature as that of adults. Following are specific predictions of each motor process in relation to the spatial characteristics in children's speech production of IM. Refinement (or Integration) is predicted by the SMD hypothesis; whereas, in children's speech production of IM any of these processes, or all of them, are expected by the P M D hypothesis. Differentiation: English IM is produced using multiple gestures. Specifically, the tongue tip is raised and fronted while the tongue dorsum is pulled backward. When spatial Differentiation has not been developed fully, these distinct regions of the tongue are not independently controlled, and the whole body of the tongue will move in the same direction (e.g., Kelso, 1977; Kent, 1992; Green et al., 2000). 73 Integration: In a child whose speech is still undergoing motor Integration, either a TT or TD gesture dominates in the production of IM. Even in adult speech, a tongue tip gesture may be dominant in the initial position, while a tongue dorsum gesture may dominate in the final position (e.g., Hardcastle & Barry, 1989). In a case where the spatial Integration of IM is lacking in children's speech, one of the two gestures will dominate more heavily than in adult speech, and this predominance will carry across syllable positions, contrary to the adult pattern. The existence of this dominance pattern is supported by the fact that some children tend to produce a "lighter" IM (more consonantal), while others produce a "darker" IM (more vocalic), consistently across syllable positions. Likewise, an adult-like gesture may be avoided altogether: e.g., TT constriction can be qualitatively modified, and a different tongue tip movement can appear instead of tongue tip raising.1 Refinement: Spatial motor "refinement" describes a process whereby a gesture or combination of gestures that have been correctly learned undergo incremental, quantitative improvement. During this ongoing process, children produce appropriate spatial characteristics for IM in terms of number and place of gestures, and allophonic variation, even though these are produced with less consistency and accuracy than adults. For example, while small TT and TD gestures may occur in unrefined speech, the tongue may not reach the degree of "stretch" typical of an adult tongue shape some, or all, of the time. ' Control of the tongue tip would be difficult, especially when flexibility is lacking. Measurement of the tongue tip can only be produced hydrostatically, using the internal compression of the intrinsic muscles of the tongue (Kent, 1992). 74 4.3 Results The spatial characteristics of English IM in children ages 3;11 to 5;9 were investigated. Qualitative measurements of the spatial characteristics such as number of gestures, tongue shape, and constriction were observed across and within children. Overall results suggest that children use different spatial organization for the speech production of IM. Quantitative measurement for magnitude change (tongue dorsum backing) across the allophones was also made. While overall results suggest a similar pattern to that of adults, individual results indicate that all general motor processes are active in these children's production of IM. 4.3.1 Number of Gestures and Constrictions A l l children in the experiment were capable of producing two lingual movements for IM in all syllable positions. In adult speech, a tongue tip raising and fronting gesture, and a tongue dorsum backing gesture are present (Browman & Goldstein, 1995; Gick, Campbell, Oh, & Tamburri-Watt, in press; Sproat & Fujimura, 1993). In children's speech, tongue tip and dorsum movements were present, but not all gestural components were identical to those found in adult productions of IM. While both tongue tip and tongue dorsum movements were present in most of the children's speech, the movements appeared less constricted compared to IM in adult English. In adult speech, the tongue tip has a raising and fronting gesture, and the tongue dorsum has a backing gesture (Browman & Goldstein, 1992, 1995; Gick, 2003; Sproat & Fujimura, 75 1993). In these children, tongue tip closure often was not achieved. Examples of the constrictions of tongue tip and tongue dorsum of IM in adult speech are shown in figure 4-1. Initial "lemon" Mrtu H N U V M i" " .. H tl li i > TT a. TT closure at the alveolar ridge b. slight TD backing Final "peel" ..... f " ^ '1 ...» ' TD a. magnitude reduction of TT, with the tongue body lowering b. greater TD backing Figure 4-1. Constrictions of English IV: adult (F) Results from children's speech showed two different lingual movements for the tongue tip. While one of these two tongue tip movements (either a raising or fronting) was present in all positions in most children, tongue tip lowering was also observed. The tongue dorsum, on the other hand, had a more consistent backing movement. Nevertheless, the tongue dorsum was visibly less stretched than in adult English IM. For instance, even though most children used a tongue dorsum backing movement, they were likely to move the entire tongue body (or at least the posterior area of the tongue) backward, instead of pulling it backward by lowering the middle of the tongue. 76 Interestingly, for the tongue dorsum, one child A D (and RC for some tokens) did not have tongue dorsum backing for most of his utterances of IM. Instead, he moved the tongue dorsum forward in the same direction of tongue tip movement. In addition, some tokens in two children (MC, RC) showed a palatal constriction (e.g., "tail," "yellow"). Examples of ultrasound images with tongue surface tracings are shown in figure 4-2. Initial /!/ FinalIM TT raising TD backing TT lowering TD backing I " i TT raising TD backing TT lowering TD backing Figure 4-2. Constrictions of IV: children. Along with tongue surface tracings, constriction locations and tongue shapes are shown in the mid-sagittal view of the ultrasound. The tongue tip is facing right. 77 Constrictions of other coronals (e.g., IM and /n/) in the initial position are also compared with that of IM. While IM and Inl have a similar tongue shape in the initial position with the tongue tip raising, the tongue shape and constriction of IM (especially the tongue tip) in the initial position seem rather different from the former two (see figure 4-3). Initial IM and Inl Initial IM II Ii 1 Figure 4-3. Constrictions of It/, Inl, IM: children. Along with tongue surface tracings, constriction locations and tongue shapes of coronals in the initial position are shown in the mid-sagittal view of the ultrasound. The tongue tip is facing right. 78 4.3.2 Tongue Shape Tokens for four syllable categories (V#LV, VL#V, VL#C, V L V ) were qualitatively analyzed by overlaying tracings of the surface of the tongue (generated from tokens in each category) to show a general tongue shape for each allophone of IM. However, the number of tokens collected from each child was not the same, and some categories did not have enough tokens or lacked data, due to experimental and/or technical difficulties. This makes comparisons of syllable positions across children and within children impossible. Nevertheless, individual results provide qualitative information about the spatial characteristics of children's IM at these ages. The number of tokens used is shown in table 4-1. Table 4-1. Distribution of Allophones of IV RC AD JL FL JN LR M C KR (3;11) (4;4) (4;7) (4;9) (5;1) (5;6) (5;6) (5;9) #LV 6 7 8 10 2 ' 3 0 6 V# LV 0 0 0 9 10 3 4 14 V L # C 3 0 6 18 14 0 2 24 VL# 15 8 2 12 8 7 1 8 V L # V 3 0 17 13 6 10 10 29 VLV 8 10 8 10 9 8 8 17 Total 35 25 41 72 49 31 25 98 Note: # L V : word-initial, V#LV: syllable-initial, VL#C: syllable-final, VL#: word-final, VL#V: ambisyllabic, V L V : intevocalic 79 The tongue shape that the children used in the allophones of IM indicates that at these ages, they use different tongue movements for IM. Ultrasound images of the tongue for each allophone were grouped together. A tracing of the mid-sagittal section of the tongue was drawn on each image, then overlaid on top of each tracing that was used to determine the general tongue shape in allophonic variation. In figures 4-4 to 4-10, ultrasound images on the left column selectively show the tongue shape for each syllable position, and the tracings on the right column show cumulative results of the overlaid tracings from all tokens in each syllable category. Also, ultrasound images shown in the figures are smaller than the tracings shown next to them, since the images were reduced in size after the tracings. Figure 4-4a illustrates how tracings were made. FL (4;9), M Results for FL showed an adult-like pattern in at least some tokens in all syllable positions. The tongue tip and tongue dorsum constrictions were present. For some tokens in the final position, no tongue tip raising was present. Figure 4-4 shows tongue shape at the maximum constriction of TD. 80 Tokens Tongue shape at TD constriction Tracing of individual token Overlaid tracings from all tokens Bell bed ' ' 1 / / j .•* •* ; : / / : Jail baby I * • • * • . -* • • •** » • Snail baby • • '• Whale bee ' 1 Tracings of other tokens in this category also were combined. Figure 4-4a. Tongue Shape (Final VL#C): F L The ambisyllabic allophones (VL#V) showed a rather different tongue shape than the final allophones followed by a consonant (VL#C). With less TD backing, the tongue shape in the ambisyllabic allophones was more similar to that in the initial allophone (V#LV) than to the final allophone (VL#C). 81 Tokens Tongue shape at TD Tracing of individual Overlaid tracings constriction token from all tokens Bell egg Nail eight Figures 4-4b. Tongue Shape (Ambisyllabic VL#V): FL Tokens Tongue shape at TD constriction Tracing of individual token Overlaid tracings from all tokens A ladybug . « * -! A lake i Figures 4-4c. Tongue Shape (Initial V#LV): FL For both allophones (shown in figures 4-4b and 4-4c), tongue tip raising often was not present in speech. 82 Tokens Tongue shape at TD constriction Tracing of individual token Overlaid tracings from all tokens Elephant Helicopter Yellow fe ^ ' i Figure 4-4d. Tongue Shape (Intervocalic V L V ) : F L The intervocalic allophones showed least tongue dorsum backing in FL. The tongue shape in these allophones was similar to the tongue at rest position, although tongue tip raising was more present. 83 JN (5;1), F A l l allophones showed a similar tongue shape. Tongue dorsum backing in the final allophone (VL#C) seemed to be slightly greater than that of other allophones. A slight tongue tip raising was observed, but it was impossible to tell whether complete tongue tip closure occurred. Syllable position Tongue shape at TD constriction Overlaid tracings from all tokens Final VL#C Bell baby El Ambisyllabic VL#V Snail angel Initial V#LV J ladybug Intervocalic Yellow Figure 4-5. Tongue Shape: J N 84 JL(4;7), F The stretching of JL's tongue showed an adult-like backing gesture. In her speech, the ambisyllabic allophone (VL#V) had a greater tongue dorsum backing than that in the other final allophone (VL#C). This indicates that JL treated the ambisyllabic allophone similar to the final allophone, and that both postvocalic allophones behave as syllable-final allophones rather than syllable-initial. Whereas the intervocalic allophone seemed to pattern with the postvocalic allophones, the tongue shape of the word-initial allophone was different from the others. Syllable position Tongue shape at TD constriction Overlaid tracings from all tokens Final VL#C Bell bed Ambisyllabic VL#V Nail angel # Initial Leg Intervocalic Elephant Figure 4-6. Tongue Shape: J L 85 KR (5; 9), F K R seemed to use a relatively similar tongue shape for the initial and final allophones, and the intervocalic allophones, even though they were patterned a little differently from others. The tongue tip was missing in most of the images, and tongue dorsum backing was relatively fast compared to that of the other children. Syllable position Tongue shape at TD constriction Overlaid tracings from all tokens Final VL#C Bell bed Ambisyllabic VL#V Jail A Initial V#LV A lemon Intervocalic Yellow Figure 4-7. Tongue Shape: K R 86 LR (5; 6), F Ultrasound images for LR showed similar tongue shapes for all allophones. No significant tongue dorsum backing was seen across allophones. No tokens appeared in the syllable-final position, as a result of her adding "and" between the two words. The word-final allophone seemed to have less tongue dorsum backing than the other allophones. Images of the tongue for the initial allophone were not clear. Syllable position Tongue shape at TD constriction Overlaid tracings from all tokens Final # Angel and whale Ambisyllabic VL#V Snail and angel Initial V#LV A Loon Intervocalic V L V Yellow Figure 4-8. Tongue Shape: L R 87 MC (5;6), F No significant tongue dorsum backing was observed across allophones. However, the initial allophone was generally more anterior than the final allophones. Images for the initial allophone (V#LV) were generally less clear, and she often added "and" between the two words. The final allophone followed by a consonant (VL#C) appears to have less tongue dorsum backing. The initial and intervocalic allophones seem to have more anterior raising than the final allophones. Syllable position Tongue shape at TD constriction Overlaid tracings from all tokens Final V L # C Tail baby un — i Ambisyllabic V L # V Jail and bed Intial V # L V Jleg m • ^ Intervocalic Yellow Figure 4-9. Tongue Shape: M C 88 RC(3;11), F Similar tongue shapes for all allophones were observed. The syllable-initial allophone (V#LV) was not available, because this child did not know the letter " A " in " A # leg." Tongue tip raising was present, but no apparent tongue dorsum backing was observed for most tokens in all allophones. Interestingly, she substituted [w] for some initial IM words (e.g., leaf, leash, ladybug). However, the tongue tip constrictions for [1] and [w] were distinct from each other. The intervocalic allophone showed the least tongue dorsum constriction. [w] substitution Figure 4-10. Tongue Shape: R C 89 AD (4;4), M A l l tokens used were IM words in isolation. Ultrasound images for A D showed no significant tongue dorsum backing, and the tongue tip constriction was more consistent in the initial than in the final position. Interestingly, the direction of tongue dorsum movement was opposite to that found in other children. For instance, the tongue dorsum was pushed forward from back to front, instead of front to back. Coupling of the gestures was often observed, and tongue tip and tongue dorsum movement was less coordinated with respect to timing. Syllable position Tongue shape at TD constriction Overlaid tracings from all tokens Final # Wheel # Initial Lemon l 1 Intervocalic Yellow Figure 4-11. Tongue Shape: A D 90 4.3.3 Tongue Dorsum Backing First, each child's tongue was measured at the speech rest position to see whether the tongue position was stabilized, relative to the head, during the experiment. A N O V A results showed that the tongue at speech neutral position is highly significant across subjects (F (7, 368) = 41.225, p < .0001). A two-tailed t-test showed that tongue at speech rest position (within subjects) was significantly different from the null hypothesized mean zero (p < .0001). For one child (MC), the tongue rest position was placed in two different locations, because the recording had to be paused to fix head movement. Bivariate Line Chart Split By: subject 22 6 10 15 20 25 neutral x 30 35 40 Figure 4-12. Tongue Location at Speech Rest Position (scattergraph) 91 AD KR JN 0 5 10 15 20 25 30 35 Figure 4-13. Speech Rest Position: Distance from reference point (mm) Figures 4-12 and 4-13 indicate that speech rest position for each subject is stable (except for MC) , suggesting that it can be used as a reliable reference point (Gick, Wilson, Kock, & Cook, submitted). For quantitative analysis of magnitude change, the alpha level for probability of a Type 1 error was set to .05 (p < .05). Results for quantitative analysis of the height and backing of the tongue dorsum (TD) across syllable position showed that the height of TD gestures in postvocalic allophones is less than the TD height in prevocalic and intervocalic allophones. A N O V A results showed that magnitude change across syllable positions is highly significant (F (5, 370) = 15. 311, p < .0001). Results for the TD backing distance from the tongue rest position showed that backing is greater in postvocalic allophones than pre- and intervocalic allophones. The intervocalic allophone showed the least backing. A N O V A results for the backing showed that distance change in syllable position is significant (F (5, 370) = 4.081, p = .0013). Fisher's PLSD post-hoc tests showed that tongue dorsum backing in isolated initial and final allophones was not 92 significantly different (p >.05); however, tongue dorsum backing in ambisyllabic allophones was significantly different from the others (p < .05). # intervoc Figure 4-14. TD backing and height (scale shown in mm). Isolated words are shown in grey, and two-word combinations are shown in black. The abbreviations used for allophones are as follows: # initial (word-initial), v-initial (syllable-initial, V#LV), final-v (ambisyllabic, VL#V), final-c (syllable-final, VL#C), final # (word-final), and intervoc (intervocalic). 