UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The Gestural organization of North American English /r/ : a study of timing and magnitude Campbell, Fiona Margaret 2004

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

Item Metadata

Download

Media
831-ubc_2005-0018.pdf [ 16.02MB ]
Metadata
JSON: 831-1.0091851.json
JSON-LD: 831-1.0091851-ld.json
RDF/XML (Pretty): 831-1.0091851-rdf.xml
RDF/JSON: 831-1.0091851-rdf.json
Turtle: 831-1.0091851-turtle.txt
N-Triples: 831-1.0091851-rdf-ntriples.txt
Original Record: 831-1.0091851-source.json
Full Text
831-1.0091851-fulltext.txt
Citation
831-1.0091851.ris

Full Text

T H E G E S T U R A L ORGANIZATION O F N O R T H A M E R I C A N E N G L I S H hi: A S T U D Y OF TIMING A N D M A G N I T U D E by FIONA M A R G A R E T C A M P B E L L  B.A., The University of British Columbia, 2002  A T H E S I S SUBMITTED IN PARTIAL F U L F I L M E N T O F THE REQUIREMENTS FOR THE D E G R E E OF MASTER OF A R T S in T H E F A C U L T Y OF G R A D U A T E S T U D I E S (Department of Linguistics)  T H E UNIVERSITY O F BRITISH C O L U M B I A December 2004  © Fiona Margaret Campbell, 2004  Abstract The syllable-based allophonic variation in the relative timing and magnitude of two gestures which has been observed for a number of complex segments, including III, /w/ and nasal consonants (Sproat and Fujimura, 1993; Browman and Goldstein, 1995; Krakow, 1999; Gick, 2003), provides the generalization that more anterior gestures tend to appear at syllable peripheries. However, because only two gestures are examined it is not clear how to characterize the gestures involved, nor whether timing offsets are categorical or gradient. North American English Ixl is a particularly variable and complex segment with three constrictions - tongue root (TR), tongue tip/blade/body (TB), and lips (Lip), perhaps the reason it has been the subject of numerous studies (Delattre and Freeman, 1968; etc.). That said, technological limitations have hindered the ability of articulatory studies to accurately examine the timing and magnitude of all three gestures of Ixl (Gick and Campbell, 2003). With the object of better understanding the nature and role of timing offsets within phonetics and articulatory phonology, this thesis reports the results of an experiment examining productions of Ixl by nine speakers of Canadian English. The present study uses a novel combination of M-mode ultrasound to measure lingual gestures and Optotrak to measure lip gestures - a significant improvement over previous methods, particularly with respect to temporal resolution (Campbell, Gick, & Namdaran, 2004). Despite variable tongue position/shapes across subjects, cross-subject results show a position-based reduction in gestural magnitude similar to that observed in previous studies (see Krakow, 1999); the Lip and T B gestures were reduced in syllable-final position and the TR gesture was reduced in syllable-initial position. Results also indicate that there is a pattern of front-to-back timing in syllable-initial position, with the Lip occurring before the T B , which occurs before the TR. In syllable-final position the order observed was TR and Lip, followed by TB. These findings are not wholly consistent with any of the theories advanced thus far to explain syllable-based allophonic variation. It is proposed that the relative magnitude of gestures is a better predictor of timing than the relative anteriority of a gesture or an assigned phonological classification.  ii  T A B L E OF CONTENTS Abstract  ii  Table of Contents  iii  List of Tables  v  List of Figures  vi  Acknowledgements  CHAPTER I  vii  Introduction  1  1.1 Syllable-based timing and magnitude effects within Articulatory Phonology .. 1 1.1.1 Articulatory Phonology 1 1.1.2 Syllable-based allophonic variation 2 1.2 Previous work: /r/ 4 1.3 Predictions for hi, timing and magnitude 6 1.4 The present study 10  C H A P T E R II  Methodology  12  2.1 Participants 2.2 Stimuli 2.3 Apparatus 2.4 Procedure 2.4.1 Data Collection 2.4.2 Ultrasound System 2.4.2 Optotrak System 2.4.3 Audio/Synchronization of Signals 2.5 Analysis 2.5.1 Timing Measures and Calculations 2.5.1.1 Ultrasound Data 2.5.1.2 Optotrak Data 2.5.2 Magnitude Measures 2.5.2.1 Ultrasound 2.5.2.2 Optotrak 2.5.3 Qualitative Observations 2.5.3.1 Tongue shape 2.5.3.2 Glottal insertion in the Resyllabifiable context  iii  12 12 13 14 14 15 17 18 19 19 21 22 24 24 25 27 27 27  C H A P T E R III  Results  29  3.1 Timing 3.1.1 Cross-subject Timing Results 3.1.1.1 Cross-subject timing results by position 3.1.1.2 Cross-subject timing results by articulator 3.1.2 Individual Timing Results 3.1.2.1 Individual timing results by position 3.1.2.2 Individual timing results by articulator 3.2 Magnitude 3.2.1 Cross-subject Magnitude Results 3.2.2 Individual Magnitude Results 3.3 Production Variation (Qualitative Results) 3.3.1 Results for tongue shape 3.3.2 Results for glottal insertion in the Resyllabifiable context  C H A P T E R IV  Discussion  42  4.1 Summary of Overall Quantitative Results 4.2 Comparison of results with predicted patterns 4.3 Proposal 4.4. Discussion of qualitative results for tongue shape 4.5 Implications for Articulatory Phonology 4.6 Potential problems with the experiment 4.7 Suggestions for future work  CHAPTER V  29 29 30 31 32 32 34 34 34 36 38 38 41  Conclusion  42 42 46 48 49 50 51  52  References  53  Appendix I: Individual subject information  57  Appendix Appendix Appendix Appendix  58 59 68 84  II: List of stimuli by vowel context III: Individual results for Timing IV: Individual A N O V A s testing for ^syllabification effects V: Individual results for Magnitude  iv  List of Tables Table 1.1 Summary of predicted categorization of gestures and predictions of relative timing by position Table Table Table Table Table Table  3.1 3.2 3.3 3.4 3.5 3.6  One group t-tests for differences between lingual gestures and Lip Unpaired t-tests for differences between T B and TR by position Summary of individual order of gestures by position Frequency of timing patterns by position Individual production variation: tongue shape and Lip gesture Percentage of glottals inserted in Resyllabifiable position  Table 4.1 Summary of predicted categorization of gestures and predictions of relative timing by position (with results)  v  9 31 31 33 33 40 41  43  List of Figures Figure 1.1 Gestural score and movement trajectories for American English III Figure 1.2 Association of 'seemingly consonantal' (C) and 'seemingly vocalic' (V) elements  3 8  Figure 2.1 Context for target Ixl Figure 2.2 Experimental Set-up Figure 2.3 B/M-mode Ultrasound Image of Ixl Figure 2.4 Frame showing M-mode trace of Ixl Figure 2.5 Placement of Optotrak markers Figure 2.6 Audio Synchronization Set-Up Figure 2.7 Locating the 0 point in the four signals Figure 2.8 Ultrasound timing calculation illustration Figure 2.9 Sample measures of timing Figure 2.10 Sample of Optotrak data illustrating selection of 0 point Figure 2.11 Selection of Optotrak Timing and Magnitude Measures Figure 2.12 M-mode image of Ixl with sample magnitude measurement locations for a) T B raising and c) TR retraction Figure 2.13 Plot of Vertical Lip Movement during Ixl Figure 2.14 Degree of Lip Approximation during Ixl Figure 2.15 Sagittal diagram of idealized tongue shapes for American Ixl Figure 2.16 Examples of a) no glottal stop, b) glottalization, c) full glottal stop  25 26 26 27 28  Figure 3.1 Cross-subject timing of achievement of T B and TR gestures (relative to Lip), by position Figure 3.2 Cross-subject differences in TR timing (relative to Lip) across positions Figure 3.3 Lip aperture across syllable positions Figure 3.4 Magnitude of T B gesture across syllable positions Figure 3.5 Magnitude of TR gesture across syllable positions Figure 3.6 Individual Lip aperture results Figure 3.7 Individual magnitude results for TB Figure 3.8 Individual magnitude results for TR Figure 3.9 Overlaid tracings of Ixl for A G L (contrasting vowels) Figure 3.10 Overlaid tracings of Ixl for MIY (contrasting position)  30 31 35 35 36 36 37 38 39 39  Figure 4.1 Illustration of predicted categories of gestures, based on the results  44  vi  12 14 15 16 18 19 20 21 22 23 24  Acknowledgments First and foremost, I would like to thank my supervisor, Dr. Bryan Gick, for his support throughout my undergraduate and graduate research career. Likewise Dr. Patricia Shaw. I also need to acknowledge the significant contribution of Ian Wilson to this research project as I don't think it would have been possible without him and certainly wouldn't have been as fun. Thanks Ian. Without the support of my committee members, Dr. Patricia Shaw and Dr. Sonya Bird, this thesis would be a far less interesting document, and most probably not as complete. I appreciate your willingness and ability to lend new perspectives to this work. Next I want to thank Eric Vatikiotis-Bateson, Shaffiq Rahemtulla, and Nahal Namdaran for assistance with equipment, data collection, and analysis. And lastly, I owe a great deal to the people who were there for late night crises, the ones who brought sustenance and made sure I left the lab periodically, those who listened to me ramble or read drafts for the millionth time, and those who simply had the good grace to put up with me while I was writing. Thank you to all my friends and family, especially my Mum and Da, Mike Young, Ramona McDowell, Jeff and Clare, Miranda Huron, and the rest at the office who tracked my progress and let out a big sigh of relief when I finished each and every chapter. All mistakes are my own. This research was funded by an NSERC Postgraduate Scholarship to Fiona Campbell and an NSERC Discovery Grant to Bryan Gick.  vii  1.0 Introduction Previous work on syllable-based allophonic variation has produced a number of studies which indicate that relative timing of the two component gestures in III, /w/ and nasals (see Krakow, 1999 for a summary, also Gick, 2003) is such that the more anterior gestures, those occurring physically further forward in the vocal tract, appear at syllable peripheries. In addition, allophonic variation of complex segments has been linked to position-dependent spatial reduction of gestures (e.g. Sproat and Fujimura, 1993), such that the less anterior gesture of the two gestures has a smaller magnitude in syllableinitial position, and the more anterior gesture shows a reduction in magnitude in syllable-final position. It is not clear what drives these phenomena, particularly whether they are categorical effects encoded in a speaker's phonology or phonetic effects resulting from perceptual or bio-mechanical requirements. Additionally, while timing and magnitude patterns have often been examined in tandem, it is not clear whether these are independent or linked. A better understanding of these issues will help clarify the role played by physical gestures in speech production and allow for further development of articulation-based theories.  1.1 Syllable-based timing and magnitude effects within Articulatory Phonology In the following section, a basic introduction to Articulatory Phonology, the theoretical framework in which this study is conducted, is provided. This is followed by a discussion of previous work on syllable-based allophonic variation for both timing and magnitude effects internal to various segments in English in section 1.1.2.  1.1.1 Articulatory Phonology The theory of Articulatory Phonology is based around phonological units called 'gestures'. A linguistic gesture is roughly defined as a coordinated set of motor events whose goal is a single articulatory target (e.g., jaw and tongue tip raising for l\l closure) (Kelso, Saltzman & Tuller, 1986; Browman and Goldstein, 1992, 1995). As a phonological unit, a gesture like an alveolar closure can be found in more than one segment (e.g. the tongue tip gesture in It/, Id/, In/ and III could be considered the same gesture) and segments may be composed of more than one gesture (for example English III generally has both tongue dorsum backing and a tongue tip/alveolar ridge closure). Gestures overlap temporally, both within a complex segment, and as coarticulation across separate segments, and the temporal organization of the gestures within a segment may vary. Much of the research conducted within the framework of Articulatory Phonology has been directed at a physiological definition of the syllable and explaining syllable-based allophonic variation. As such, Articulatory Phonology is a particularly useful theoretical paradigm in which to examine complex segments because it offers a principled method of describing and examining the components of the articulatory configuration as discrete elements. However, many questions remain as to both how gestures should be defined and classified in this framework, and how the spatial and temporal organization of gestures in running speech might best be described.  