4.3.3.1 Individual Results Individual results of TD backing also are discussed in § 4.3.2.2. These individual results may not be statistically meaningful, nor sufficient for comparison across and within subjects, because (a) a relatively small set of tokens for each category were collected, and (b) not all children produced IM in all categories of allophones. Individual results are presented here to show various patterns of TD backing across children at these ages. 93 Results show different patterns of tongue dorsum backing for each participant. In FL, the ambisyllabic allophone had greater TD backing than other allophones, but in the syllable-initial and syllable-final allophones, TD backing was similar. Also, syllable-based final allophones had greater TD backing than isolated IM. In LR, word-initial IM showed greater TD backing than the rest of allophones, and TD backing did not distinguish allophones. In M C , TD backing in the syllable-final allophone was different from that in the ambisyllabic and word-final allophones. In RC, TD backing did not distinguish allophones, although the v-initial allophone behaved similar to the intervocalic allophone. In isolated forms, IM had greater TD backing than syllable-based allophones. For JL, while the TD backing in the word-final IM was greater than the final allophones (final-v and final-c), the final allophones were not significantly different from each other (p > .05). Also, the intervocalic allophones did not have a TD backing movement. In K R , the word-initial and intervocalic allophones patterned similarly, whereas the syllable-initial allophone patterned similarly to the final allophones. For JN, TD backing in syllable-final vs. word-final, ambisyllabic vs. word-initial and intervocalic allophones was similar. For A D , both word-final and intervocalic allophones showed a negative change in magnitude, resulting from reversed tongue body movement (forward movement instead of backing). 94 # initial v-initial final-v final-c final # intervoc FL 25 20 • # initial v-initial final-v final # intervoc LR # initial final-v final-c final # intervoc JL # initial v-initial final-v final-c final # intervoc KR # initial v-initial final-v final-c final # intervoc JN Figure 4-15. TD backing: individual results (scales shown in mm) 95 4.3.3.2 Age and Gender The results of TD backing in syllable position were compared to determine the effect of age and gender. Eight children who participated in the study were divided into two groups by age (4, 5). The four children in the age 5 group were JN (5;1), LR (5;6), M C (5;6), and K R (5;9). The four in the age 4 group were RC (3;11), A D (4;4), JL (4;7), and FL (4;9). Note that RC was included in the latter group, although she was one month short of being age 4. The six allophones of IV were divided into three categories: initial (#LV, V#LV), final (VL#, VL#C), and intervocalic ( V L V , VL#V). Results for TD backing in syllable position by age showed a significant difference (p < .0001). TD backing in syllable position was also significant between the final and initial allophones (p = .0004), and between the final and intervocalic allophones (p = .0002). However, syllable position between the initial and intervocalic allophones was not significantly different (p > .5). TD backing in the age 5 group was greater than that in the age 4 group, especially for the final allophones. . • final - • initial O intervoc Age 4 Age 5 Figure 4-16. TD backing by Age 96 TD backing in the syllable position was also compared between male and female participants. Only two male children (AD and FL) participated in the current study. Again, results showed that syllable position was significant between the final and initial allophones (p = .0002), and between the final and intervocalic allophones (p = .0002), whereas no significant difference occurred between the initial and intervocalic allophones (p > .5). TD backing in syllable position in the speech of male and female children was significantly different by gender (p < .0001). It showed that TD backing in girls' speech was greater than that in boys' speech. Figure 4-17. TD backing by Gender 4.4 Discussion Results for the spatial organization of IM indicate that children often use two lingual movements for IM. However, the constrictions and tongue shapes for the allophones often do 97 not pattern with those found in adult production of IM. This indicates that the coordination of the gestures of IM are not yet mastered, at least by the children participating in this study. For the anterior movement, lowering of the tongue tip, or fronting of the tongue tip without raising, was frequently evident, and tongue tip closure seemed to be missing in some cases within and across subjects. Note that the intervocalic allophone had a more consistent tongue tip raising than other allophones, although tongue tip constriction was still missing in many tokens across subjects. A lack of tongue tip closure is also evident in adult speech in British dialects or Midwest American dialects (pc with Scobbie, 2004, and Davis, 2003), or in fast and casual speech (e.g., Hardcastle & Barry, 1985). However, a more detailed study of the tongue tip is required regarding this issue, and an EPG (or E M M A ) study done in conjunction with ultrasound techniques would be able to provide a better answer. In fast or casual adult speech, tongue tip closure often is missing or is less tightly constricted in the final position. This suggests that tip raising requires more refined and detailed control of the tongue muscles than tongue tip lowering or fronting. Except in such cases, the loss of tongue tip closure in children's speech indicates the lack of motor "Integration" needed to simplify or modify the degrees of freedom of the tongue. The anterior movement of the tongue appears to be reduced in the articulation of IM to accommodate the posterior movement of the tongue (dorsum backing). Further, results from these children's IM production also support ongoing "Differentiation" for tongue tip fronting, since tongue tip raising is not yet differentiated from fronting. These data support 2 In a recent study, Scobbie (pc, April 2004) also mentioned that tip closure can be missing in both adult and child British English. 98 the prediction that each process of motor development is active at the same time for different types and parts of movements. As for TD backing in children's speech, it is interesting to note that overall results pattern similarly to that of adults across syllable positions. The tongue dorsum gesture was present within and across allophones in most cases; however, the degree of constriction was not as great as that in adult speech (showing ongoing Refinement). Except for JL and JN, tongue dorsum backing in most cases was somewhat fast and transition was not smooth, especially in the postvocalic allophones. This indicates that the movement is less flexible or smaller than that in adult speech. Since children's coordination skills are not fully matured, it is possible that the action of pulling the posterior part of the tongue backward in slow motion may be difficult to accomplish. While all children exhibited a variety of patterns, some favoured a particular process over others. Thus, while all children showed at least some adult-like abilities undergoing "Refinement," for some children (particularly A D and to some extent RC), most instances of IM could best be characterized spatially in terms of the "Differentiation" process. For example, most of A D ' s IM productions showed coupling of the tongue tip and tongue dorsum movements, with both moving in the same direction as a unit, either forward or backward. For most children's IM, a single lingual movement was evident in some tokens, indicating an ongoing "Integration" process in which movements are simplified or modified to become a functionally dominant gesture (either TT or TD). These observations correspond with the adult listeners' perception judgments that some children's IM was lighter or darker than other children's across syllable positions. Working in parallel, all the three motor processes were evident in most children's IM production. Thus, results from this study 99 support the PMD hypothesis that all of these motor processes are active in all ages, both within and across children, and further imply that younger children of these ages can coordinate speech motor processes in ways similar to older children. Until around the age of six, children reportedly have less stability in generating patterned speech output, and their speech performance of the gestural components is likely to display distinctive developmental constraints (Smith & Goffman, 1996; Thelen & Smith, 1994). Possible difficulties in the production of IM can reside in complex tongue movements and in coordination of the articulatory gestures. Thus, the spatial coordination of IM in children's speech can be different from that of adults, due to motor restrictions and the difficulty of multiple coordination (e.g., Kelso, 1977; Kent, 1992). 4.5 Conclusion In this chapter, the spatial characteristics of IM — with respect to the number of gestures, constriction, tongue shape, and TD backing — were investigated using ultrasound imaging. We have seen that children at around ages four to five use a variety of spatial characteristics for IM. The results in this chapter suggest that the articulatory gestures of IM at these ages are not mastered fully, and adult-like lingual gestures of IM are not yet consistently present. Although some children used adult-like lingual gestures and distinguished allophones, their tongue movements were generally less flexible and more variable across syllable positions. For easier articulation, others were likely to modify or simplify the tongue tip gesture. Overall, while a good deal of variation exists across children, the patterns seen in the spatial organization of IM are evidence (both across and within children) that multiple 100 and different motor processes are simultaneously active, as supported by the P M D hypothesis for motor development. 101 CHAPTER 5 TEMPORAL CHARACTERISTICS OF IV IN CHILDREN'S SPEECH 5.1 Introduction This chapter studies the temporal characteristics of the articulatory gestures of IM in children's speech, and tests two motor development hypotheses—the Serial Motor Development (SMD) and Parallel Motor Development (PMD)—for the speech development of IM at ages 4 to 5. The main question addressed is: how does the relative timing between gestures of IM differ in each syllable position? In adult speech, the relative timing of the gestures of IM is responsive to syllable position: syllable-initially, the tongue tip (TT) raising and tongue dorsum (TD) backing gestures occur almost simultaneously, whereas in syllable-finally, a TT delay occurs, as the TD gesture starts earlier than the TT gesture, causing a positive lag between them (Gick, 1999; Gick et al., in press; Sproat & Fujimura, 1993). In other words, the relative timing between gestures is distinct across syllable positions: a positive lag in the final position, and no lag or a slight negative lag in the initial position. For the intervocalic position, previous studies of ambisyllabicity have found that the ambisyllabic allophone (VL#V) has an intermediate positive timing, although it may not be significantly different from that of the initial allophone (Gick, 2003, Gick et al., in press). The difference in gestural timing based on syllable position adds another level of complexity to the speech acquisition of IM. Children not only must know that two lingual 102 gestures are involved in the articulation of IM, but also must control the movement in the opposite direction (the tongue tip is moving forward and upward, whereas the tongue dorsum is moving backward) with changes in magnitude according to syllable position. In addition to the spatial coordination, children must time the gestures appropriately according to syllable position. Thus, for children, representing the multiple gestures of IM in different timing, along with the spatial coordination, can be a challenging motor task. Given this, the present study attempts to find out whether intergestural timing is manifest in syllable position in children's speech production of IM. If children coordinate the articulatory gestures of IM similarly to that of adults, distinct intergestural timing is expected for the syllable-based allophones of IM. However, i f children coordinate the gestures differently from what is found in adult speech, different timing patterns might be observed, or timing may not distinguish all the allophones of IM. Considering the requirements of complex coordination skills in the articulation of IM, I predict that children's temporal coordination of the articulatory gestures of IM will be different from adults. Based on three motor processes—Differentiation, Integration, and Refinement—the following predictions can be made (see also chapters 2 and 4). First, previous literature in motor development suggests that the temporal differentiation of distinct movements is complex, so that at young ages these movements should be likely to occur simultaneously, i.e., coupling (e.g., Kelso, 1977; Kent, 1992; MacNeilage & Davis, 1990). If children lack a differentiated motor skill (Differentiation), they might not be able to coordinate the tongue movements in a time varying sequence. Thus, syllable-based allophones may not be distinguished by intergestural timing, resulting in simultaneous movements between the gestures across syllable positions. 103 Second, i f children tend to use a particular movement dominantly due to difficulty in multiple coordination (Integration), gestures will be coordinated in similar timing across all allophones. Thus, a consistent lag (either positive or negative) will occur in all syllable positions. That is, the relative timing between the gestures in initial and final positions can be either in simultaneity (which is generally expected for initial allophones) or with a common lag (which is expected for final allophones) regardless of syllable position. This further predicts that children who show such patterns simply lack a particular allophonic distinction at a phonological level. For instance, i f children have not learned syllable-based allophonic variation, both initial and final allophones could be produced as either "light" or "dark" 1 or randomly as either (Bernhardt & Stemberger, 1998). Third, if children coordinate the gestures similarly to adults but with more variability (Refinement), the relative timing will be different for at least some of allophones of IM, if not all. Moreover, intergestural timing can be shorter or longer than that of adult' speech, as children's motor coordination skills are undergoing a process of refinement. In the following section, I will use two hypotheses to elaborate on predictions based on motor development. The organization of this chapter will be as follows: predictions based on two hypotheses on the temporal characteristics of IM are laid out in § 5.2; experiment results are presented in § 5.3, discussion follows in § 5.4, and a summary concludes the chapter in § 5.5. The experiment used for the temporal characteristics of IM is same as that for the spatial characteristics of IM in chapter 4. For the experimental methodology, please see chapter 3. 104 5.2 Hypotheses Intergestural timing between the tongue tip and tongue dorsum gestures of allophones of IM is investigated based on the processes of motor development (Differentiation, Integration, Refinement) referred to in Green et al. (2000). Two motor development hypotheses—the Serial Motor Development (SMD) and Parallel Motor Development (PMD)—are tested with respect to the speech production of IM in children at ages 4 to 5. While both hypotheses assume that the motor abilities of children at these ages are not as mature as that of adults, their views regarding the age effect on motor development are different (see also chapter 4, § 4.2). The SMD predicts that speech motor development is age-related, and one process (either Integration or Refinement) will describe the speech development of IM at these ages. In the study of oromotor coordination in labials, Green et al. (2000) refer to age 1 as lack of differentiation, age 2 as integration, and age 6 as refinement. Based on this, the SMD would predict that motor coordination for IM at ages between 4 and 6 would be described as either Integration or Refinement. Considering that lingual gestures are mastered later than lips and jaw movements, it is likely that Integration still will be active at ages 4 to 6. Conversely, due to a lack of study of ages between 2 and 6, we do not know exactly when, in the course of motor development, Refinement would take place. Thus, it also is predicted that Refinement will be active at ages 4 to 5. On the other hand, the PMD hypothesis predicts that motor development is not affected by age, and, thus, all general processes of motor development can be active (i.e., the tongue) as they are, in fact, in adults who are learning a new motor skill. That is, 105 Differentiation, Integration, and Refinement processes will appear in the speech production of IM in children ages 4 to 5.1 predict that the temporal coordination of the gestures of IM will be different in children, and that all three motor processes will be present at these ages. Regarding the relative timing of IM in children, specific predictions for each process are suggested below. Differentiation: English IM is produced using multiple gestures in different timing relationships. When temporal differentiation is not fully developed, relative timing between two lingual gestures/movements is not distinguished across syllable positions. In adult IM production, the intergestural timing in the initial allophone is roughly simultaneous, while that in the final allophone shows a lag (e.g., Browman & Goldstein, 1995; Sproat & Fujimura, 1993). If children's motor development is in this process, the two gestures will occur simultaneously in some cases, regardless of syllable position, showing that the gestures are not yet temporally differentiated. Integration: Coordinating two different parts of the tongue, with different timing according to syllable position, is a complex task. In speech where motor integration still is being developed, movements may be differentiated, but not coordinated in an expected way; or, a single, dominant pattern may be adopted across allophones. Thus, for a child whose speech still is undergoing motor integration, a single timing pattern with either positive (TT preceding TD) or negative (TD preceding TT) lag will dominate, and this predominance will carry across syllable positions, contrary to the adult pattern. The existence of this dominance pattern is supported by the fact that some children tend to produce "lighter" IM (more consonantal) while others produce "darker" IM (more vocalic) consistently across syllable 106 positions. Stemberger (1988) notes that children's syllabification across word boundaries need not be adult-like, and there can be greater resyllabification. Refinement: "Temporal motor refinement" describes a process whereby appropriate intergestural timing between gestures is present, but is undergoing incremental, quantitative improvement. While this process is ongoing, children will produce appropriate timing relations in terms of allophonic variation; however, these are produced with less consistency and accuracy than adults. Thus, for example, while allophones of IM are distinguished by intergestural timing (simultaneous/negative lag vs. positive lag), the positive lag in the final allophone could be longer or shorter than that in adults; and/or a negative lag in the initial allophone may be longer than that in adult's IM. If an ambisyllabic allophone behaves like a final allophone, it should have a positive lag as in adult speech (Gick, 2003); i f it is resyllabified and behaves like an initial allophone, it should not have a positive lag (e.g., Gick et al., in press; Sproat & Fujimura, 1993). The following section presents the results of the temporal characteristics of IM. 5.3 Results An ultrasound study was conducted to measure the timing relations between the articulatory gestures in allophones of IM in children aged 4 to 5, and to test how the temporal coordination of the gestures of IM are accounted for in terms of motor development hypotheses. A total of 435 tokens were analyzed: 240 tokens for postvocalic allophones, 111 tokens for prevocalic, and 84 tokens for intervocalic (see table 5-2). The alpha level for the probability of a Type 1 error was set experiment-wise to .05 (p < .05). Overall results in 107 relative timing across subjects are presented first, followed by individual results. Note that individual results provide within-subject variation in timing; however, due to a relatively small number of tokens per syllable position, and uneven numbers of categories across children, the results do not have statistically strong interpretations. 5.3.1 Relative Timing Results for the relative timing between the gestures of IV in child speech suggest that gestural timing is distinct across syllable positions. The relative timing between the gestures in all syllable positions was significantly different from a hypothesized mean difference of zero in two-tailed / tests (p < .0001) (see Appendix D). Averaged relative timing between TT and TD gestures in initial (V#LV, #LV), final (VL#C, VL#), and intervocalic (VLV, VL#V)) positions showed that the timing in each syllable position is significantly different from each other: the lag found in the final position was longer than that in both the initial and intervocalic positions. A positive lag was found in the postvocalic position (41.2 ms); and no significant lag was found in the intervocalic (7.5 ms) and prevocalic (-6.9 ms) positions. A N O V A results for the mean values comparing lag in syllable positions showed that the overall effect of syllable position is significant (F (2, 432) = 40.99, p < .0001). Fisher's PLSD post-hoc test also indicates that the mean differences between the post- and prevocalic (48. 1 ms) and between the post- and intervocalic (33.7 ms) lags are highly significant (p < .0001), whereas the mean difference between the pre- and intervocalic lag (-14.4 ms) is marginally significant (p <.05). 108 120 -80 -1— ' 1 ' final initial intervoc Means Table for TT-TD (ms) Effect: syllable Count Mean Std. Dev. Std. Err. final 240 41.209 52.826 3.410 initial 111 -6.900 47.967 4.553 intervoc 84 7.532 39.146 4.271 Fisher's PLSD for TT-TD (ms) Effect: syllable Significance Level: 5 % Mean Diff. Crit. Diff. P-Value final, initial 48.109 11.108 <0001 S final, intervoc 33.677 12.268 <0001 S initial, intervoc -14.432 13.995 .0433 S Figure 5-1. Intergestural Timing (TT-TD) in Three Syllable Positions (ms) The relative timing in syllable position was also compared for the effect of age and gender. The eight children that participated in the study were divided into two groups by age (4, 5). Four children in the age 5 group were JN (5;1), L R (5;6), M C (5;6), and K R (5;9); and four in the age 4 group were RC (3;11), A D (4;4), JL (4;7), and FL (4;9). Note that RC was included in the latter group although she was one month younger than 4 years. Allophones were divided into three basic syllable positions: initial (#LV, V#LV), final (VL#, VL#C), and intervocalic ( V L V , VL#V). While relative timing between the initial and final allophones (p < .0001), and between the final and intervocalic allophones (p < .0001), was significant in both age groups, as well as in both gender groups, the relative timing between the initial and intervocalic allophones was not significant (p > .05). Neither age nor gender effect was found 109 regarding the relative timing in the three syllable positions (p > 05). However, the relative timing in the final allophones was longer in the age 4 group and in the male group; and in respect to age and gender, the opposite patterns were found in the initial allophones. I male fj female Figure 5-2. Intergestural Timing by Age and Gender (ms) 5.3.1.1 Isolated Words Results for allophonic variation in isolated forms show a similar pattern to that found in adult speech: a relatively long positive lag for final, a negative lag for initial, and a short intermediate lag for intervocalic allophones. A total of 233 tokens in isolated form (final: n = 89, initial: n = 60, intervocalic: n = 84) were analyzed. A N O V A results for the allophones in isolated form showed that the effect of syllable position is significant (F (2, 230) = 47.89, p < .0001). Means of intergestural timing (with standard deviations in parentheses) for allophones in isolated form were 55.4 ms (58.96), -23.87 ms (48.69), and 7.5 110 ms (39.15) for final, initial, and intervocalic positions, respectively. Fisher's PLSD post-hoc test also indicates that the mean differences between the final and initial (79.2 ms, p < .0001), between the final and intervocalic (47.8 ms, p < .0001), and between the initial and intervocalic (-31.4 ms, p = .0002) were all highly significant. 5.3.1.2 Two-word Combination A total of 202 tokens in combination were analyzed: VL#C: n = 63, VL#V: n = 88, V#LV: n = 51. The ambisyllabic allophones (VL#V) behaved more similarly to the initial (V#LV) than to the final (VL#C) allophones. A N O V A results for the ambisyllabic allophones showed that the effect of syllable position is significant (F (2, 199) = 7.41, p = .0008). Means of intergestural timing (with standard deviations in parentheses) for allophones in two-word combinations were 13.06 ms (38.87), 24.6 ms (42.39), and 44.4 ms (51.1) for initial, ambisyllabic, and final positions, respectively. In other words, the ambisyllabic allophones (VL#V) showed an intermediate timing relation between the gestures, as in adult speech. Fisher's PLSD post-hoc test also indicates that while the mean differences between final and ambisyllabic allophones (19.8 ms, p = .0076) and between final and initial allophones (31.3 ms, p = .0002) were significant, the mean difference between initial and ambisyllabic allophones were not significant (11.5 ms, p >.05). Means of intergestural timing in both isolated and ambisyllabic allophones combined with all tokens by all children are tabularized in table 5-1. 111 Table 5-1. Mean of Intergestural Timing # initial initial ambisyllabic final final # intervocalic Mean -24 - 14 - ' 25 44 55 8 SD 48.8 38.9 42.4 51.1 60 0 39.1 n 51 88 63 90 84 Note: Mean, shown in milliseconds, SD (standard derivation), n (number of tokens) Figure 5-3, compares the timing relation between the gestures of IM in isolated forms and two-word combinations. While the final allophones (VL#C) patterned similarly to the word-final allophones (VL#), the initial allophones in isolation (#LV) and the two-word combinations (V#LV) showed opposite timing patterns. For instance, in syllable-initial position, the TD reached its peak before the TT reached its maximum constriction (14 ms); but in word-initially, the TT reached its peak before the TD reached its maximum constriction, causing negative relative timing (-24 ms) (see also table 5-1). That is, although the relative timing in the both allophones was considered as simultaneous movement (less than 33 milliseconds), the gestures were coordinated in a different order. 112 # initial initial ambisyl final final # intervoc Figure 5-3. Intergestural timing (TT-TD): Isolated vs. Two-word Combination (ms): Allophones of IXI in isolated form are shown in grey and allophones of IXI in two-word combination are shown in black. 5.3.2 Individual Results The relative timing in syllable position by each child also suggests a similar pattern to the overall result: a positive lag appeared in the postvocalic position, but almost simultaneous timing between the gestures (either negative or positive) appeared in the pre- and intervocalic positions. A long positive lag in the postvocalic position was found in all children except for one; a negative lag was found in the prevocalic position for five children; in six children, a lag in the intervocalic position was relatively longer (a shorter negative lag) than its prevocalic counterparts. A N O V A results for the mean values of lag by syllable position and subject showed that the overall effects of syllable position (F (2, 411) = 33.08, p < .0001) and 113 of subject (F (7,411) = 7.18, p < .0001) are highly significant. The two-tailed t test (p < .05) also indicates that the intergestural timing between gestures (TT-TD) was significantly different in six children from a null hypothesized mean of 0 (df (434) = 8.73, p < .0001), but not significant in two children (KR, LR) (p > .05). H i final |. \ initial I' *| intervoc JN KR AD FL RC LR MC JL Figure 5-4. Intergestural timing (TT-TD): All children (ms) With respect to relative timing, results for each child suggest two distinct patterns. Four children used timing to distinguish allophones in a way similar to what adults do (JL, FL, KR, RC), and the other four children used a single dominant timing for most allophones (JN, LR, M C , AD). As for allophonic variation, the former four children as well as JN distinguished allophones by timing; and three children in the latter group did not distinguish allophones by timing for most tokens. Overall means and number of tokens for each syllable category for each child are tabulated in table 5-2. A complete list of statistics for each child is in Appendix D. 114 Table 5-2. Distribution Chart: Intergestural Timing (TT-TD) in Six Syllable Positions RC(F:3;11) '$0s AD (M: 4;4) JL (F: 4:7) FL (M: 4;9) JN(F- 4.10) JN(F: 5;1) LR (F. 5,6) MC(F: 5;6) KR (F. 5.9) M SD N M SD n M SD n M SD n M SD n M SD N M SD n M SD n M n # Initial -18 53.7 ' 13 17 45.9 6 -33 " ' 64.5 9 -31 41.5 17 80 50.5 5 -44 38.4 3 -14 26.2 7 n/a n/a n/a -42 568 4 V-Initial a a n/a n/a n/a n/a n/a -4 • 45.1 8 11 37.2 9 109 37.0 7 42 36.7 11 -33 33.3 3 21 30.5 8 6' 27.8 12 Final-V 0.0 . * 0.0 i n/a n/a n/a 44 38.5 19 10 .20.9 13 "171 33.8 7-- 100 38.4 7 60 13.4 11. 36 42.2 14 -3 29.6 21 Final-C 11 19.2- 3 n/a n/a n/a 17 : 50.5 6 33 40.0 19 147 53.9 7 93 29.2 18 n/a n/a n/a 33 47.0 2 20 54.5 15 Final # 49 62.6 26 52 26.2 7 166 47.0 2 81 40.2 21 3d 53.1 10 78 55.2 9 -2 53.2 17 50 23.5 2 72 49.0 6 Intervoc 10 16.0 10 30 30.9 9 12 305 8 -9 15.5 11 ' 28 80.2 7 50 75.5 8 60 13.4 11 -7 14.0 10 -8 47.8 17.. Total 435 55 22 * 52 90 43 56 49 36 75 ' Note: M (mean, shown in milliseconds), SD (standard derivation), n (number of tokens). L R and M C are twins. JN (4;11) and JN (5;1) are the same child. Four children showed adult-like relative timing between the gestures for at least some allophones. JL(4;7),F While JL had the most adult-like tongue dorsum backing gesture with a fully stretched tongue, she had a longer positive lag for the ambisyllabic allophone (VL#V) than the final allophone (VL#C). She also had a very long positive lag for the isolated final IM. This indicates that the ambisyllabic allophone patterns similarly to the final allophone, and did not go through resyllabification. Results for JL showed a great positive lag in the final allophone in isolated words (166.5 ms); however, neither the final (17 ms) nor ambisyllabic (44 ms) allophones in two-word combinations had a long positive lag. Note that the ambisyllabic allophone (VL#V) had a longer lag than the final allophone (VL#C). On the other hand, the initial allophone in isolated words (-33 ms) had a short negative lag, and the intergestural timing in the intervocalic allophone (12 ms) was near synchronicity. The initial allophone in the two-word combinations (V#LV) and isolated words patterned similarly. A total of 52 tokens were analyzed. FL (4;9), M According to the results, F L ' s speech was faster than the rest of the subjects and had the most adult-like allophonic distinction. A total of 90 tokens were analyzed. Timing pattern was similar to that of JL. A N O V A results showed that the effect of syllable position is significant (F (5, 84) = 21.39, p < .0001). Results for FL also showed a great lag in the final (81 ms) and initial (-31 ms) allophones in isolated words. The intergestural timing in the 116 intervocalic (-9 ms), ambisyllabic (10 ms), and initial (11ms) allophones was near synchronicity. This suggests that the ambisyllabic allophone is resyllabified and behaves similar to the initial allophone. The final allophone in two-word combination had a short positive lag (33 ms). #initial initial ambisyll final f inal* intervoc # initial initial ambisyll final f inal* intervoc FL JL Figure 5-5. Intergestural Timing (TT-TD): JL and FL KR (5; 9), F K R was the oldest participant. Results for K R (5;9, F) seemed to show a similar pattern in timing relation between the gestures to that in adults. However, the two-tailed / test showed that the mean difference was not significantly different from a null hypothesized mean of 0. This indicates that in her speech, the relative timing is not distinct in syllable position. A total of 75 tokens were analyzed. A N O V A results showed that the effect of syllable position is significant (F (5, 69) = 4.73, p = .0009). Results for K R showed a great positive lag in the isolated final allophone (72 ms), similar to that of FL (81 ms), and a relatively long negative lag (-42 ms) in the isolated initial allophone. However, the relative 117 timing in all four allophones in two-word combination was not distinguishable. For example, the timing between the gestures was very short, although the intervocalic (-8 ms) and ambisyllabic (-3 ms) had negative timing, while the initial (6 ms) and final (20 ms) allophones had positive timing. RC(3;11), F RC was the youngest among the participants, yet her intergestural timing showed a similar pattern (adult-like relative timing) to that found in the older children in her section. A total of 55 tokens were analyzed. A N O V A results showed that the effect of syllable position is significant (F (4, 50) = 3.9, p = .0078). Five allophones, except the initial (V#LV) allophone, were collected.1 Results for RC also showed a positive lag in the isolated final allophone (49 ms), although shorter than that of the other three children mentioned above. The relative timing among the isolated initial allophone (-18 ms), final (11 ms), and intervocalic (10 ms) allophones was not distinct. The intergestural timing in ambisyllabic allophone was simultaneous (0 ms). Fisher's PLSD post-hoc tests showed that only the mean difference between the isolated initial and final allophones was significant (p = .0004), and the mean difference between all other allophones was not significant (p > .05). 1 RC could not identity the letter "A" in the picture book and could not do v-initial IM sequence (e.g., "A # ladybug"). The other word used for this target was "bee" as in "bee # leaf or "bee # lips," which she did not do. 118 # initial initial a m b i s y | | f i n a | r m a * * intervoc KR # initial ambisyll final RC Figure 5-6. Intergestural Timing (TT-TD): KR and RC final # intervoc Three children (JN, M C , AD) had a tendency to have a long positive lag across allophones; and one child (LR) mostly had a very short or negative lag across allophones. Results for LR and M C are shown together in order to compare them as twins. JN(4;11, 5,1), F JN participated in two sessions. One distinction between JN and the other children was the darkness of prevocalic allophones: a extremely long positive lag was found in all allophones, except for the isolated final and intervocalic allophones at 4;11; and still a great positive lag in all allophones, except for the initial allophones (#LV, V#LV) at 5;1. Between 4;11 and 5;1, the relative timing range for the initial and final allophones has dramatically changed and fits into the correct timing range for each allophone. Nevertheless, in terms of relative timing, all allophones patterned similarly as to a final allophone. This indicates that JN does not distinguish allophones by timing. 119 Figure 5-7. Intergestural Timing (TT-TD): JN (4; 11 and 5;1) LR(5;6) andMC (5;6), F LR and M C are twins. However, results for relative timing for allophones showed opposite patterns between them. M C used a positive intergestural timing for most allophones (except for intervocalic), whereas LR used a negative intergestural timing or very short positive lag across allophones. Interestingly, both tended not to differentiate across allophones. A total of 36 tokens were analyzed for M C and 49 for LR. A N O V A results for LR showed that the effect of syllable position is not significant (F (4, 44) = 1.07, p > .05). The initial allophone (V#LV) had a greater negative lag than that of the isolated counterpart (#LV). No final allophone (VL#C) was analyzed. A N O V A results for M C showed that the effect of syllable position is significant (F (4, 31) = 2.85. 