1  1.1.2 S y l l a b l e - b a s e d a l l o p h o n i c v a r i a t i o n A number of segments have been identified as having more than one component gesture and a number of studies have examined the relative timing of those gestures in English. This includes the relationship between the velum and the oral gesture associated with nasals (Krakow, 1989, 1999), the tongue tip and tongue dorsum for III (Sproat and Fujimura, 1993; Browman and Goldstein, 1995) and the lips and tongue dorsum for /w/ (Gick, 2003). In her 1999 review of the literature on the physiological organization of syllables, Krakow summarizes the findings of her 1989 study of American English nasals. She found that for /ml the velum gesture was larger and longer in syllable-final position (regardless of stress pattern). In addition, the velum preceded the lip gesture in syllablefinal position and followed it in syllable-initial position. In a 1993 study of North American English III, Sproat and Fujimura found that the tongue tip gesture had a greater magnitude and preceded the tongue dorsum gesture in syllable-initial position, and that the tongue dorsum had a greater magnitude and preceded the tongue tip gesture in syllable-final position. To account for this (and Krakow's observations of timing in nasals) they proposed that, based on the width of the constriction, a gesture could be classified as either [consonantal] (having an extreme constriction) or [vocalic] (having a less extreme constriction or an opening, as with the velum) and that the timing pattern observed was due to the fact that vocalic gestures are attracted to the syllable nucleus, while consonantal gestures are attracted to syllable margins. They add that consonantal gestures are 'stronger' in syllable-initial position and 'weaker' in syllable-final position while vocalic gestures show the opposite pattern. Browman and Goldstein (1995) followed this argument and found similar results for American English /l/, except their results indicated that in syllable-initial position, the gestures tended toward simultaneity with a greater timing offset in syllable-final position. This result is consistent with their 1992 proposal that there is a "single syllable-final organizational pattern in which the wider constrictions always precede the narrower constrictions" (p. 167). In addition, they observed syllable-final reduction of the tongue tip gesture in lil, In/, and III, as compared to syllable-initial position, and called this a 'general positional effect'. Note that an alternate view in which the effect is actually syllable-initial augmentation (Fougeron and Keating, 1995) is possible. However, since there are cases in which a gesture is apparently completely absent (e.g., Gick, 1999), and this is more straightforwardly explained as 'reduction to nothing' than 'augmentation from nothing', the effect will be referred to as reduction in this thesis. A visual aid to understanding the patterns observed, a 'gestural score' overlaid on movement trajectories of the articulators plotted over time for American English III, is given in figure 1.1 (figure from Gick and Goldstein, 2002, adapted from Browman and Goldstein 1995). Note the increased magnitude of the tongue tip gesture in syllableinitial position and the tongue rear in syllable-final position as well as the timing offset between the peaks of the two gestures in syllable-final position.  2  Figure 1.1 Gestural score and movement trajectories for American English l\l  (figure from Gick and Goldstein, 2002, adapted from Browman and Goldstein 1995) In a study of American English glides, Gick (2003) found evidence of a similar timing pattern in /w/, with the labial gesture occurring earlier than the tongue dorsum gesture in syllable-initial position, and later in syllable-final position (where it also displayed a reduction in magnitude). Given that both gestures have relatively wide constrictions, and neither seems inherently more 'consonantal' than the other, to account for the consonantal-like pattern of the labial gesture, Gick proposes that it is assigned the phonological category 'C-gesture' while the tongue dorsum gesture is assigned the category 'V-gesture'. The defining characteristics of a C-gesture are "(1) final reduction, (2) intermediate magnitude under ^syllabification, and (3) a tendency to occur farther away from the peak vowel." (Gick, 2003, p. 13) In a study of timing patterns for liquids in six different languages, Gick, Campbell, Oh, and Tamburri-Watt (2004b) found a slightly different pattern for l\l in Western Canadian English. In this case, a bigger difference in timing was seen between gestures in pre vocalic (preceding a word boundary) position than was seen in postvocalic (following a word boundary) position. However, the tongue tip gesture did precede the tongue dorsum gesture in prevocalic position and follow it in postvocalic position, as in previous studies. In addition, this study identified some cross-linguistic generalizations, including the presence of a dorsal constriction in postvocalic liquids across languages, some difference between gestural timing and/or magnitude in prevocalic vs. postvocalic positions, and that one can largely predict intergestural timing across languages based on more anterior gestures tending to be found at syllable margins. Due to the range of results across languages, they conclude that language-specific phonology must play a role, but that there are different kinds of phonetic-based factors that may play a role in different syllable positions. It is proposed that perceptual recoverability (e.g. Mattingly, 1981; Silverman, 1995; Kochetov, 2002; Chitoran, Goldstein, & Byrd, 2002) plays a bigger role in determining gestural timing patterns in syllable-initial position (based on the importance of onsets in lexical retrieval processes), whereas biomechanical factors such as the jaw cycle (e.g. Lindblom, 1983; Keating, 1983; MacNeilage, 1998) are more important in syllable-final position.  