2 LR added "and" between two words; hence target words in /final # c/ were produced in /final # v/ form instead (e.g., "bell bed" was produced as "bell and bed"). 120 p = .039); however, timing did not distinguish allophonic variation. A l l allophones except the intervocalic showed a similar positive intergestural timing. Due to poor images, no isolated initial allophone was analyzed. LR MC Figure 5-8. Intergestural Timing (TT-TD): LR and M C AD (4;4), M A D showed somewhat different gestural patterns from the rest of the children in the experiment. The relative timing was not significant; however, since he only produced IM in isolated words, it is impossible to compare intergestural timing in syllable position with that of the other children. Also note that although he had a positive lag in word-initial, -final, and word-internal positions, the gestures were not coordinated in the same way as in the other children's speech. In most of his utterances, A D had a tongue dorsum fronting instead of backing; and a positive lag occurs when the movement of the tongue dorsum precedes the tongue tip movement. A D showed the least differentiated motor coordination among all the subjects, which was reflected in the temporal and spatial characteristics of IM in his speech. 121 A total of 22 tokens were analyzed. A N O V A results for A D showed no effect of syllable position (F (2,19) = 1.84, p > .05). A l l allophones showed a positive lag: initial (17 ms), final (52 ms), and intervocalic (30 ms), respectively. Fisher's PLSD post-hoc tests showed no significant differences in the mean intergestural timing between allophones (p>.05). 100 -i < < • 80 - T 60 - T T 40 - ' ' •' 20 - I • . ' 0 - ' — ' ' — — ' ' — ' --20 --40 --60 --80 -I . • • # initial final # intervoc Figure 5-9. Intergestural Timing (TT-TD): AD 5.4 Discussion Two general observations are made: first, across subjects, relative timing between the gestures in children's IM shows significant patterns similar to that in adults' IM; second, isolated words and two-word combinations seem to have distinct patterns. To begin with, overall results of relative timing (TT-TD) reflect the pattern found in adults' IM: a long positive lag (TD precedes TT) appears in the final position, whereas a short positive or a negative lag (TT preceded TD) occurs in the initial and intervocalic positions. This shows 122 that, taken as a whole, children at this age are successfully approximating adult-like timing, despite a great deal of variation. However, when examining individual results, different tendencies are found regarding the temporal characteristics of IM. While the relative timing in four children suggested resyllabification of the ambisyllabic allophone (VL#V), in the other four, it did not. It is not clear whether this distinction is age-dependent. It also is interesting to note that children in the latter group were more likely to use a greater positive lag in both pre- and post-vocalic allophones compared to children in the former group, except for one child (LR). With a relatively long lag in prevocalic allophones, as well as in the postvocalic position, allophonic distinction by relative timing may not be possible, and IM may sound like dark IM across syllable positions. The findings in this chapter support the PMD hypothesis that children at these ages are undergoing the refinement process, both across and within subjects; whereas, within subjects, they still are showing evidence of the processes of integration and differentiation. A l l of the following transpire: an adult-like pattern occurs, but with over- or under-sized lag (refinement); relative timing distinguishes only some allophones, and/or one dominant timing pattern (positive lag) is present across syllable positions (integration); and timing relation is not distinct across syllable positions (differentiation). Second, while relative timing distinguishes allophones in isolated words, IM in prevocalic (V#LV) and postvocalic (VL#C, VL#V) positions in two-word combinations shows no difference in timing, even for the children with the refinement process. This suggests that ambisyllabic allophones behave differently from allophones in isolated words. Although no statistically significant difference has been found, the ambisyllabic allophone 123 tends to be resyllabified and patterns similarly to the initial (V#LV), rather than its postvocalic counterpart (VL#C). (For a study of ^syllabification in adult IM, see Gick, 2003.) This indicates that relative timing differentiates IM in isolated words from ambisyllabic allophones, but still needs to be refined to distinguish resyllabification of allophones across word boundaries. Thus, children at these ages use a similar timing pattern of III to that found in adult speech (Sproat & Fujimura, 1993; Browman & Goldstein, 1995; Gick et al., in press): a positive lag (TD backing occurs earlier than TT raising) in postvocalic allophones, a negative lag (TT raising occurs earlier than TD backing) in prevocalic allophones, and an intermediate lag in intervocalic allophones. According to a recent study of the relative timing in the allophones of IM, Gick et al. (in press) found that in Western Canadian English a great negative lag occurs in the prevocalic position (-58 ms), a relatively short positive lag in the postvocalic position (18 ms), and an intermediate lag (< 10 ms) in the ambisyllabic position. In the present study, a relatively short positive lag for the prevocalic (14 ms), a great positive lag for the postvocalic (44 ms), and the intermediate lag for the ambisyllabic (25 ms) allophones were found in two-word combination. A similar pattern is also observed in the isolated allophones. This indicates that the relative timing for the syllable-based allophones in children is similar to that of American English speaking adults, not Western Canadian English speakers. However, direct comparison between adult's and children's IM in speech needs further research. 5.5 Conclusion 124 In this chapter, I have investigated the temporal characteristics of IM in children aged 3; 11 to 5;9 to observe whether children use relative timing between gestures similar to that found in adult speech. Results showed that adult-like relative timing patterns occur significantly across children. However, relative timing across syllable position within children shows more variation, reflecting the multiple ongoing processes of motor development. This study further suggests that relative timing is more distinct in IM in isolation, than in two-word combinations: a long positive lag in word-final position, a negative lag in word-initial, and a very short lag in intervocalic were found. However, the relative timing in resyllabified allophones does not show a consistent pattern across children, suggesting that intergestural timing is not consistently used to distinguish allophonic variation. Both across and within subjects, children at these ages show evidence of being in the refinement process, since they can produce IM timing patterns qualitatively similar to that in adult speech, although with too large or too small lag values. Also, within subjects, children tend to use a dominant timing pattern across allophones—as expected in the integration process—or in some cases, show inappropriate simultaneity as they are in the differentiation process. Thus, as predicted in the P M D hypothesis, all three types of processes of motor development appear to be active in this age group. 125 CHAPTER 6 ACOUSTIC AND PERCEPTUAL EVALUATION OF IM 6.1 Introduction This chapter discusses acoustic information of children's IV, and perceptual judgements by adult listeners. Previously in chapters 4 and 5, we have seen that the articulatory gestures of IM of children ages 4 and 5 are spatially and temporally coordinated differently than in adults. Spatially, the tongue shape and direction of the movement in the speech production of III varied, yet the variation reflected three motor development processes (Differentiation, Integration, and Refinement). Temporally, the intergestural timing differed in syllable-based allophones in most children; however, the direction and range of the timing lag were varied across subjects, and not consistent across syllable positions, which also mirrored the three motor development processes. While this dissertation takes a primarily articulatory approach to investigating the speech development of IM, and has shown that the articulatory characteristics of IM have distinct patterns in the process of language development, it also is important to see how the acoustic counterparts of/1/ articulations are related to this study's findings. This chapter will attempt to determine whether the production of III by the children in the study varies acoustically across subjects and syllable positions, as is the case for articulation, or whether children show more consistent acoustic patterns similar to adults' III. 126 In addition, perceptual judgments by phonetically trained adults are presented to see how the children's speech production of IM is perceived by adults. It is important to observe how the children's actual speech production of IM would be perceived and transcribed by adults, especially i f it. was not produced close to the IM of adults. However, note that the acoustic and perceptual evaluations in this chapter are post-experiment and provide only limited information based on the analyses of a random sample of tokens. In the following sections, the formant and spectrographic analysis for the acoustic study is presented in § 6.2, and perception judgments by adults follow in § 6.3; discussion is laid out in § 6.4, and a conclusion is presented in § 6.5. 6.2 Acoustic Analysis The "lateral / l / " is a sound produced with the tip of the tongue placed on the alveolar ridge, while the side(s) of the tongue is (are) left in open position. Thus, air travels from the glottis to the lips passing around the lateral edge(s), while the centre airway is blocked in the midline by the tongue tip closure. This causes splitting of the airway, especially in the prevocalic position, which renders distinct formant frequencies for the second (F2) and third (F3) formants (Stevens, 2000). The formant structure of IM shows similarities to that of the glide Iwl in that both FI and F2 have very low intensity, while the bandwidth of FI is broad (Ladefoged, 1993, 2001; Stevens, 2000). However, the constriction of IM is much shorter than t that of the glides, as is indicated by a slightly higher FI frequency than that for a high vowel, but lower than that of glides (Stevens, 2000, p. 533). A vowel-like low F2 also is an indication of the dorsal gesture of IM (Walsh-Dickey, 1997, p. 51). While FI and F2 of 127 English IV are relatively low, F3 has a relatively high frequency with strong amplitude, resulting in a wide gap between F2 and F3 (Ladefoged & Maddieson, 1996; Stevens, 2000). In addition to formants, IV is characterized with an anti-resonance (zero in the spectrum) caused by the airspace under the tongue. Although a zero for the lateral configuration depends on the vowel environment and on the individual speaker, the lowest zero can be in the frequency range of 2200 to 4400 Hz. (see Stenvens, 2000). 5000T ; : 1 1 . 40001 Time (s) 'whale' 'leg' Figure 6-1. Spectrograms of Initial and Final IV: Adult Speaker (JN's mother) Figure 6-1 shows that, for this speaker's IV in prevocalic position, the formants F2 and F3 could move directly into the adjacent vowel without any clear frequency variations. 128 but IM shows less energy than the vowel. There also is a spectral discontinuity when the tongue tip touches the alveolar ridge. In postvocalic position, it shows as an abrupt change in formant transition for F2, which is signalled by the down slope of F2 with FI near 1000 Hz, and with F3 simultaneously moving up toward 3000 Hz, separating F2 from F3. For children's speech, a recent master's thesis by Gormley (2003) tested the formant frequencies of FI and F2 of the initial III in an English-German bilingual child (3;3), and results showed higher frequencies for both FI and F2 than in adults. Children's speech is less accurate than that of adults in general, and even in adults, great variability—in terms of individual differences and gender—can be expected in the lateral configuration (Stevens, 2000). However, the formants frequencies of III in male and female adults and children are roughly in the frequency range shown in table 6-1. Table 6-1. Formant Frequencies (Hz) of IM by Adults and Child (prevocalic) FI F2 F3 F4 Adult (male) Ladefoged(1993) 250 1200 2400 3200 Stevens (2000) 360 900 2800 3400-3900 Adult (female) Stevens (2000) 350 1180 — TN's mother 380 1000 1800 3500 Child* (male) Gormley (2003) 530 1650 *A child is German-English bilingual (3;3). 129 Following are the formant analysis of children's IM from the sample tokens. 6.2.1 Methods Tokens were randomly selected from the main experiment for three syllable positions. As mentioned in the methodology in chapter 3, eight children who participated in the study were pre-screened for their production of IM, and those who produced adult-like IM by ear— consistently without substitutions (except for one child, RC) as confirmed by their parents— were chosen for the study. For the acoustic analysis, both IM words checked and perceived as IM by naive adult listeners before measurement as well as tokens with unclear pronunciation of IM, which were previously excluded from the articulatory measurement, were included. For comparison, the tokens were chosen as closely as possible across subjects. The list of words used for the analysis is shown in table 6-2. Usually, three formants are required to describe liquids (Borden, Harrison, & Raphael, 1994). The formant frequencies of FI , F2, and F3 of children's IM in the sample tokens were analyzed at 50% of the duration of IM. 130 Table 6-2. Word List JL (4;7) FL (4;9) JN (5;1) LR (5;6) MC (5;6) KR (5;9) AD (4;4) RC (3;11) wheel V V V V V V whale V 1 tail V V V V snail V bell V V V V V. V V lip(s) V V V V V V V 1 ladybug V V V V V A/ 1 lemon V. ladder V 1 leg V yellow V v . V V V V V l[w]eash V Total 12 7 9 7 7 9 7 8(1)* *[w] substitution is counted separately. 6.2.2 Results The results of the means of three formant frequencies of IM in children are shown in table 6-3. A total of 66 tokens was measured. 131 Table 6-3. Mean of Formant Frequencies: FI, F2, and F3 of IM FI F2 F3 JL (4;7),F 789 1356 3876 FL (4;9), M 599 1740 3782 JN(5;1),F 623 1598 4094 LR (5;6),F 673 1725 3459 M C (5;6), F 643 1416 3874 K R (5;9),F 694 1356 3734 A D (4;4), M 722 1685 3639 RC (3;11),F 717 1396 3510 Average 685 Hz 1519 Hz 3757 Hz In general, the formant frequencies of children's IV were higher than those in adults, but in a similar range with that found in younger children (Gormley, 2003), as in table 6-1. As for the high frequencies, both F2 and F3 were in the range of 1500 to 4000 Hz (F2) and of 2500-4000 Hz (F3) that identifies the lateral configuration in adult speech (Stevens, 2000). In RC (3;11), the formants of /w/ in the glide substituted "leash [wish]" showed lower frequencies than in III: FI was about 550 Hz; F2 was about 1200 Hz; and F3 was about 2900 Hz. This separates [w] from IV in terms of the formant frequency. Figure 6-2 shows the distribution of the frequencies of FI, F2, and F3 of /!/ in these children. 132 Count Count Count ere c » I • o ore " i ss 3 o -«1 o 3 as a a" e SB N 6.2.2.1 Age There was no age effect in the formant frequencies (F (2, 63) = .877, p > .05), although the low-frequency of FI decreased by age (e.g., 730 Hz (3 yr.), 700 Hz (4 yr.), 660 Hz (5 yr.), and the high-frequency of F2 and F3 increased by age (e.g., F2: 1400 Hz (3 yr.), 1500 Hz (4-5 yrs.), F3: 3500 Hz (3 yr.), 3800 Hz (4-5 yrs.)). 6.2.2.2 Gender While no significant difference was found between male and female children for the frequencies of FI or F3, the frequency of F2 was significantly different between them (F( l ,64) = 4.242,p<.05). 134 2500 2250 -2000 -male female FI F2 F3 male 630 1710 3710 female 700 1470 3770 Figure 6-4. Frequency of F2 by Gender 6.2.2.3 Individual Results Examples of the spectrograms of initial and final IV from the sample tokens are presented in this section. 135 JL (4;7), F The spectrograms show a sudden change in the frequency of F2 for both initial and final IM. A long duration of IM in the postvocalic position is shown in "whale." In "leg," F3 is not clearly shown. SOOOr 40001 30001 20001 10001 t -§990, 4 Tme(s) 'whale' Q77 Time® 1.05624 TITB(S) 'leg' Figure 6-5. Spectrograms of IV: JL 136 FL (4; 9), M The spectrograms show the formant transition of F2 in both initial and final IV, but a more abrupt change in F2 is found with the initial IM word "a leaf." The frequency of F3 is consistent and high. 5000 4000 3000 2000 1000 059465 0.66T44 Time (s) Time (s) 'a tail ' 'a leg' Figure 6-6. Spectrograms of IV: F L 137 JN(5;1), F Her speech was relatively slower than the other children's in the experiment. It was also noticed that her initial allophones sounded like dark IV and that final allophones were exaggerated. The spectrograms show a very high frequency of F3 in both initial and final IV. 5000 4000 3000 2000 1000 w I . . . .! 2.09846 Time (s) 'a whale' ' J leg' Figure 6-7. Spectrograms of HI: J N 138 LR (5:6) F In LR, the formant transitions of the postvocalic 'tail' form a rather smooth slope for F2, with a much widened bandwidth of FI . The prevocalic "ladybug" shows extra resonance above F2 and an abrupt slope change of F2. 5000 4000J, 3000 2000^ 100QJ, 2.13388 T i m e (s) ' tail ' 'ladybug' Figure 6 - 8 . Spectrograms of /I/: L R 139 MC (5;6), F The spectrograms of IM show formant transitions of F2 similar to those found in other children. A broad bandwidth of FI is observed in MC' s IM. 5000 4000 3000H 2000 1000 1.83725 T i m e (s) ' tail ' 'ladybug' Figure 6-9. Spectrograms of IM: M C 140 KR (5;9), F The formant frequency of F2 in both prevocalic and postvocalic IM shows similar transitions with a stable high frequency of F3. 