3  Unfortunately, because all of the previous studies examine timing and magnitude differences between only two gestures, they do not provide enough information to discriminate between the theories presented above. The observation that a less anterior gesture precedes a more anterior gesture in syllable-final position could be due to the need to preserve perceptual cues and no matter how many gestures there are, they will be realized in a strictly back-to-front order. Alternately, the same pattern may surface because each gesture is (phonologically) categorized as either a C-gesture or a Vgesture and in syllable-final position any number of V-gestures would precede any number of C-gestures. A similar case could be made for magnitude - an 'extreme' constriction may be reduced syllable-finally due to its less salient position, or because it is a C-gesture and C-gestures are reduced in that position. Lastly, it is not known whether timing and magnitude effects are part of the same phenomena or if they operate independently of one another. When only two gestures are studied with the effects judged on a 'present or absent' basis, they could equally be linked (only one need be realized but both may be) or acting independently (both or only one effect could be operative). It would be possible to disambiguate some of the predictions made by the above proposals regarding syllable-based variation if a segment with more than two gestures could be examined for relative timing and magnitude. Fortunately, such a segment does exist in English, the rhotic liquid Ixl. As it is produced in most North American dialects of English, Ixl is unusually complex (Hagiwara, 1995) in that, although the exact lingual configuration varies widely, it is generally composed of three independent oral constrictions: one between the tongue root and the pharyngeal wall (TR), one between the tongue tip/body and the palate or the alveolar ridge (TB), and one between the lips (Lip). This makes it uniquely suited for a test intended to determine whether the observed gestural timing and magnitude patterns are gradient (likely phonetic) or categorical (likely phonological) effects. Unfortunately, the difficulty of imaging and measuring movements of the lips, tongue, and pharynx simultaneously during speech has (thus far) proved to be prohibitive for such a study. Difficult though it may be to look at all the articulators involved in productions of Ixl, it is the subject of a considerable literature and several studies do address certain syllable position-based effects from an articulatory perspective. These findings are outlined in section 1.2  1.2 Previous work: Ixl North American English Ixl has been the subject of relatively intense scrutiny. A number of articulatory studies have examined various aspects of the production of North American English Ixl using a variety of methods. This is likely largely due to the relatively high degree of variation in tongue shapes that can be used in the production of the English rhotic. Much of the work on Ixl has focused on correlating tongue shape with the acoustic characteristics of Ixl (e.g. Delattre and Freeman, 1968; Hagiwara, 1995, Alwan, Narayanan, & Haker, 1997; Hashi, Honda, & Westbury, 2003; Guenther, Espy-Wilson, Boyce, Matthies, Zandipour, & Perkell, 1999; Espy-Wilson, 2004), or in cataloguing the variety of tongue shapes observed (e.g. Uldall, 1958; Delattre and Freeman, 1968; 1958; Lindau, 1985; Westbury, Hashi, & Lindstrom, 1998, Tiede, 4  Boyce, Holland & Choe, 2004). Nevertheless, several studies do bear directly on questions of syllable-based magnitude effects for the lips and tongue tip/blade and these are discussed in more detail below. The earliest simultaneous observations of all three gestures of Ixl came from the use of cineradiograms (x-ray films). In one such study, Delattre and Freeman (1968) present a relatively comprehensive examination of the variation observed in North American English Ixl across dialects and within individual subjects. They identify eight types ranging from the classically described 'retroflex' Ixl with the tip curled back, through a variety of 'bunched' configurations commonly ascribed to North American speakers, to a schwa-like tongue configuration, although the two most extreme types were only produced by British subjects. In addition to explicitly identifying the three areas of constriction, Delattre and Freeman claim that it is the constrictions observed that were responsible for the acoustic characteristics of Ixl (esp. F3 lowering), rather than tongue shape per se. With regards to syllable-position based variation, it was noted that lip rounding was more likely to occur in prevocalic pre-stress position and that the retroflex tongue shape (or the nearest thing to it) is used most frequently in a strong syllabic position (e.g. prevocalic pre-stress). These observations are supported by two other studies. Also using cineradiograms, Zawadzki and Kuehn (1980), observed both variation in tongue shape, and a difference between prevocalic (initiates a syllable) and post-vocalic (terminates a syllable) allophones. Specifically, "the prevocalic allophone was characterized by greater lip rounding, a more advanced tongue position, and less tongue dorsum grooving." (p.253). Gick (1999) used electro-magnetic midsagittal articulography (EMMA), a point tracking technique to look at the magnitude of the more anterior lingual gesture (tongue tip or tongue blade) across positions and found a reduction in syllable-final allophones. Based on a relatively simple probe-contact experiment, Hagiwara (1995) found that while 'tip up' (retroflex) was a stable tongue shape across positions for a given speaker, subjects who used a ' blade up' configuration in syllable-initial position were likely to use a 'tip down' configuration in syllable-final position. This would likely represent a reduction in magnitude in syllable-final position if measured. One study, Alwan et al (1997), used Magnetic Resonance Imaging (MRI) to look at Ixl and found no positional differences between syllable-initial and syllabic Ixl. This is almost certainly due to the fact that while MRI is uniquely able to image the whole vocal tract in 3D, it requires sustained productions of the target sound, something that is likely to obscure any timing or magnitude differences observed in natural speech. The observation of natural speech is further impeded by the subject's supine position during this procedure.  5  Given these findings and the various proposals for explaining syllable-based variation in complex segments, there are several potential patterns we might expect to find for English III. The following section examines these proposals and their predictions for timing and magnitude for III in more detail.  1.3 Predictions for Ixl, t i m i n g a n d magnitude Several theories come out of the studies discussed in section 1.1.2 which may account for the observed patterns; Sproat and Fujimura (1993) propose that gestures are organized based on a categorical degree of constriction (consonantal vs. vocalic gestures), Browman and Goldstein (1992, 1995) suggest that the relative width of the constrictions are important, Gick (2003) argues that the attested pattern is more likely driven by an abstract phonological categorization (C-gestures vs. V-gestures), and Gick et al (2004b) propose that the organization of gestures has more to do with physiological and phonetic factors that affect production and perception of multiple gestures in a syllable/segment than phonological categories of gestures. An additional theory is considered, that put forward by Carter (2002), which proposes the association of gestures with hierarchical phonological structures in syllable-final position (consonantal with the syllable coda and vocalic with the syllable rime). Each of these theories makes specific predictions about what patterns for timing and magnitude will look like for a segment with three gestures. These predictions are laid out in detail below, while taking into account previous work on the magnitude of the two more anterior gestures of III (lips and tongue tip/blade). Under the proposal of Sproat and Fujimura (1993), all of the gestures of III would likely be considered [vocalic] as none exhibit an 'extreme constriction'. Given this, all of the gestures should be attracted to the nucleus (presumably to the same degree) and therefore would be expected to be essentially simultaneous in both positions. With respect to the magnitude of the gestures, they can all be expected to pattern together as 'stronger' in syllable-final position. This does not seem very likely, given the results of Gick (2003) for /w/ and the observation that both the lip and tongue tip/blade gestures seem to reduce in final position (Delattre and Freeman, 1968; Zawadzki and Kuehn, 1980; Gick, 1999), a feature of [consonantal] gestures according to Sproat and Fujimura (1993). If the relative width of constriction drives timing offsets and gestural reduction patterns in Final position, with narrower constrictions closer to the edges of syllables and showing a reduction in magnitude, as suggested by Browman and Goldstein (1995), then the T B gesture of III would likely follow the TR gesture in this position. The actual relative degree of constriction at the Lip, as compared to the other two constrictions under consideration is not known, however since lip rounding has been observed to act in a manner similar to a consonant-like gesture (when occurring with a tongue dorsum retraction gesture) in previous studies (Gick, 2003), for the purposes of this discussion it will be considered to have a narrower constriction than is apparent at the TR. Any prediction about the relative constriction widths at the Lip and T B would be purely speculative, so for syllable-final position this study can predict that the order of gestures 6  would be TR, followed by Lip and TB (in either order or simultaneous, depending on the relative width of constriction) with reduction in the magnitude of both of these gestures. Browman and Goldstein (1995) claim that the generalizations about gestural organization are position-specific, applying to syllable-final position, and state that in syllable-initial position gestures should tend toward simultaneity. No claim is made regarding the syllable-initial reduction of the tongue rear backing apparent in their data (see figure 1.1 for an example). According to the proposal by Gick (2003), that gestures have an assigned languagespecific (phonological) status as either a C-gesture or a V-gesture, any of the three gestures could (theoretically) belong to either category. This means that any of the gestures may show the characteristics of a C-gesture (final reduction and a tendency to appear further from the peak vowel) and therefore be classified in that way. It would not be expected however, that a gesture would show a C-gesture characteristic in one position and a V-gesture characteristic in another. If the language-specific designation is consistent throughout a given language, (i.e. once categorized a gesture retains its class regardless of the segment it is in) then the results of Gick (2003) on /w/ suggest that the labial gesture will act as a C-gesture, and the results of Delattre and Freeman (1968) and Zawadzki and Kuehn (1980) (more lip rounding in prevocalic position) are consistent with this. The 'more advanced tongue position' of Zawadzki and Kuehn (1980), greater magnitude in syllable-initial position from Gick (1999), and the increased likelihood of a retroflex in Delattre and Freeman (1968) suggest that the more anterior lingual gesture is a C-gesture as well. This theory would therefore predict that for Ixl the Lip and tongue blade will pattern together in showing final reduction and occurring farther away from the peak vocalic element of the syllable. Crucially, if phonological categorization is responsible for the patterns observed in segments with two gestures, there should be no more than a two way distinction in timing and magnitude for the