5000 4000J 3000 20001 1000 1.611367 T i m e (s) 'whale' 'leg' Figure 6-10. Spectrograms of IM: KR 141 AD (4:4), M While AD's spectrograms compared to the other children's show similar formant transitions of F2 for the postvocalic IM, F2 in "whale" merges with FI , leaving a possibility of syllabic [1,]. The frequency of F3 in pre- and postvocalic IM is not clear. Figure 6-11. Spectrograms of IV: AD 142 RC(3;11), F As seen in A D , a similar merge occurs between F2 and FI in the postvocalic IM of "whale." The prevocalic IM in "leg" in figure 6-12 is also compared with [w] substituted "leaf." in figure 6-13. The spectrogram of "l[w]eaf shows a great slope change for F2 with a shorter duration than that of /!/, whereas the frequency transition of F2 in "leg" is not as steep as the former. 5000 4000 3000 20001 1000 1.23964 Time (s) whale' 'leg' Figure 6-12. Spectrograms of /!/: R C 143 : : T i m e (s) 'lMeaf] 0.76T571 Figure 6-13. [w] Substitution for Initial IM: R C The spectrographic analysis of IM is somewhat difficult, even in adults' speech, as the formant frequencies of IM often are confused with glides or back vowels. Nevertheless, in these samples, the formant patterns in the prevocalic and postvocalic IM largely fall under the range of lateral configuration. This implies that the production of IM in these children is acoustically grouped in the same category. For more accurate quantitative analysis of the acoustic characteristics of IM in speech development, future study will be required 144 6.3 Perceptual Judgment 6.3.1 Methods Perceptual judgments by three adults were made to find out how IV in these children is impressionistically perceived by adult listeners. A l l three adults were female English speakers, who had been trained in phonetic transcription: a linguistics graduate student (Al ) , an acoustic phonetician (A2), and a speech-language pathologist (A3). The former two had no experience in transcribing children's utterances; the latter had exclusive experience working with the deaf and hard-of-hearing. They were asked to listen to each child's audio file consisting of 14-17 tokens collected from the experiment, including the initial, intervocalic, and final IV. The list of words was made as similar as possible across the children. Both words in isolation and two-word combinations were used. The tasks were two-fold: first, they were asked to write down all the words they heard; second, when that was done, they were asked to listen to the files again and transcribe the words that contain IV. They were asked to describe whether IV sounded like light [1], dark [1], or other sounds, using narrow transcription. 6.3.2 Results In general, all three of them mentioned that the perception judgments of child utterances were difficult, and they were not confident about their evaluations. However, the initial IV received more consistent judgments from all adult listeners than the final allophones. 145 The final IM in several children's utterances were often perceived as glide [w] or back vowels. The intervocalic IM was perceived mostly the same as the initial allophone. Some words were not recognized by the adults. Results of the percentage of the initial, final, and intervocalic allophones perceived as light 1, dark [fj, or other sounds are presented in table 6-4. A complete list of the adult's perception judgments for each child is provided in Appendix B. 146 Table 6-4. Perceptual Judgment of /!/ by Adult Listeners A l A2 A3 light dark other light dark other light dark other JL Initial (5) 100 0 0 100 0 0 80 o 20 (4;7) Intervocalic (4) 50 50 0 100 0 0 100 0 0 Final (6) 0 16.6 83.4 0 100 0 0 16.6 83.4 FL Initial (6) 100 0 0 100 0 0 100 0 0 (4;9) Intervocalic (4) 75 25 0 100 0 0 100 ...;.',a.;,; 0 Final (6) 0 100 0 0 83.4 16.6 0 100 0 JN • Initial (5) 100 0 0 100 0 0 100 0 0 (5;i) Intervocalic (3) 100 00 0 100 0 0 100 0 . 0 Final (7) 85.7 14.3 0 0 71.4 28.6 0 71.4 28.6 LR Initial (4) 100 0 0 100 0 0 100 0 0 (5;6) Intervocalic (4) 100 . 0 0 100 0 0 100 0 0 Final (6) 0 100 0 16.6 83.4 0 0 66.8 33.2 M C Initial (5) 100 0 0 100 0 0 100 0 0 (5;6) Intervocalic (4) 100 0 0 75 0 25 7 5 0 25 Final(6) 0 100 0 33.3 33.3 33.3 0 83.4 16.6 K R Initial (6) 83.4 16.6 0 50 50 0 100 0 0 (5;9) Intervocalic (4) 75 0 25 50 25 25 100 0 o Final (7) 0 28.6 71.4 0 100 0 0 57.1 42.9 A D Initial (6) 100 0 0 100 0 0 100 0 o (4;4) Intervocalic (3) 33.3 33.3 33.3 66.6 0 33.3 100 0 0 Final (6) 16.6 40.6 42.8 0 50 50 0 16.6 83,4 RC Initial (5/1') 80 0 20 80 0 20 80 0 20 (3;11) Intervocalic (4) 50 25 25 50 25 25 50 o 50 Final(5) 20 0 80 0 40 60 : : : o :. 20 80 ' [w] substituted "leaf was not included in the table. All three listeners perceived and transcribed it as [w]. 147 JL(4;7),F A l l the adults perceived the initial IM well. A2 expressed that some of JL's initial /1/s sounded dark [i], although she did not transcribe it as [i]. Two adults perceived JL's final IM as vowel-like sounds ( A l , A3), while the other (A2) described it as dark I. The intervocalic IM was perceived mostly as light IM. Some of the initial 1-words were not recognized correctly by the adults: leg was transcribed as "lake" (Al ) , and leaf was transcribed as "leap" ( A l ) or "lake" (A2) or not recognized at all (A3). FL (4;9), M The perception judgments showed that all three adults perceived most of FL 's IM as an adult-like IM, and the intervocalic IM patterned with the initial IM. JN(5;1), F While JN's initial and intervocalic IM was perceived 100 percent as light 1 by all three adults, the split results came for the final IM, as one adult (Al ) perceived it as light 1, suggesting that JN's IM sounded similar across syllable positions; while the other two perceived it mostly as dark [t]. LR (5:6). F A l l three adults mentioned LR's IM sounded like an adult's IM. They perceived the initial and intervocalic IM 100 percent as light 1, and most of final IM as a dark 1. A3 did not recognized the word wheel. 148 MC (5; 6) F While all three adults perceived the initial IM well, the intervocalic IM in jellybeans was transcribed as [w] by two (A2 and A3). For the final IM, A l and A3 perceived it as dark 1, and A2 transcribed it as either light, dark, or other sounds. KR (5:9), F K R is the oldest child among the group. Two adults ( A l and A2) perceived some of the initial 1-words as dark 1, some of the intervocalic 1-words as dark or other, and A3 perceived both initial and intervocalic IM as light 1. The perception of the final IM also caused splitting results: while A2 perceived it as dark 1, A l perceived most of the final 1-words as a vowel, as did A3 for the half of them. AD (4;4), M The adults' perception judgments also noted some nasalized /1/s and that AD ' s final IM sounded like a vowel. A l l three perceived all initial 1-words as light 1, as well as most of the intervocalic 1-words, although two listeners ( A l and A2) transcribed IM in jellybeans as [j] or [w]. AD's final IV was perceived by all the adults as either dark 1 or [w] or [ou]. Some final 1-words were not recognized correctly by the adults: whale was perceived as "real" (A l and A2) or "wheel" (A3); snail was perceived as "nail" ( A l and A3) or "mail" (A2); and tail was perceived as "pail" by one adult (A3). 2 In the main experiment, the tokens with unclear pronunciation of /I/ such as "tail," "whale," and "snail" were excluded from the analysis. 149 RC(3;11), F While all three adults perceived the initial IM as light 1, ladybug was not perceived clearly. RC had glide substitution for " leaf [wif], but other initial 1-words were produced as IM and perceived as [1]. The intervocalic IM showed mixed results, and most of the final 1-words were perceived as a vowel [o] by all three adults. Also, IM in whale was perceived as "wheel" by two adults ( A l and A2): one described it as light 1 and the other described it as [w]. A l l of them noticed [w] substitution for "leaf." 6.4 Discussion Generally, identification of IM is more difficult than other approximants. A perception study of three-formant synthetic patterns for IM, Ixl, /w/ and/j/ in the intervocalic position (V_V) with the vowels IM, /a/, and lui showed that the lowest percent of listeners recognize IM, while Iwl gets the highest percent of identification, followed by Ixl and /j/ (Lisker, 1957, cited in Borden et al., 1994). While /w/ and IM share similar formant patterns, the frequency of F2 of/w/ is lower than that of IM. Between the liquids IM and Ixl, F3 differentiates one from the other, since the frequency of F3 of Ixl is lower than that of IM, while both have F2 in the same range (Borden etal., 1994; Stevens, 2000). Some of the final IM in children's speech was perceived as a back vowel [o] or a glide [w] by adult listeners, which, however, should have lower first- and second-formant frequencies than IM. Also, some judgments made by these adults were not consistent within the category. Nevertheless, the perceptual judgments have provided useful information showing that the allophones of IM sound different by syllable position, and that the final allophones of /!/ in children's speech are more variable than the 150 initial or intervocalic ones. This suggests that children coordinate the articulatory gestures of IM differently for these syllable-based allophones, although each child may show individual variation. The formant study of children's IM also shows relatively consistent patterns of formant frequencies and transitions of IM across children. Findings indicate that the absolute value of formants in children is relatively higher than in adults. This was first noted in the original vowel study of Peterson and Barney (1952), although they did not mention how to define the characteristics in children that are distinct from that of adults. In a recent study of speech development, Callan, Kent, Guenther, and Vorperian (2000) found the correlation between the developmental change in the size and shape of the vocal tract and auditory feedback in speech production. It is predicted by the source-filter theory of speech: small size of the vocal tract in children leads to higher resonant frequencies, and hence to higher formant frequencies. While the acoustic findings of IM in children are distinct from that of adults in terms of the absolute value of formants, the formant transitions and frequencies of IM in these children fall under the same range of laterals. This indicates that the articulatory characteristics of IM in the production of children ages 4 to 5 may be more varied than the acoustic characteristics of IM, and that adult-like motor coordination skills may take longer to acquire than acceptable acoustic production. 6.5 Conclusion This chapter has presented the acoustic analysis of three formant frequencies of IM, and the perceptual judgment of IM by adult listeners. Overall, all children showed /l/-like 151 formant patterns for both initial and final IM, with a noticeable formant transition in F2 while F3 stayed in a high frequency range. The formant frequencies of IM were higher than those in adults, but similar to those found in one bilingual child in a previous study (Gormley, 2003). Adult listeners also recognized most of the 1-words in these children; however, quite a few instances of final IM were likely to be perceived as vowels. 152 CHAPTER 7 C O N C L U S I O N 7.1 Summary In this dissertation, I have examined the spatial and temporal characteristics of the articulatory gestures of English III as produced by children aged 3; 11 to 5;9 across different syllable positions. Based on previous articulatory studies of IM in adult speech, the following questions have been tested: (a) what are the gestures used, and what are their spatial and temporal characteristics?; (b) how do these gestures vary in magnitude and timing across syllable positions?; and (c) what process(es) of speech motor development is/are present in the speech acquisition of III? The findings are: 1. Spatial generalizations: a) Two lingual constrictions, tongue tip and tongue dorsum, are generally used in the children's articulation of III, but in terms of spatio-temporal organization, they appear with both qualitative and quantitative variation compared to that in adult speech. b) Tongue tip closure often is missing: fronting or lowering of the tongue tip often is present. c) Tongue dorsum backing is not as dramatic as that in adult speech. 153 2. Timing generalizations: a) While all of the children showed a variety of distinct patterns in respect to allophonic distinction, four out of eight distinguished allophones of IM relatively consistently by using qualitatively adult-like intergestural timing; the other four only sporadically distinguished allophones using timing. b) Most children showed an intergestural timing lag of some kind across most allophones of IM, although the direction of the lag often was not the same—relative to syllable position—as in adult speech. 3. A l l children in this study showed evidence of ongoing processes of differentiation, integration, and refinement. However, two general patterns were observed in terms of these motor development processes: a) For most of the children, the majority of differences from adult-like IM took the form of "refinement": children had qualitatively similar gestural patterns to that in adult speech, although the gestures showed more variability. b) In contrast, for a smaller group of children, the dominant processes were "differentiation" and/or "integration." These cases showed, for example, a single dominant movement or timing pattern, or no differentiated movement at all. The current study supports the PMD hypothesis for some children, showing evidence that at these ages they are continuing to develop via all three types of motor development processes (differentiation, integration and refinement). For children, the spatial and temporal 154 coordination of multiple gestures is not easy to master. The present study suggests that the tendency for late acquisition of IM may be due at least in part to a combination of the complex articulatory behaviours required in the production of IM and limited motor coordination skills. 7.2 Limitations While the current study has provided the foundations of the static and temporal organization of IM in children's speech production, it also has limitations. Ultrasound imaging provides good visual information of articulatory gestures and tongue movements during speech. However, ultrasound is not always able to show the tongue tip image clearly, which makes it difficult to measure the relative timing between tongue tip and dorsum movements, as well as to define the constriction location of the tongue tip. In addition, since the application of ultrasound imaging techniques to speech science is relatively new, establishing a way to normalize the tongue shape and location across subjects, as well as within subjects, is still in progress. Until standardized measurement across subjects is available, it would be difficult to define the size and location of a child's tongue relative to that of an adult's. Thus, direct comparison of the articulatory gestures in size and location of the tongue between children and adults has to remain for future research. The design of stimuli can be another issue. The stimuli used in this dissertation have been carefully made to replicate as closely as possible those used in previous articulatory studies of adult speech. For this age group of children, pictures of objects containing IM were used with other pictures, so that IM appears in different syllable positions, just as if it would have appeared in a carrier phrase. However, combining two pictures that are not necessarily 155 related to each other may confuse some children, and thus affect their production. Therefore, the Picture Identification Task used in the present study may not provide a direct comparison between adults and children's speech production of IM. In addition, in future research, modification of stimuli for children is expected when comparing speech across different age groups. That is, for a group of younger children, the experiment design used in the present study may not be easy to follow; and an older group of children, who already know how to read, may not be interested in the present study's design. Making stable and consistent stimuli for all age groups is not an easy task, yet it is an empirical question to consider carefully in future work. 7.3 Implications This dissertation has provided articulatory information of children's IM at 3;11 to 5;9. The determination of a baseline for the articulation of IM (indeed for any sound) in children's speech has been a much needed endeavour to help with the further study of the development of speech production. The ultrasound techniques used in the current study—minimizing invasiveness, danger, and experimental set up, and requiring a relatively simple and short session—are promising, and open up new possibilities for articulatory work with young children. Prior to the present work, no research had directly measured children's lingual gestures in relation to speech motor development at any age, nor had there existed a baseline study of motor behaviour for any articulators in children's speech between the ages of 2 and 6. Thus, the present study of lingual motor development for children around 4 to 6 years 156 contributes greatly to bridging the large gap in the literature between younger and older children with respect to the development of speech motor control in general. While clearly much work remains to be done in this area—including production studies with children at a wider variety of ages—the gestural baseline of the production of IM provided by this dissertation lays the groundwork for a more thorough articulatory account of speech development, and contributes directly to identifying the causes for the difficulties that may reside in the acquisition of certain sounds. This present work is based on the speech of eight monolingual English speaking children, and future research aims to expand it to younger and older groups, and to compare developmental patterns. With this baseline for IM, younger children with glide substitutions now can be compared to see how articulatory gestures of glides and laterals are different and similar; for older children, gestural characteristics of IM and Irl can be compared to find similarities and differences that can account for issues in liquid acquisition in general. The methods used in this study also can be applied to L2 and bilingual populations to test how children and adults exposed to more than one language produce and perceive IM in the ambient language. Moreover, by providing a baseline from normal children, the results of this study can be applied to clinical research in speech pathology. Although this study is only a first step toward a better understanding of the links between motor development and speech production in children's speech, it has begun to lay down some important foundations. 