three gestures of Ixl. Another proposal with a phonological basis for explaining intergestural timing is Carter's (1999, 2002). In these studies he observed that in some dialects of British English a 'dark' III, is found in syllable-initial position. He suggests that liquids may be composed of a 'seemingly consonantal' element and a 'seemingly vocalic' element and that these are associated with different parts of a (hierarchical) phonological/prosodic structure in syllable-final position; specifically, the more consonant-like gesture is associated with the coda, and the more vowel-like with the rime, as in figure 1.2 (modified from Carter, 2002). This structural distinction would thus assure that the seemingly-vocalic gesture is phased prior to the seemingly-consonantal gesture in syllable-final position, as he observed across dialects. In syllable-initial position he claims that both gestures would be associated with a single node, the Onset, and that this lack of an intermediate level would allow for any order of realization, accounting for the variation observed across dialects.  7  Figure 1.2 Association o f seemingly consonantal' (C) and 'seemingly vocalic' (V) elements (modified from Carter, 2002)  Given the hierarchical nature of the base syllable structure he uses and the fact that the different types of gesture (apical/seemingly-consonantal and dorsal/seemingly-vocalic) are separately associated with the coda and the rime (respectively) in syllable-final position, the tongue blade would be expected to always follow the tongue root in syllable final position, since the rime will always (temporally) precede the coda. The labial gesture should pattern with one of the other two gestures, depending on its structural affiliation, however it is important to note that there appear to be only two timing positions available so there can be a maximum two-way distinction in timing, (i.e. either Lip patterns with T B or Lip patterns with TR, it shouldn't be different from both). A fourth explanation of the timing patterns observed comes from Gick et al (2004b), who consider the role of perceptual recoverability requirements along with biomechanical constraints on anterior gestures in light of jaw cycles. It is claimed that perception-based studies, when considered together, make two basic predictions about timing. The first is that gestures in syllable-initial position should prefer to be realized simultaneously because recoverability of consonants in this position is dependent on the transition to the vowel (Chitoran et al, 2002). The second prediction is that in syllablefinal position the perceptual recoverability of gestures would be improved if gestures occurred in the order of least anterior to most anterior (back-to-front), due to the tendency of more anterior sounds to obscure less anterior sounds when produced simultaneously (Kochetov, in press). A potentially confounding bio-mechanical factor, the jaw cycle, introduces other priorities for the timing and holds that speech production is based around the opening and closing of the jaw from an early age (MacNeilage, 8  1998) and consonants are fit into this cycle later. If this is the case, then more anterior consonants or gestures would be expected to occur when the jaw is at its highest position (of either opening or closing), thus restricting them to the time in the syllable furthest from the vocalic peak (which would presumably generally be when the jaw is most open). Less anterior gestures, being located closer to the 'hinge' of the jaw, would have less distance to make up should they occur when the jaw is more open and so they would not be expected to have the same restrictions. If timing were based around the jaw cycle, we would then expect to see timing distinctions in both syllable-initial and syllable-final positions with the degree of dependence on the jaw closure (more anterior, more dependent in general) determining the order in which gestures are realized. Gick et al (2004b) conclude that (cross-linguistically) the temporal organization of the gestures in syllable-initial position is largely dependent on the need for perceptual recoverability, but that in syllable-final position it is more likely that biomechanical factors (esp. the jaw cycle) play this role. If this is the case, then for Ixl we would expect to see all three gestures occurring simultaneously in syllable-initial position, and the tongue root preceding the tongue blade in syllable-final position, with the lips, being more anterior, likely patterning with the tongue blade. Table 1.1 summarizes the predictions of each of the theories with regards to Ixl. The categories each gesture would be assigned within a given theory are presented first, followed by the predictions for timing and magnitude (if applicable) in syllable-initial and syllable-final positions. Table 1.1 Summary of predicted categorization of gestures and predictions of relative timing by position. Sproat & Fujimura (1993) Browman & Goldstein (1995) Gick (2003) Carter (2002) Gick et al (2004b)  LIP [vocalic]  TB [vocalic]  TR [vocalic]  narrower constriction (than TR) C-gesture  narrower constriction (than TR) C-gesture  wider constriction  ?  seemingly consonantal less anterior  seemingly vocalic least anterior  anterior  V-gesture  9  INITIAL All three simultaneous, All three reduced All three simultaneous. Lip / TB > TR, TR reduced Any order: dialect dependent All three simultaneous  FINAL All three simultaneous, No reduction TR > Lip TR > TB TB & Lip reduced TR > Lip / TB Lip & TB reduced (Lip)/TR > TB/(Lip) TR > TB > Lip  1.4 The present study The goal of the present paper is to examine the syllable-based variation in timing and magnitude of gestures in the composite segment North American Ixl and thereby disambiguate the conclusions that can be drawn from the results of previous studies of the organization of gesturally complex segments. A combination of B/M-mode Ultrasound imaging (for lingual data) and Optotrak tracking (for labial position data) is used to track the movements of the three gestures concerned over real time at a natural speech rate. Ultrasound uses the "echo patterns of ultra-high frequency sound both