157 Bibliography Armstrong, D., Stokoe, W., & Wilcox, S. (1995). Gesture and the nature of language. Cambridge: Cambridge University Press. Aslin, R.N., Pisoni, D.B., Hennessy, B.L., & Perey, A.J . (1981). Discrimination of voice onset time by human infants: New findings and implications for the effects of early experience. Child Development, 52, 1135-1145. Barlow, J. (1995).The development of on-glides in American English. In Proceedings of Boston Univeristy conference on language development 20. Boston University. Barton, D. (1980). Phonemic perception in children. In G. Yeni-Komshian, J.F. Kavanagh, & C A . Ferguson (Eds.), Child Phonology, 2: Perception. New York: Academic Press. Bernhardt, B., & Stemberger, J. (1998). Handbook of phonological development: From the perspective of constraint-based nonlinear phonology. Academic Press. Bernhardt, B., Gick, B., Bacsfalvi, P., & Ashdown, J. (2003). Speech habitation of hard of hearing adolescents using electropalatography and ultrasound as evaluated by trained listeners. Clinical linguistics & Phonetics, vol. 17, no. 3, 199-216. Borden, G., Harris, K., & Raphael, L. (1994). Speech Science Primer: Physiology, Acoustics, and Perception of Speech. Baltimore, M D : Williams & Wilkins. Bosch, L. , & Sebastian-Galles, N . (1997). Native-language recognition abilities in 4-month-old infants from monolingual and bilingual environments. Cognition, 65, 33-69. Boysson-Bardies, B. de. (2001). How language comes to children: From birth to two years. Cambridge: MIT Press. 158 Boysson-Bardies, B. de, Vihman, M . , Roug-Hellichius, L., Durand, C, Landberg, I., & Arao, F. (1992). Material evidence of infant selection from the target language : A cross-lingusitc phonetic study. In C. Ferguson, L . Menn & C. Stoel-Gammon (Eds.), Phonological development: Model research implications, 369-391. York Press. Browman, C., & Goldstein, L. (1989). Articulatory gestures as phonological units. Phonology, 6, 201-251. Browman, C., & Goldstein, L. (1990a). Representation and reality: Physical systems and phonological structure. Journal of Phonetics, 18, 411-424. Browman, C., & Goldstein, L. (1990b). Tiers in articulatory phonology, with some implications for casual speech. In T. Kingston, & M . E. Beckman (Eds.), Papers in Laboratory Phonology I: Between the grammar and physics of speech, 341-376. Cambridge: Cambridge University Press. Browman, C , & Goldstein, L. (1992). Articulatory phonology: An overview. Phonetica, 49, 155-180. Browman, C , & Goldstein, L. (1995). Dynamics and articulatory phonology. In Port & Van Geider (Eds.), Mind as motion: Explorations in the dynamics of cognition, 175-193. Cambridge, M A : MIT Press. Byrd, D. (1996a). Influences on articulatory timing in consonant sequences. Journal of Phonetics, 24 (2), 209-244. Byrd, D. (1996b). A phase window framework of r articulatory timing. Phonology, 13, 139-169. Byrd, D., & Saltzman, E. (2002). Speech production. In M . Arbib (Ed.), The Handbook of Brain Theory and Neural Networks, 1072-1078. Cambridge: MIT Press. 159 Callan, D.E., Kent, R.D., Guenther, F.H., & Vorperian, H.K. (2000). An auditory-feedback-based neural network of speech production that is robust to developmental changes in the size and shape of the articulatory system. Journal of Speech, Language, and Hearing Research, 43, 721-736. Campbell, F., & Gick, B. (2003). A labial gesture for A poster presentation at 145 lh meeting of the Acoustical Society of America. Nashville, May, 2003. Caplan, F. (1978). The first twelve months of life. New York: Bantam Books. Cutler, A. , Mehler, J., Norris, D., & Segui, J. (1989). Limits on bilingualism. Nature, 340, 229-330. Cutler, A . , Mehler, J., Norris, D., & Segui, J. (1992). The monolingual nature of speech segmentation by biinguals. Congitive Psychology, 24, 381-410. Dinnsen, D. (1992). Variation in developing and fully developed phonetic inventories. In C A . Ferguson, L. Menn & C. Stoel-Gammon (Eds.), Phonological development: models, research, and implications, 191-210. Timonium, M D : York Press. Dinnsen, D. (2001). A reconsideration of children's phonological representations. In B. Skarabela, S. Fish & A. Do (Eds.), Proceedings of Boston University Conference on Language Development, 26, 1-23. Cascadilla Press. Eimas, P.D, Siqueland, E.R., Jusczyk, P.W., & Vigorito, J. (1971). Speech perception in infants. Science, 171, 303-306. Espy-Wilson, C. Y . (1992). Acoustic measures for linguistic features distinguishing the semi vowels /w j r 1/ in English. Journal of the Acoustical Society ofAmerica, 92, 736-751 Farnetani, E. (1999). Coarticulation and connected speech processes. In W. Hardcastle & J. Laver (Eds.), The handbook of phonetic sciences, 371-404. 160 Fougeron, C, & Keating, P. A. (1997). Articulatory strengthening at edges of prosodic domains. Journal of the Acoustical Society of America, 101, 3728-3740. Gafos, A. I. (1999). The articulatory basis of locality in phonology. New York: Garland. Genesee, F. (1989). Early bilingual development, one language or two? Journal of Child Language, 16, 161-179. Genesee, F. (2001). Bilingual first language acquisition: Exploring the limits of the language faculty. In M . McGroarty (Ed.), Annual Review of Applied Linguistics: Language and Psychology, Vol.21, 153-168. Cambridge: Cambridge University Press. Genesee, F. (2002). Bilingual acquisition. Article from http://www.earlychi Idhood.com. Gick, B. (1999). The articulatory basis of syllable structure: a study of English glides and liquids. Doctoral thesis. Yale University. Gick, B. (2002a). The use of ultrasound for linguistic phonetic fieldwork. Journal of International Phonetics Association, 32 (2), 113-121. Gick, B. (2002b). Why liquids are hard to pronounce: The development of tongue motor differentiation. Cognitive Science Colloquium Series, John Hopkins University. October, 2002. Gick, B. (2003). Articulatory correlates of ambisyllabicity in English glides and liquids. Laboratory Phonology VI: Constraints on Phonetic Interpretation. Cambridge: Cambridge University Press. Gick, B., & Bird, S. (submitted).Techniques for field application of lingual ultrasound imaging. Gick, B., & Campbell, F. (2003). Intergestural timing in English Ixl. In Proceedings of the XVth International Congress of Phonetics Sciences. Barcelona. 161 Gick, B., Campbell, F., Oh, S., & Tammburri-Watt, L. (in press). Toward universals in the gestural organizations of syllables: A cross-linguistic study of liquids. Journal of Phonetics. Gick, B., Kang, A. M., & Whalen, D. H. (2002). MRI evidence for commonality in the post-oral articulations of English vowels and liquids. Journal of Phonetics, 30, 357-371. Gick, B. , & Wilson, I. (to appear). Pre-liquid excrescent schwa: what happens when vocalic targets conflict. Laboratory Phonology 8. Gick, B. , & Wilson, I. (2001). Pre-liquid excrescent schwa: what happens when vocalic targets conflict. In P. Dalsgaard, B. Lindberg, & H. Benner (Eds.), Proceedings of the 7th Eurospeech, 273-276. Gick, B., Wilson, I., Koch, K., & Cook, C. (submitted). Giles, S. B., & Moll , K. L. (1975). Cinefluorographic study of selected alophones of English IV. Phonetica, 31, 206-227. Goad, H. (1996). Consonant harmony in child language: evidence against coronal underspecification. Proceedings of the UBC International Conference on Phonological Acquisition, 187-200. Cascadilla Press. Gormley, A. (2003). The production of consonant harmony in child speech. Master's thesis. University of British Columbia. Gracco, V . (1994). Some organisational characteristics of speech movement control. Journal of Speech and Hearing Research, 37, 4-27. Gracco, V. , & Abbs, J. (1986). Variant and invariant characteristics of speech movements. Experimental Brain Research, 65, 156-166. 162 Green, J., Moore, C., Ruark, J., Rodda, P., Morvee, W., & VanWitzenburg, M . (1997). Development of chewing in children from 12 to 48 months: a longitudinal study of E M G patterns. Journal of Neurophysiology, 77, 2704-2716. Green, J.^  Moore, C., Higashikawa, M . , & Steeve, R. (2000). The physiologic development of speech motor control: Lip and jaw coordination. Journal of Speech, Language, and Hearing Research, 43, 239-255. Grosjean, F. (1998). Studying bilinguals: Methodological and conceptual issues. Bilingualism: Language and Congnition, 1, 131-149. Grunwell, P. (1982). Clinical Phonology. London: Croom Helm. Hardcastle, W. (1985). Some phonetic and syntactic constraints on lingual coarticulation during /kl / sequences. Speech Communication, 4, 247-263. Hardcastle, W., & Barry, W. (1985). Articulatory and perceptual factors in l\l vocalisation in English. Work in progress. Phonetics Laboratory, University of Reading, 5, 31-44. Hawkins, S. (1984). On the development of motor control in speech: evidence from studies of temporal.coordination. In N . J. Lass (Ed.), Speech and Language: Advances in basic research and practice, 11. New York: Academic Press. Houde, J., & Jordan, M . (1998). Sensorimotor adaptation in speech production. Science, vol. 279. 1213-1216. Houwer, A. , de. (1995). Bilingual language acquisition. In P. Fletcher & B. MacWhinney (Eds.), The handbook of child language (pp. 219-250). Oxford: Blackwell. Ingram, D. (1974). Phonological rules in young children. Journal of Child Language, I, 49-64. 163 Ingram, D. (1981). The emerging phonological system of an Italian-English bilingual child. Journal of Italian Linguistics, 2, 95-113. Jakobson, R. (1968). Child language, aphasia, and phonological universals. The Hague: Mouton. Jusczyk, P.W., Friederici, A.D. , Wessels, J., Svenkerud, V . Y . , & Jusczyk, A . M . (1993). Infants' sensitivity to the sound patterns of native language words. Journal of Memory and Language, 32, 402-420. Kahn, D. (1976). Syllable-based generalizations in English phnology. New York: Garland Publishing. Kalverboer, A. , Hopkins, B., & Geuze, R. (1993). Motor development in early & later childhood: longitudinal approaches. Cambridge University Press. Keating, P. A. (1995). Effects on prosodic position on It, dl tongue/palate contact. Proceedings of Internalioanl Congress on Phonetic Sciences 95 Stockholm, 3, 432-435. Kelso, J.A.S. (1977). Motor control mechanisms underlying human movement reproduction. Journal of Experimental Psychology: Human Perception and performance, 3, 529-543. ' Kelso, J.A.S., Saltzman, E., & Tuller, B. (1986). The dynamic perspective on speech production: Data and theory. Journal of Phonetics, 14, 29-59. Kelso, J.A.S., Southard, D.L., & Goodman, D. (1979). On the nature of human interlimb coordination. Science, vol. 203. 1029-1031. 164 Kelso, J.A.S., & Tuller, B. (1987). Intrinsic time in speech production: Theory, methodology, and preliminary observations. In E. Keller & M . Gopnik (Eds.), Motor and sensory processes of language (pp. 203-222). Hillsdale, NJ: Lawrence Erlbaum. Kent, R. D. (1976). Anatomical and neuromuscular maturation of the speech mechanism: Evidence from acoustic studies. Journal of Speech and Hearing Research, 19, 421 -447. Kent, R. D. (1983). The segmental organisation of speech. In P. MacNeilage (Ed.), The production of speech (pp. 57-89). New York: Springer-Verlag. Kent, R. D. (1992). The development of phonological development. In C. Ferguson, L. Menn & C. Stoel-Gammon (Eds.,), Phonological development: Model research implications, ch.2. York Press. Kochetov, A. (2002). Production, perception, and emergent phonotactic patterns: a case of contrastive palatalization. New York & London: Routledge. Kochetov, A. (2003). Syllable position effects and gestural organization: articulatory evidence from Russian. In Papers in Laboratory Phonology VIII: Varieties of Phonologial Competence. Krakow, R. A. (1989). Inter-articulator timing and syllable structure. Meeting handbook, Linguistic Society of America, 38. Krakow, R. A. (1999). Physiological organisation of syllables: A review. Journal of Phonetics, 27, 23-54. Kuhl, P. K., & Meltzoff, A . N . (1982). The bimodal perception of speech in infancy. Science, 2/5,1138-1141. Ladefoged, P. (1993). A course in phonetics, third edition. Harcourt Brace. 165 Ladefoged, P. (2001). Vowels and consonants: An introduction to the sounds of languages. Oxford: Blackwell. Ladefoged, P., & Maddieson, I. (1996). The sounds of the world's languages. Oxford: Blackwell. Lazarus, J. A. , & Todor, J. I. (1987). Age differences in the magnitude of associated movement. Developmental Medicine and Child Neurology, 29, 726-733. Leonard, L. (1995). Phonological Impairment. In P. Fletcher & B. MacWhinney (Eds.), The handbook of child language (pp. 573-602). Oxford: Blackwell. Lindblom, B. (1983) Economy of speech gestures. In P. MacNeilage (Ed.), The production of speech, 2.17-246. New York: Springer-Verlag. Locke, J. (1980). The inference of speech perception in the phonologically discordered child. Part II: Some clinically novel procedures, their use, some findings. Journal of Speech and Hearing Disorders, 4, 445-468. Locke, J. (1983). Phonological acquisition and change. New York: Academic Press. Locke, J. (1993). The child's path to spoken language. Oxford: Blackwell. Leonard, L. B. (1995). Phonological Impairment. In P. Fletcher & B. MacWhinney (Eds.), The handbook of child language, 573-602. Oxford: Blackwell. MacKay, I. (1977). Tenseness in vowels: an ultrasonic study. Phonetica, 34, 325-351. Macken, M . A. (1980). Aspects of the acquisition of stop systems: A cross-linguistic perspectives. In G. Yeni-Komshian, J.F. Kavanagh, & C.A. Ferguson (Eds.), Child Phonology, 1: Production. New York: Academic Press. MacNeilage, P. F. (1999). Acquisition of speech. In W. Hardcastle & J. Laver (Eds.), The handbook of phonetic sciences (pp. 301-332). Oxford: Blackwell. 166 MacNeilage, P. F., & Davis, B. F. (1990). Acquisition of speech production: Frames, then content. In M . Jeannerod (Ed.), Attention and Performance 13: Motor representation and control, 453-476. Hillsdale, NJ: Lawrence Erlbaum. Menn, L. (1983). Development of articulatory, phonetic, and phonological capabilities. In B. Butterworth (Ed.), Language Production, 2. London: Academic Press. Menn, L. , Mattei, E. (1992). The "two-lexicon" account of child phonology: Looking back, looking ahead. In C. A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological Development: Models, research, implications. Timonium, MD:York Press. Menn, L., & Stoel-Gammon, C. (1995). Phonological development. In P. Fletcher & B. MacWhinney (Eds.), The handbook of child language (pp. 335-360). Oxford: Blackwell. Menyuk, P., Liebergott, J., & Schultz, M . (1986). Predicting phonological development. In B. Lindblom & R. Zetterstrom (Eds.), Precursors of early speech. New York: Stockton Press. Miesel, J .M. (1989). Early differentiation of languages in bilingual children. In K. Hyltenstam & L. Obler (Eds.), Bilingualism across the lifespan: Aspects of acquisition, maturity and loss (pp. 13-40). Cambridge: Cambridge University Press. Narayanan, S., Alwan, A. , & Haker, K. (1997). Toward articulatory-acoustic models for liquid approximants based on MRI and EPG data. Part 1. The laterals. Journal of Acoustical Society of America,\0l (2), 1064-1077. Narayanan, S., Bryd, D., & Kaun, A. (1999). Geometry, kinetics, and acoustics of Tamil liquid consonants. Journal of Acoustical Society of America, 106 (4), pt. 1, 1993-2007. 167 Nazzi, T., Bertoncini, J., & Mehler, J. (1998). Language discrimination by newborns: Toward an understanding of the role of rhythm. Journal of Experimental Psychology, 24, 1-11. Nittrouer, S. (1993). The emergence of mature gestural patterns is not uniform: Evidence from an acoustic study. Journal of Speech and Hearing Research, 36, 959-971. Nittrouer, S., Studdert-Kennedy, M , & McGowan, R. S. (1989). The emergence of phonetic segments: Evidence from the spectral structure of fricative-vowel syllables spoken by children and adults. Journal of Speech and Hearing Research, 32, 120-132. Oh, S. (2002). Cross-language blending of IM gestures by Korean-English bilingual children. In Proceedings of Boston University conference on language development, 26, 473-484. Somerville, M A : Cascadilla Press. Oh, S. (in press). Liquid acquisition in Korean: A case study of monolingual and bilingual children. In J. Song and J. Yeon (Eds.), Frontiers in Korean Language Acquisition. The Saffron Korean Linguistics Series, 120-141. London: Eastern Art Publishing. Oh, S., & Gick, B. (2002). An ultrasound study of articulatory gestures of Korean liquids. In Proceedings of International Conference on Speech Sciences I, 236-242. Oiler, D. K. (1975). Simplification as the goal of phonological processes in child speech. Language Learning, 24, 299-303. Oiler, D. K. (1978). Infant vocalisation and the development of speech. Allied Health and Behavioural Sciences, 1, 525-549. Oiler, D. K. (1980). The emergence of the sounds of speech in infancy. In Yeni-Komshian, Kavanagh, & Ferguson (Eds.), Child Phonology I. Paradis, M . (1981). Neurolinguistic organization of a bilingual's two languages. In J. Copeland (Ed.), The seventh LACUSforum. Columbia, SC: Hornbeam Press. 168 Paradis, M . (1990). Language lateralization in bilinguals: enough already! Brain and Language, 39, 576-586. Pater, J., & Paradis, J. (1996). Truncation without templates in child phonology. In Preceedings of the Boston University conference on language development, 20, 540-552. Somerville, M A : Cascadilla Press. Perkell, J. S. (1980). Phonetic features and the physiology of speech production. In B. Butterworth (Ed.), Language production, 337-372. London: Academic Press. Perkell, J. S., Cohen, M . , Svirsky, M . , Matthies, M , Garabieta, I., & Jackson, M . (1992). ' Electromagnetic midsagittal articulatometer (EMMA) systems for transducing speech articulatory movements. Journal of the Acoustical Society of America, 92, 3078-3096. Peters, A . M . (1977). Language learning strategies: Does the whole equal the sum of the parts? Language, 53, 560-73. Peters, A . M . (1983). The units of language acquisition. Cambridge: Cambridge University Press. Peterson, G.E., & Barney, K. L. (1952). Control methods in a study of the vowels. Journal of the Acoustical Society of America, 24, 175-184. Piske, T. (1997). Phonolgical organisation in early speech production: Evidence for the importance of articulatory patterns. Speech Communications, 22, 279-295. Pouplier, M . (2003). Units of phonological encoding: empirical evidence. Doctoral dissertation. Yale University. Pye, C. (1986). One lexicon or two? An alternative interpretation of early bilingual speech. Journal of Child Language, 13, 591-593. Romaine, S. (1995). Bilingualism (2 n d Ed.). Oxford: Blackwell. 169 Rizzoratti, G., & Arbib, M . (1998). Language within our grasp. Trends in NeuroSciences, 21 (5), 188-194. Amsterdam: Elsevier. Saltzman, E. (1995). Intergestural timing in speech production: data and modeling. In K. Elenius, & P. Branderud (Eds.), Proceedings of theXIIIth International Congress of Phonetics Sciences, 84-91. Stockholm: K T H and Stockholm University. Saltzman, E., & Byrd, D. (2000). Task-dynamics of gestural timing: phase windows and multifrequency rhythms. Human Movement Science, 19, 499-526. Sanders, E. K. (1972). When are speech sounds learned? Journal of Speech and Hearing Disorders, J 7 , 55-63. Schnitzer, M . L. , & Krasinski, E. (1994). The development of segmental phonology production in a bilingual child. Journal of Child Language, 21, 585-622. Schuster, C. S. & Ashburn, S. S. (1992). The process of human development: A holistic life-span approach. (Third edition). Philadelphia: J. B. Lippincott Company. Scobbie, J., & Wrench, A. (2003). An articulatory investigation of word final IM and IV-sandhi in three dialects of English. In Proceedings of the XVth International Congress of Phonetics Sciences. Barcelona. Sharkey, S. G., & Folkins, J. W. (1985). Variability of lip and jaw movements in children and adults: implications for the development of speech motor control. Journal of Speech and Hearing Research, 25 (1), 8-15. Smith, A . (1992). The control of orofacial movements in speech. Critical Review in Oral Biology and Medicine, 3,233-267. Smith, A. , & Goffman, L. (1996). Stability and patterning of speech movement sequences in children and adults. Journal of Speech, Language, and Hearing Research, 41(\), 18-170 30. Smith, A. , Goffman, L., & Stark, R. (1995). Speech motor development. Seminars in Speech and Language, 16 (2), 87-99. Smith, B. (1978). Temporal aspects of English speech production: A developmental perspective. Journal of Phonetics, 6, 37-67. Smith, B. (1994). Effects of experimental manipulations and intrinsic contrasts on relationships between duration and temporal variability in children's and adults' speech. Journal of Phonetics, 22, 155-175. Smith, B., & McLean-Muse, A. (1986). Articulatory movement characteristics of labial consonant productions by children and adults. Journal of Acoustical Society of America, 80(5), 1321-1328. Smith, N . (1973). The acquisition of phonology. Cambridge: Cambridge University Press. Sproat, R., & Fujimura, O. (1993). Allophonic variation in English III and its implications for phonetic implementation. Journal of Phonetics, 21, 291-311. Stampe, D. (1969). The acquisition of phonetic representation. Proceedings of the fifth regional meeting of the Chicago Linguistic Society, 433-444. Stark, R. E. (1980). Stages of speech development in the first year of life. In G. Yeni-Komshian, J.F. Kavanagh, & C A . Ferguson (Eds.), Child Phonology, 1: Production. New York: Academic Press. Stemberger, J. P. (1988). Between-word processes in child phonology. Journal of Child Language, 75,39-61. Stemberger, J. P. (1992). A connectionist view of child phonology: Phonological processing 171 without phonological processes. In C. A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological Development: Models, research, implications. Timonium, M D : York Press. Stemberger, J., & Stoel-Gammon, C. (1991). The underspecification of coronals: Evidence from language acquisition and performance errors. In C. Paradis & J.-F. Prunet (Eds.), The special status of coronals, 181-199. San Diego: Academic Press. Stevens, K. (2000). Acoustic Phonetics. Cambridge, M A : MIT Press. Stoel-Gammon, C. (1983). Constraints on consonant-vowel sequences in early words. Journal of Child Language, 10, 455-457. Stoel-Gammon, C. (1985). Phonetic inventories, 15-22 months: a longitudinal study. Journal of Speech and Hearing Research, 28, 505-512. Stoel-Gammon, C. (1992). Prelinguistic vocal development: Measurement and predictions. In C. A . Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological Development: Models, research, implications. Timonium, M D : York Press. Stoel-Gammon, C. (1996). On the acquisition of velars in English. Proceedings of the UBC international conference on phonological acquisition, 201-214. Cascadilla Press. Stone, M . (1990). A three-dimensional model of tongue movement based on ultrasound and x-ray microbeam data. Journal of the Acoustical Society of America, 87, 2207-2217'. Stone, M . (1993). Measuring speech production, Part II: Indirect articulatory measurements. Acoustical Society of America Video Series. Stone, M . (1997). Laboratory techniques for investigating speech articulation. In W. Hardcastle & J. Laver (Eds.), The handbook of phonetic sciences, 11 -32. Blackwell. 172 Stone, M . (1999). Imaging techniques. In W. Hardcastle & L. Hewlett (Eds.), Coarticulation: Theory, Data, and Techniques, 246-259. Cambridge University Press. Straight, H. S. (1980). Auditory versus articulatory phonological processes and their development in children. In G. Yeni-Komshian, J. Kavanagh, & C A . Ferguson (Eds.), Child Phonology, 2: Perception. New York: Academic Press. Strange, W., & Jenkins, J.J. (1978). Role of linguistic experience in the perception of speech. In R.D. Walk, H.L. Pick (Eds.), Perception and Experience (pp. 125-69). New York: Plenum. Strange, W., & Broen, P. A. (1980). Perception and production of approximate consonants by 3-year-olds: A first study. In G. Yeni-Komshian, J. Kavanagh, & C A . Ferguson (Eds.), Child Phonology, 2: Perception. New York: Academic Press. Studdert-Kennedy, M . (1991). Comment: The emergent gesture. In I. G. Mattingly & M . Studdert-Kennedy (Eds.), Modularity and the motor theory of speech perception: Proceedings of a conference to honour Alvin M. Leberman. Hillsdale, NJ: Lawrence Erlbaum. Studdert-Kennedy, M . , & Goldstein, L . (2003). Launching language: The gestural origin of discrete infinity. In M . H . Christiansen & S. Kirby (Eds.), Language Evolution: The states of the art. Oxford University Press. Thelen, E. (1995). Motor Development: A new synthesis. American Phychologist, 50, 79-95. Thelen, E., & Smith, L. (1994). A dynamic systems approach to the development of cognition and action. Cambridge, M A : MIT Press/Bradford. Tobin, Y . (1997). Phonology as human behavior. London: Duke University Press. 173 Trehub, S. E. (1976). The discrimination of foreign speech constrasts by infants and adults. Child Development, 47,466-412. Tremblay, S., Shiller, D, & Ostry, D. (2003). Somatosensory basis of speech production. Nature, vol. 423, 866-869. June 19, 2003. Trevarthen, C. (1984). Signs of motivation for speech in infants, and the nature of a mother's support for development of language. In B. Llindblom & R. Ztterstrom (Eds.), Precursors of Early Speech. New York: Stockton Press. Tuller, B., & Kelso, J.A.S. (1984). The timing of articulatory gestures: Evidence for relational invariants. Journal of the Acoustical Society of America, 76, 1030-1036. Tuller, B., & Kelso, J.A.S. (1990). Phase transition in speech production and their perceptual consequences. In Jeannerod (Ed.), Attention & Performance XIIL (pp. 429-451). Vihman, M . (1985). Language differentiation by the bilingual infant. Journal of Child Language, 12, 297-324. Vihman, M . (1996). Phonological Development. Oxford: Blackwell. Volterra, V. , & Taeschner, T. (1978). The acquisition and development of language by bilingual children. Journal of Child Language, 5, 311-326. Walsh-Dickey, L. (1997). The phonology of liquids. Doctoral dissertation, University of Massachusetts. Werke, J., Gilbert, J.H.V., Humphery, K. , & Tees, R.C. (1981). Developmental aspects of cross-language speech perception. Child Development, 52, 349-355. Werker, J.F., & Tees, R.C. (1984). Cross-language speech perception: evidence for perceptual; reorganization during the first year of life. Infant Behavior Development, 7,49-63. 174 Werker, J.F., & Tees, R.C. (1999). Influences on infant speech processing: toward a new synthesis. Annual Review. Psychology, 50, 509-535. Yeni-Komshian, G., Kavanagh, J.F., & Ferguson, C.A. (Eds.) (1980). Child Phonology, I: Production; 2: Perception. New York: Academic Press. 175 A P P E N D I X A . List of Utterances JN(4;11) A Lilo A L i l o A leg A ladybug A leaf A leash A lips seal hey whale (error) pail hey snail baby jail (error) nail baby wheel baby shell baby tail baby seal eight whale eight snail eight jail eight wheel eight shell angel tail angel salad yellow yellow elephant elephant elephant elephant JN(5;1) J ladybug Jleaf J lips Jleg A leg J light J ladder bee ladybug bee leaf bee lip bee leg seal baby seal baby whale baby whale baby a bell bell baby bell baby snail baby snail baby jail baby jail baby nail baby nail baby a wheel wheel baby wheel baby shell baby shell baby tail baby seal bed whale bed bell bed snail bed jail bed bell egg shell egg shell egg snail angel jail angel nail angel FL (4;7) seal seal bee whale whale bee bell bell bee snail snail bee jail jail bee nail nail bee wheel wheel bee shell shell bee tail tail bee seal baby whale baby bell bed snail baby jail baby nail bed wheel bee shell bed tail baby tail seal eight whale eight bell eight bell egg snail eight jail eight nail eight wheel eight wheel ear shell ear shell egg tail egg 176 wheel ear ladybug seal ear a ladybug apple angel a lake elephant a leaf elephant a lips teletubbies a leg teletubbies a lamb teletubbies a lamp jellybeans lemon jellybeans a lemon jelly ladder yellow a ladder air plane television elephant elephant jellybeans helicopter helicopter willy wonka willy wonka yellow yellow 177 JL (4;6) M C (5;6) a and a whale seal and bumblebee whale a seal bee whale a whale and baby snail a whale and baby snail a bell a man stuck in jail snail and baby jail jail and baby jail a nail and baby wheel bee jail and bed wheel bee tail seal and um a bee tail baby seal bee seal ear shells and bed whale angel shell bed bell and egg bell bed snail angel bell bed jail and eight tail nail eight tail a shell and.eight ball orange tail eight ball orange ladybug girl orange a ladybug girl orange J ladybug whale angel bumblebee leaf whale angel C and leaf snail angel C and lip snail angel Jleg a man stuck in jail and Jleg angel J lamp a man stuck in jail and J lemon angel J lemon nail J ladder nail angel A ladder tail lollipops monkey's tail and angel lollipops monkey's tail and angel pillows seal ear grapefruit seal ear teletubbies wheel ear teletubbies wheel ear elephant bell egg jellybeans bell egg jellybeans egg clam blue a little girl red little girl helicopter teletubbies yellow LR (5;6) lilo bellybutton belly button seal seal bumblebee a bumblebee and a seal seal seal and B bell bell and bed S and baby snail baby and snail snail and baby jail jail and baby nail and baby shell and bed monkey and baby tail tail and baby laughing laughing A and a ladybug B and leaf lip and B B and lip leg and A A and leg lamp and A A and lamp lemon lemon and A A and lemon ladder and A A and ladder teletubbies elephant jellybeans jellybean and elephant blue and red helicopter grapes pillows 178 teletubbies teletubbies elephant elephant jellybeans jellybeans yellow yellow ladybug a aladybug leaf leaf lip lip leg sheep lamp lamp ladder ladder a lamp a lamp a leaf a leg a leg collar red willy wonka silly red green loonie lollipops tiger lily yellow collar willy wonka ear ear and dolphin angel and whale whale and angel egg and bell bell and egg snail and angel jail and angel nail and angel shell and egg monkey and angel tail tail and angel loon a man trying to hear somebody listening to music 179 KR (5;9) RC (3;11) A D (4;4) bell baby ear and wheel whale bell bed in jail snail bell bed jail jail jail baby nail nail jail baby nail wheel nail baby sea shell shell nail baby sea shell tail seal baby sea shell lips seal bee snail light seal bee snail leg seal bee snail lemon seal bee tail ladder shell baby whale letter shell bed whale yellow shell bed wheel yellow shell bed man in jail baby jellybean shell bed nail baby jelly snail baby snail baby jello snail baby sea shell egg jello tail baby sea shell egg whale baby wheel ear whale baby ladder whale bee ladybug whale bee ladybug wheel baby leaf wheel bee leash wheel bee leg basketball orange light basketball orange lips bell egg lips bell egg look at the stars bell egg elephant doll orange elephant doll orange helicopter jail a helicopter jail a jellybeans jail angel jellybeans jail angel pillow nail a pillow nail a yellow nail angel yellow seal a bear seal angel door snail a dear snail a otter 180 snail angel bumble bee snail egg doggie tail a dolphin tail a tail a tail a tail angel tail angel whale a whale a whale angel whale angel wheel ear wheel ear a ladder a ladder aladybug a lamp a lamp a lamp a leaf a leg a leg a lemon a lemon a lemon a loonie a loonie or a toonie a loonie or a toonie bee leaf bee leaf bee lip bee lip leash leash elephant elephant helicopter helicopter jello jello jellybeans jellybeans jellybeans lollipops lollipops 181 lollipops looking at the stars up a telescope looking up at the stars in the telescope pillows pillows silly silly smiling teletubbies teletubbies yellow yellow yellow and jello a man is in jail and a bell easter egg nail baby snail baby tail baby nail prince doll doll jail ladybug a ladybug a leash Lilo Lilo nail seal tail tail *Lilo, Willy Wonka, Teletubbies: children's movie/TV characters 182 APPENDIX B. Word List for Perception Judgment by Adults RC (3;11) A l A2 A3 wheel 0 1/w 0 whale l(wheel) w (wheel) 0 tail o 1 o Final snail 0 o/w 1 nail 0 o 0 lips 1 1 1 light 1 1 1 Initial leg 1 l(w?) 1 l[w]eaf w (wee) w w ladybug ? 1? ? ladder 1 1 1 yellow w 1/1 0 Intervocalic jellybeans ? 1 1 helicopter 1 1 1 elephant 1 w 1/u 183 A D (4;4) A l A2 A3 wheel l /o w ow whale 1 /o(real) 1 (real) ow (wheel) tail l /o 1 ow (pail) Final snail 1 (nail) o (mail) 1 (nail) nail 1 w o bell 1 1 w lips 1 1 1 lemon 1 1 1 Initial leg 1 1 1 leaf 1 1 1 ladder 1 1 1 yellow 1 1 1 Intervocalic jellybeans y w? 1 Jello l 1 1 JL (4;7) A l A2 A3 Final wheel l /o ] 1/a whale o 1 0 tail a • * 3 snail a * O nail 0 * 0 bell o (bill) U Initial lip 1 1 1 leg 1 (late) 1 (leg?) 1 leaf 1 (leap) 1 (lake?) ladybug 1 1 1 ladder 1 1 1 Intervocalic yellow 1 1 jellybeans 1 1 helicopter 1 1 ' elephant 1 1 1 185 FL (4;9) A l A2 A3 Final wheel 1 w 1 whale 1 1 tail 1 1 snail 1 1 nail 1 1 1 bell 1 1 1 Initial lips 1 1 1 lemon 1 1 1 leg 1 1 1 leaf 1 1 1 ladybug 1 1 1 ladder 1 1 1 Intervocalic yellow 1 1 1 jellybeans 1 1 1 helicopter 1 1 1 elephant 1 1 1 186 JN(5;1) A l A2 A3 wheel 1 wt 1 whale 1 wl 1 tail 1 l /o snail 1 l /o Final nail 1 1 bell 1 1 1 shell 1 1 1 lip 1 1 1 leg 1 1 1 Initial leaf 1 1 1 ladybug 1 1 1 ladder 1 1 1 yellow 1 1 1 Intervocalic jellybeans 1 1 1 elephant 1 1 1 187 LR (5;6) A l A2 A3 Final seal 1 1 l /o wheel 1 1(7) ? tail 1 1 snail * * nail 1 1 bell 1 1? 1 Initial lip 1 1 1 lemon 1 1 1 ladybug 1 1 1 ladder 1 1 1 Intervocalic yellow 1 1 1 jellybeans 1 1 1 helicopter 1 1 1 elephant 1 1 1 188 MC(5;6) A l A2 A3 Final seal 1 al 1 whale 1 al 1 tail 1 1 1 snail 1 nail 1 1 1 bell 1 * w Initial lip 1 1 . 