emitted and received by piezoelectric crystals contained in a small...transducer" (Gick, 2002, p. 115). This signal travels through soft tissue or liquid, but 'bounces off air or bone. In (2-dimensional) B-mode, with the ultrasound transducer under the chin, the screen displays information about the superior surface of the tongue from the tongue root to near the tip (Stone, 1990) along the midsagittal plane (which is recorded to video). In combined B/M-mode, the B-mode midsagittal line is displayed and its movements along one or more trajectories (chosen by the researcher) are tracked and presented over what is essentially continuous time. This ultrasound data can be collected concurrently with external tracking of markers on the lips and face using Optotrak, "an optical 3D motion measurement system that tracks the positions of infrared light emitting diodes" (E. Baxter, Northern Digital Inc., personal communication, March 29, 2004) This system is generally accurate to within 0.1mm and can be collected at a frame rate much higher than standard video. Two earlier pilot studies, Gick and Goldstein (2002) and Gick and Campbell (2003), used B-mode Ultrasound and standard video (for lip tracking) to look at the relative timing and magnitude of the three gestures of Ixl. In both of these studies, syllable-based variations in the timing of the gestures of NA English Ixl were found, although the results were not consistent and were based on relatively few subjects. Perhaps the most important problems with these studies were the coarse temporal resolution (30 frames/second (fps)) that limited the accuracy of timing results, and the relatively poor spatial resolution, particularly for the lips. Since the labial gesture was used as a baseline and was often not measurable, the results were useful in determining the design of the present study, but were not themselves conclusive. Keeping the limitations of these studies in mind, it was generally the case that in syllable-initial position, Ixl had the lip gesture preceding the tongue body gesture, which in turn preceded the tongue root gesture. In syllable-final position, the most common result was a reduced lip gesture and no significant timing difference between the tongue root and tongue tip/body. Following up on these results another pilot study, Campbell, Gick and Namdaran (2004) used an improved method for imaging both the tongue and lips that is almost identical to that used in the present study. This study only included one speaker which is, unfortunately, not enough to draw meaningful conclusions about the organization of a sound as variable as Ixl. It did, however, confirm that a combination of B/M-mode ultrasound and Optotrak data signifies a significant improvement in technique from previous studies in that it allowed more precise imaging and comparison of the timing of all three gestures of Ixl. For this subject, in syllable-initial position the order of 10  gestures found was Lip preceding the tongue blade, preceding the tongue root, and in syllable-final position the tongue root and Lip were approximately simultaneous and preceded the tongue blade. Reduction of lip approximation and T B gestures in postvocalic position was also observed and is consistent with positional reduction effects seen elsewhere. These results are not entirely consistent with any of the above theories, however they should be taken only as preliminary results. The present study expands upon the Campbell et al (2004) pilot study, including results for nine subjects. A detailed description of the methodology used is given in Chapter 2, covering both quantitative and qualitative examination of the data. The results of the experiment are given in Chapter 3. The key findings of cross-subject statistical tests are presented for measures of timing and magnitude, as are the results of similar tests for each individual subject. Finally, some individual trends in Ixl production are presented based on qualitative examination of the data. Discussion of the hypotheses presented in section 1.3, in light of the results, can be found in Chapter 4, in addition to the broader implications of these findings, problems with the study, and directions for future work.  11  2.0 Methodology An experiment was conducted using Optotrak and B/M-mode ultrasound video to record the three gestures of North American English Ixl in pre-, inter-, and post-vocalic positions in a variety of vocalic contexts.  2.1 Participants Ten subjects participated in this study, 5 female, 5 male. Ages ranged from 22 to 36 and all were chosen on the basis of being native speakers of Canadian English, and preferably monolingual. Of those who reported speaking other languages, three spoke French as a second language and one spoke Cantonese. Six were from Vancouver, two more were from other parts of Western Canada, and two were from Ontario. Please see Appendix I for information about individual subjects. Subjects were paid for their participation in this experiment. One of the male subjects from Vancouver was excluded from the analysis based on poor ultrasound image quality.  2.2 S t i m u l i Stimuli were designed, as in previous studies (Gick and Campbell, 2003), such that Ixl was flanked by maximally similar vowels. The syllable position of the Ixl was varied such that it occurred in Initial (pre-vocalic) position, Final (post-vocalic) position, and in a context where it was Resyllabifiable (post-vocalic/word-final followed by a vowelinitial word), as in figure 2.1. The Final condition included /h/, a segment with no oral gestures, in order to disambiguate results which might show a possible resyllabification effect. All vowels present in Canadian English which can occur word-finally/in open syllables were used as vocalic contexts, because each vowel has the potential to obscure one or more of the gestures of Ixl and given the variability of production for this segment (Delattre and Freeman, 1968, etc.), the best context in which to observe the timing may also vary. This meant that III, Id, Id, lol, and lul served as context vowels, but that lax vowels could not be used as they do not occur word-finally/in open syllables.  