1 lemon 1 1 1 leg 1 1 1 ladybug 1 1 1 ladder 1 1 1 Intervocalic yellow 1 1 1 jellybeans 1 1/w w helicopter 1 1 1 elephant 1 1 1 189 K R (5;9) A l A2 A3 wheel 0 1 0 whale o 1 o tail 0 0 Final snail 1 1 t/o nail e: 1 l /o bell 1 1 l /o seal o 1 " l /o lip 1 1 lemon 1 1 1 Initial leg 1 1 leaf 1 1 1 ladybug 1 1 1 ladder 1 1 1 yellow 1 1 1 Intervocalic jellybeans w 1 helicopter 1 V 1 • elephant 1 1 1 190 A P P E N D I X C. Statistics for Spatial Measurement: T D backing A . Overall Results ANOVA Table for diff x DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 7472.176 1494.435 15.311 <0001 76.553 1.000 Residual 370 36115.143 97.608 Means Table for diff x Effect: position Count Mean Std. Dev. Std. Err. # initial 42 7.810 6.862 1.059 v-initial 40 10.400 9.303 1.471 final-v 88 14.636 12.094 1.289 final-c 67 16.881 10.777 1.317 final # 61 10.770 10.952 1.402 intervoc 78 4.256 6.470 .733 Fisher's PLSD for diff x Effect: position Significance Level: 5 % Mean Diff. Crit. Diff. P-Value # initial, v-initial -2.590 4.292 .2361 # initial, final-v -6.827 3.644 .0003 # initial, final-c -9.071 3.824 <0001 # initial, final # -2.961 3.895 .1358 # initial, intervoc 3.553 3.718 .0610 v-initial, final-v -4.236 3.705 .0251 v-initial, final-c -6.481 3.882 .0011 v-initial, final # -.370 3.953 .8539 v-initial, intervoc 6.144 3.778 .0015 final-v, final-c -2.244 3.150 .1621 final-v, final # 3.866 3.237 .0194 final-v, intervoc 10.380 3.021 <0001 final-c, final # 6.110 3.438 .0005 final-c, intervoc 12.624 3.236 <0001 final #, intervoc 6.514 3.321 .0001 25 H 20 i # initial v-initial final-v final-c final # intervoc S One Sample t-test Split By: subject Hypothesized Mean = 0 s Mean DF t-Value P-Value s diff x, Total 11.043 375 19.861 <0001 diff x, JN 18.776 48 15.403 <0001 s diff x, KR 16.235 97 15.586 <0001 diff x, AD -2.160 24 -1.638 .1145 s diff x, FL 5.472 71 10.945 <0001 s diff x, RC 8.686 34 10.126 <0001 s diff x, LR 7.484 30 11.620 <0001 s diff x, MC 5.920 24 3.244 .0034 s diff x, JL 15.049 40 6.054 <0001 191 APPENDIX D. Statistics for Temporal Measurement: Relative Timing (TT-TD) A. Overall Results ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 266282.265 53256.453 23.320 <0001 116.600 1.000 Residual 429 979717.522 2283.724 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. # initial 59 -23.141 48.790 6.352 v-initial 51 13.059 38.879 5.444 final-v 88 24.597 42.391 4.519 final-c 63 44.400 51.100 6.438 final # 90 54.020 60.023 6.327 intervoc 84 7.532 39.146 4.271 Interaction Bar Plot for TT-TD (ms) Effect: position Error Bars: ?1 Standard Deviation(s) Row exclusion: all subject-timing.svd # initial v-initial final-v final-c final* intervoc Fisher's PLSDfor TT-TD (ms) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mean Diff. Crit. Diff. P-Value # initial, v-initial -36.200 17.959 <0001 s # initial, final-v -47.737 15.805 <0001 s # initial, final-c -67.541 17.017 <0001 s # initial, final # -77.161 15.734 <0001 s # initial, intervoc -30.673 15.955 .0002 s v-initial, final-v -11.538 16.530 .1708 v-initial, final-c -31.341 17.693 .0005 s v-initial. final # -40.961 16.463 <0001 s v-initial, intervoc 5.527 16.674 .5151 final-v, final-c -19.803 15.502 .0124 s final-v, final # -29.423 14.081 <0001 s final-v, intervoc 17.064 14.328 .0197 s final-c, final # -9.620 15.429 .2211 final-c, intervoc 36.868 15.655 <0001 s final #, intervoc 46.488 14.250 <0001 s One Sample t-test Split By: subject Hypothesized Mean = 0 Mean DF t-Value P-Value TT-TD (ms). Total 22.430 434 8.731 <0001 TT-TD (ms), JN 65.241 48 7.920 <0001 TT-TD (ms), KR 5.772 74 1.045 .2995 TT-TD (ms), AD 33.300 21 4.387 .0003 TT-TD (ms), FL 26.091 96 4.759 <0001 TT-TD (ms), RC 21.191 54 2.736 .0084 TT-TD (ms), LR -2.039 48 -.401 .6902 TT-TD (ms), MC 21.275 35 3.487 .0013 TT-TD (ms), JL 19.852 51 2.408 .0197 192 B. Overall results (including JN 4;11) ANOVA Table for TT-TD (ms) DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 262554.043 52510.809 17.468 <0001 87.340 1.000 Residual 472 1418885.094 3006 112 Means Table for TT-TD (ms) Effect: position Count Mean Std. Dev. Std. Err. # initial 64 -15.089 55.952 6.994 v-initial 58 24.688 49.730 6.530 final-v 95 35.403 57.843 5.935 final-c 70 54.707 59.740 7.140 final # . 100 51.615 59.564 5.956 intervoc 91 9.148 43.295 4.539 Interaction Bar Plot for T T-TD(ms) Effect: position Error Bars: 71 Standard Deviation(s) 100 T ' ' ' 1 80 " I 60 H -20 " -40 --60 " [ -80 • • • ' ' • # initial v-initial final-v final-c final # intervoc Fisher's PLSD for TT-TD(ms) Effect: position Significance Level: 5 % Mean Diff. Crit. Diff. P-Value # initial, v-initial -39.777 19.532 <0001 S # initial, final-v -50.492 17.423 <0001 S # initial, final-c -69.796 18.633 <0001 S # initial, final # -66.704 17.246 <0001 S # initial, intervoc -24.237 17.576 .0070 S v-initial, final-v -10.715 17.953 .2415 v-initial, final-c -30.019 19.130 .0022 S v-initial, final # -26.927 17.782 .0031 S v-initial, intervoc 15.540 18.102 .0923 final-v, final-c -19.304 16.971 .0259 S final-v, final # -16.212 15.436 .0396 S final-v, intervoc 26.255 15.803 .0012 s final-c, final # 3.092 16.790 .7176 final-c, intervoc 45.559 17.128 <0001 s final #, intervoc 42.467 15.609 <0001 s One Sample t-test Split By: subject Hypothesized Mean = 0 Mean DF t-Value P-Value One Sam pie t-test Split By: position TT-TD (ms), Total 2.9E1 477 10.518 <0001 TT-TD (ms), JN 6.8E1 55 9.106 <0001 Hypothesized Mean = = 0 TT-TD (ms), KR 5.772 74 1.045 .2995 Mean DF t-Value P-Value TT-TD (ms), AD 33.3 21 4.387 .0003 TT-TD (ms), Total 28.563 477 10.518 <0001 TT-TD (ms), FL 21.46 89 3.883 .0002 TT-TD (ms),# initial -15.089 63 -2.157 .0348 TT-TD (ms), RC 2.1 E1 54 2.736 .0084 TT-TD (ms), v-initial 24.688 57 3.781 .0004 TT-TD (ms), LR -2.04 48 -.401 .6902 TT-TD (ms), final-v 35.403 94 5.966 <0001 TT-TD (ms), MC 2.1E1 35 3.487 .0013 TT-TD (ms), final-c 54.707 69 7.662 <0001 TT-TD (ms), JL 2E1 51 2.408 .0197 TT-TD (ms), final # 51.615 99 8.665 <0001 TT-TD (ms), JN4 9.1E1 42 7.647 <0001 TT-TD (ms), intervoc 9.148 90 2.016 .0468 193 C. Individual Results : JN (4;11) ANOVA Table for TT-TD (ms) Row exclusion: all subject-t iming.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 134945.282 26989.056 8.420 <0001 42.098 1.000 Residual 37 118603.706 3205.506 Means Table for TT-TD (ms) Effect: posit ion Row exclusion: all subject-t iming.svd Count Mean Std. Dev. Std. Err. # initial 5 79.920 50.502 22.585 v-initial 7 109.414 37.053 14.005 final-v 7 171.257 55.816 21.097 final-c 7 147.471 53.891 20.369 final # 10 29.970 53.118 16.797 intervoc 7 28.543 80.263 30.337 225 -200 -# initial v-initial final-v final-c final # intervoc JN (4; 10) Fisher 's P L S D f o r TT-TD (ms) Effect: posit ion Significance Level : 5 % Row exclus ion: all subject-t iming.svd Mean Qff. Crit. Off. P-Value # initial, v-initial -29.494 67.172 .3794 # initial, final-v -91.337 67.172 .0090 S # initial, final-c -67.551 67.172 .0488 S # initial, final # 49.950 62.833 .1157 # initial, intervoc 51.377 67.172 .1297 v-initial, final-v -61.843 61.319 .0482 S v-initial. final-c -38.057 61.319 .2164 v-initial, final # 79.444 56.533 .0072 S v-initial, intervoc 80.871 61.319 .0111 S final-v, final-c 23.786 61.319 .4369 final-v, final #' 141.287 56.533 <0001 S final-v, intervoc 142.714 61.319 <0001 S final-c, final # 117.501 56.533 .0002 s final-c, intervoc 118.929 61.319 .0004 s final #, intervoc 1.427 56.533 .9595 One Sample t-test Split By: position Hypothesized Mean = 0 Row exclusion: all subject-t iming.svd Mean DF t-Value TT-TD (ms). Total TT-TD (ms), # initial TT-TD (ms), v-initial TT-TD (ms). final-v TT-TD (ms), final-c TT-TD (ms), final # TT-TD (ms), intervoc P-Value 90.607 42 7.647 <0001 79.920 4 3.539 .0240 109.414 6 7.813 .0002 171.257 6 8.118 .0002 147.471 6 7.240 .0004 29.970 9 1.784 .1081 28.543 6 .941 .3831 194 D. JN (5;1) ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 66498.997 13299.799 6.382 .0001 31.912 .996 Residual 50 104190.856 2083.817 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. 150 -| # initial 3 -44.400 38.452 22.200 125 " v-initial 11 42.382 36.754 11.082 100 -final-v 7 99.900 38.452 14.533 75 " final-c 18 92.500 29.244 6.893 final # 9 77.700 55.222 18.407 50 " intervoc 8 49.950 75.517 26.699 25 " Fisher's PLSD for TT-TD (ms) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mear; Diff. Crit. Diff. P-Value # initial, v-initial -86.782 59.720 .0053 S # initial, final-v -144.300 63.271 <0001 s # initial, final-c -136.900 57.178 <0001 S # initial, final # -122.100 61.126 .0002 S # initial, intervoc -94.350 62.073 .0036 S v-initial, final-v -57.518 44.331 .0120 s v-initial, final-c -50.118 35.090 .0060 S v-initial, final # -35.318 41.211 .0914 v-initial, intervoc -7.568 42.604 .7227 final-v, final-c 7.400 40.841 .7174 final-v, final # 22.200 46.207 .3392 final-v, intervoc 49.950 47.453 .0395 S final-c, final # 14.800 37.432 .4309 final-c, intervoc 42.550 38.960 .0329 s final #, intervoc 27.750 44.553 .2167 interaction Bar Plot for TT-TD (ms) Effect: position Error Bars: ?1 Standard Deviation(s) Row exclusion: all subject-timing.svd 0 -25 -50 -75 -100 # initial v-initial final-v final-c final # intervoc One Sample t-test Split By: position Hypothesized Mean = 0 Row exclusion: all subject-timing.svd Mean DF t-Value P-Value -TD (ms), Total 67.789 55 9.106 <0001 -TD (ms), # initial -44.400 2 -2.000 .1835 -TD (ms), v-initial 42.382 10 3.825 .0033 '-TD (ms), final-v 99 900 6 6.874 .0005 •-TD (ms), final-c 92.500 17 13.420 <0001 •-TD (ms), final # 77.700 8 4.221 .0029 •-TD (ms), intervoc 49.950 7 1.871 .1036 195 E. K R (5;9) ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 43278.113 8655.623 4.736 .0009 23.681 .975 Residual 69 126101.138 1827.553 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. # initial 4 -41.625 56.871 28.435 v-initial 12 5.550 27.800 8.025 final-v 21 -3.171 29.607 6.461 final-c 15 19.980 54.573 14.091 final # 6 72.150 49.016 20.011 intervoc 17 -7.835 47.866 11.609 Interaction Bar Plot for TT-TD (ms) Effect: position Error Bars: 71 Standard Deviation(s) Row exclusion: all subject-timing.svd 150 T 1 ' 1 ' 125 " 100 --75--100 --125 -1 • ' : ' ' ' • # initial v-initial final-v final-c f inal* intervoc Fisher 's PLSDfor TT-TD (ms) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mean Diff. Crit. Diff. P-Value # initial, v-initial -47.175 49.239 .0601 # initial, finai-v -38.454 46.526 .1037 # initial, final-c -61.605 47.992 .0126 # initial, final # -113.775 55.050 .0001 # initial, intervoc -33.790 47.394 .1594 v-initial, final-v 8.721 30.862 .5747 v-initial, final-c -14.430 33.030 .3865 v-initial, final # -66.600 42.642 .0027 v-initial, intervoc 13.385 32.155 .4092 final-v, final-c -23.151 28.831 .1137 final-v, final # -75.321 39.479 .0003 final-v, intervoc 4.664 27.824 .7391 final-c, final # -52.170 41.196 .0138 final-c, intervoc 27.815 30.211 .0706 final #, intervoc 79.985 40.498 .0002 One Sample t-test Split By: position Hypothesized Mean = 0 Row exclusion: all subject-timing.svd Mean DF t-Value P-Value TT-TD (ms), Total 5.772 74 1.045 .2995 TT-TD (ms), # initial -41.625 3 -1.464 .2394 TT-TD (ms), v-initial 5.550 11 .692 .5035 TT-TD (ms), final-v -3.171 20 -.491 .6289 TT-TD (ms), final-c 19.980 14 1.418 .1781 TT-TD (ms), final # 72.150 5 3.606 .0155 TT-TD (ms), intervoc -7.835 16 -.675 .5094 196 F. A D (4;4) ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 2 4321.151 2160.575 1.841 .1858 3.683 .326 Residual 19 22292.209 1173.274 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. # initial 6 16.650 45 901 18.739 final # 7 52.329 26.200 9.903 intervoc 9 29.600 30.901 10.300 Interaction Bar Plot for TT-TD (ms) Effect: position Error Bars: ?1 Standard Deviation(s) Row exclusion: all subject-timing.svd # initial final # intervoc Fisher's PLSD for TT-TD (ms) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mean Diff. Crit. Diff. # initial, final # # initial, intervoc final #, intervoc P-Value -35.679 39.886 .0766 -12.950 37.785 .4819 22.729 36.130 .2036 One Sample t-test Split By: position Hypothesized Mean = 0 Row exclusion: all subject-timing.svd Mean DF t-Value P-Value -TD (ms), Total 33.300 21 4.387 .0003 -TD (ms), # initial 16.650 .5 .889 .4150 -TD (ms), final # 52.329 6 5.284 .0019 -TD (ms), intervoc 29.600 8 2.874 .0207 197 G. FL (4;9) ANOVA Table for TT-TD (ms) Row exclusion: all subject-t iming.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 137044.543 27408 909 21.397 <0001 106.985 1.000 Residual 84 107601.233 1280.967 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-t iming.svd Count Mean Std. Dev. Std. Err. # initial 17 -31.341 41.576 10.084 v-initial 9 11.100 37.231 12.410 final-v 13 10.246 20.993 5.822 final-c 19 33.300 40.022 9.182 final # 21 80.871 40.197 8.772 intervoc 11 -9.082 15.554 4.690 # initial v-initial final-v final-c final* intervoc FL Fisher 's PLSD for TT-TD (ms) Effect: position Significance Level: 5 % Row exclusion: all subject-t iming.svd Mean Diff. ait. Diff. P-Value # initial, v-initial -42.441 29.340 .0051 S # initial, final-v -41.587 26.223 .0022 S # initial, final-c -64.641 23.761 <0001 S # initial, final # -112.213 23.221 <0001 S # initial, intervoc -22.259 27.541 .1118 v-initial, final-v .854 30.863 .9563 v-initial, final-c -22.200 28800 .1291 v-initial, final # -69.771 28.356 <0001 S v-initial, intervoc 20.182 31.990 .2131 final-v, final-c -23.054 25.618 .0771 final-v, final # -70.625 25.118 <0001 S final-v, intervoc 19.328 29.158 .1910 final-c, final # -47.571 22.535 <0001 S final-c, intervoc 42.382 26.965 .0024 S final #, intervoc 89.953 26.490 <0001 s One Sample t-test Split By: position Hypothesized Mean = 0 Row exclusion: all subject-t iming.svd Mean DF t-Value P-Value -TD(ms), Total 21.460 89 3.883 .0002 -TD(ms),# initial -31.341 16 -3.108 .0068 -TD(ms), v-initial 11.100 8 .894 .3972 -TD(ms), final-v 10.246 12 1.760 .1039 -TD(ms), final-c 33.300 18 3.627 .0019 -TD (ms), final # 80.871 20 9.220 <0001 -TD(ms), intervoc -9.082 10 -1.936 .0816 198 H. RC (3;11) ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 4 , 42435.333 10608.833 3.906 .0078 15.625 .879 Residual 50 135793.532 2715.871 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. # initial 13 -17.931 53.721 14.900 final-v 3 0.000 0.000 0.000 final-c 3 11.100 19.226 11.100 final # 26 48.669 62.640 12.285 intervoc 10 9.990 16.085 5.087 Fisher 's PLSDfor TT-TD(ms) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mean Diff. Qit. Diff. P-Value # initial, final-v -17.931 67.045 .5935 # initial, final-c -29.031 67.045 .3886 # initial, final # -66.600 35.556 .0004 # initial, intervoc -27.921 44.028 .2086 final-v, final-c -11.100 85.466 .7953 final-v, final # -48.669 63.825 .1319 final-v, intervoc -9.990 68.905 .7721 final-c, final # -37.569 63.825 .2427 final-c, intervoc 1.110 68.905 .9743 final #, intervoc 38.679 38.950 .0516 One Sample t-test Split By: position Hypothesized Mean = 0 Row exclusion: all subject-timing.svd Mean DF t-Value TT-TD (ms), Total TT-TD (ms), # initial TT-TD (ms), final-v TT-TD (ms), final-c TT-TD (ms), final # TT-TD (ms), intervoc P-Value 21.191 54 2.736 .0084 -17.931 12 -1.203 .2520 0.000 2 • • 11.100 2 1.000 .4226 48.669 25 3.962 .0005 9.990 9 1.964 .0811 199 I. LR (5;6) ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 4 5390.877 1347.719 1.095 .3707 4.379 .310 Residual 45 55398.473 1231.077 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. # initial 7 -14.271 26.200 9.903 v-initial 3 -33.300 33.300 19.226 final-v 11 6.055 13.471 4.062 final # 17 -1.959 53.268 12.919 intervoc 12 5.550 12.962 3.742 Interaction Bar Plot for TT-TD (ms) Effect: position Error Bars: 71 Standard Deviation(s) Row exclusion: all subject-timing.svd # initial v-initial final-v final # intervoc Fisher's PLSD for TT-TD (ms) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mean Diff. Crit. Diff. P-Value One Sample t-test Split By: position Hypothesized Mean -# initial, v-initial 19.029 48.766 .4360 # initial, final-v -20.326 34.168 .2371 0 # initial, final # -12.313 31.736 .4387 Row exclusion: all subject-timing.svd # initial, intervoc -19.821 33.609 .2411 Mean DF t-Value P-Value v-initial, final-v -39 355 46.029 .0919 TT-TD (ms), Total -1.998 49 -.401 .6901 v-initial, final # -31.341 44.254 .1607 TT-TD (ms), # initial -14.271 6 -1.441 .1996 v-initial, intervoc -38.850 45.616 .0932 TT-TD (ms), v-initial -33.300 2 -1.732 .2254 final-v, final # 8.0 (3 27.345 .5580 TT-TD (ms), final-v 6.055 10 1.491 .1669 final-v, intervoc .505 29.499 .9727 TT-TD (ms), final # -1.959 16 -.152 .8814 final #, intervoc -7.509 26.645 .5731 TT-TD (ms), intervoc 5.550 11 1.483 .1661 200 J. MC (5;6) ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 4 12643.546 3160.887 2.859 .0398 11.438 .707 Residual 31 34268.661 1105.441 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. v-initial 8 20.813 30.507 10.786 final-v 14 35.679 42.252 11.292 final-c 2 33.300 47.093 33.300 final # 2 49.950 23.547 16.650 intervoc 10 -6.660 14.041 4.440 Interaction Bar Plot for TT-TD (ms) Effect: position Error Bars: ?1 Standard Deviation(s) Row exclusion: all subject-timing.svd v-initial final-v final-c final # intervoc Fisher's PLSD for TT-TD (rr.s) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mean Diff Crit. Diff. P-Value v-initial, final-v -14.866 30.054 .3209 v-initial, final-c -12.488 53.609 .6381 v-initial, final # -29.138 53.609 .2762 v-initial, intervoc 27.473 32.165 .0914 final-v, final-c 2.379 51.260 .9252 final-v, final # -14.271 51.260 .5742 final-v, intervoc 42.339 28.076 .0044 final-c, final # -16.650 67.810 .6201 final-c, intervoc 39.960 52.525 .1309 final #, intervoc 56.610 52.525 .0355 One Sample t-test Split By: position Hypothesized Mean = 0 Row exclusion: all subject-timing.svd Mean DF t-Value P-Value TT-TD (ms), Total 21.275 35 3.487 .0013 TT-TD (ms), v-initial 20.813 7 1.930 .0950 TT-TD (ms), final-v 35.679 13 3.160 .0075 TT-TD (ms), final-c 33.300 1 1.000 .5000 TT-TD (ms), final # 49.950 1 3.000 .2048 TT-TD (ms). intervoc -6.660 9 -1.500 .1679 201 K. .IL (4;7) ANOVA Table for TT-TD (ms) Row exclusion: all subject-timing.svd DF Sum of Squares Mean Square F-Value P-Value Lambda Power position 5 84457.462 16891.492 8 114 <0001 40.571 1.000 Residual 46 95758.488 2081 706 Means Table for TT-TD (ms) Effect: position Row exclusion: all subject-timing.svd Count Mean Std. Dev. Std. Err. # initial 9 -33.300 64.485 21.495 v-initial 8 -4.163 45.162 15.967 final-v 19 43.816 38.536 8.841 final-c 6 16.650 50.502 20.617 final # 2 166.500 47.093 33.300 intervoc 8 12 488 30.507 10.786 # initial v-initial final-v final-c final* intervoc JL Fisher's PLSDfor TT-TD (ms) Effect: position Significance Level: 5 % Row exclusion: all subject-timing.svd Mean Diff. Crit. Diff. P-Value # initial, v-initial -29.138 44.626 .1953 One Sample t-test # initial, final-v -77.116 37.163 .0001 S Split By: position # initial, final-c -49.950 48.404 .0434 S Hypotnesizea Mean = u Row exclusion: all subject-timing.svd # initial, final * -199.800 71.794 <0001 S # initial, intervoc -45.788 44.626 .0446 S v-initial, final-v -47.978 38.707 .0162 S Mean DF t-Value P-Value v-initial, final-c -20.813 49.599 .4027 TT-TD (ms), Total 19.852 51 2.408 .0197 v-initial, final # -170.662 72.606 <0001 S TT-TD (ms), # initial -33.300 8 -1.549 .1599 v-initial, intervoc -16.650 45.920 .4692 TT-TD (ms), v-initial -4.163 7 -.261 .8018 final-v, final-c 27 166 43.008 .2100 TT-TD (ms), final-v 43.816 18 4.956 .0001 final-v, final # -122.684 68.273 .0007 S TT-TD (ms), final-c 16.650 5 .808 .4560 final-v, intervoc 31.328 38.707 .1101 TT-TD (ms), final # 166.500 1 5.000 .1257 final-c, final # -149.850 74.987 .0002 S TT-TD (ms), intervoc 12.488 7 1.158 2849 final-c, intervoc 4.163 49.599 .8666 final #, intervoc 154.012 72.606 <0001 S 202 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0092349/manifest

Comment

Related Items