Figure 2.1 Context for target Ixl The syllabic context for Ixl varied as follows: a. Initial Ixl b. final (Resyllabifiable) Ixl ...V,#RV,... ...V,R#V,...  c. Final Ixl ...V,R#hV,...  Where V ! = III, Id, Id, lol, Id  These stimuli were presented within the carrier phrase '...said " " each...' as an emphasized two syllable nonsense phrase (with the Ixl in the middle). This is as close to equal stress on the two syllables as it was possible to elicit. Recognizable mono-syllabic words were used where possible in order to prompt the appropriate vowel and to ease 12  the difficulty of the reading task. A complete list is given in Appendix II. An item from a randomized list of names preceded each test phrase and an item from a randomized list of times followed each test phrase. The presence of Ixl within the carrier sentences was minimized, particularly within syllables adjacent to the target phrase. Examples are given below: a. Casey said "hay ray" each evening. b. Mike said " hair A " each day. c. Joan said "hair hay" each month.  The one hundred and fifty sentences necessary for this experiment (ten tokens for each of 15 conditions (5 vowel contexts, 3 syllable positions)) were randomized, along with 20 sentences required for a separate experiment. These sentences were divided into six sets and several 'dummy' sentences were added to the beginning of each set in order to avoid list effects and bring the number of sentences in each set up to thirty. The full set of stimuli was therefore 180 sentences.  2.3 Apparatus Subjects were seated in a modified American Optical Co. ophthalmic examination chair (model 507-A) adjusted to maximize head stability and Ultrasound probe contact. This included a headrest located at the back of the head, just above the neck, and a forehead stabilizing head restraint, which was secured in a position where it was in contact with the subject's head, but not with enough pressure to cause discomfort. The ultrasound transducer, mounted on a mechanical arm attached to the chair, was secured in a position where it pressed against the subject's neck in such a way as to provide a consistent midsagittal (B-mode) image of the subject's tongue from root to tip. Twelve infrared light-emitting diode markers were attached to the subject and apparatus, as described in section 2.4.2. These were tracked by the three LED-sensing cameras in the Optotrak camera bar, which was approximately 2m in front of the subject at roughly head height. Subjects read from the 17inch monitor of a Macintosh G4 PowerBook computer positioned directly in front of the subject on a shelf below the camera bar. A hinged wooden clapper, which was used for synchronization of signals, as detailed in section 2.4.3, was located immediately to the left of the subject. Audio information was recorded via a cardioid microphone directly in front of the subject. Figure 2.2 (from Wilson, in preparation) provides a visual summary of the experimental set-up and more detailed descriptions of the equipment are given below.  13  Figure 2.2 Experimental Set-up (from Wilson, in preparation)  2.4 Procedure 2.4.1 Data collection Subjects were asked to read the stimuli sentences at a comfortable/natural rate. The sentences were presented as PowerPoint slides (96 pt font) on the monitor of the PowerBook G4 computer. Having the stimuli presented individually just over 2 meters from the subject and at approximately eye level has the effect of minimizing head movement (Stone, 2004) and preventing the subject from tilting the head up or down. This is particularly important because no correction was applied to the data based on head position. Work by Gick, Bird, and Wilson (2004a) showed no evidence of a correlation between head position and tongue depth, as viewed with the ultrasound, and so submental tissue compression (from contact with the Ultrasound transducer) was not factored in to measures taken either. A tone played for approximately 300 ms with the presentation of each new stimuli and the sentence was displayed for three seconds with one second of blank screen between each sentence. As described above, six sets of thirty sentences, lasting approximately 130 seconds/set were collected for each subject, although in some cases extra sets were collected as needed and as subject comfort allowed. Breaks between sets allowed for marker 14  position to be verified, data to be processed, and the subject to drink water. Including set-up and breaks, the experiment took approximately 1.5 hours (per subject) to complete. Each subject read the sentences in the same (randomized) order. The first sentence was excluded from the analysis, however since the subject did not know which sentence was the final one, end list effects were avoided and these tokens were included. In addition, prior to data collection, subjects were given a set of 15 practice sentences in order to familiarize them with the format of the sentences, get them settled into the chair, verify that the marker placements were secure, and allow time for adjustments to the position of the M-mode cursor lines.  2.4.2 Ultrasound System Ultrasound data were collected via an Aloka ProSound SSD-5000 Ultrasound machine using the combined B/M-mode setting. With a UST-9118 endo-vaginal 180-degree probe/transducer, this mode allows for 2-D imaging of the tongue surface from tip to root (B-mode) as well as tracking of the lingual movements along a certain trajectory over what can be considered essentially continuous time (M-mode). Three cursors (A, B, C) were positioned such that they intersected with constrictions visible on the BMode ultrasound image of the tongue in order to track the movements of the individual articulators shown in the M-mode signal (see Figure 2.3). Cursor A was placed on the tongue blade/body (TB) located between the tongue tip (TT) and the tongue mid (TM), B was placed on the T M , located in the middle region of the tongue, and C was placed on the tongue root (TR) which was often as far back as was visible throughout the utterance. Figure 2.3 B/M-mode Ultrasound Image of Ixl U5 C 1 5 P L  *  , F 3 27-JUL-04 14:02:10 ?2Hz 9118 6.0 DUfl:100* HI = 0 . 3 9  10cm-  Dislance  (cm)  B-Mode Ultrasound \ Image of AV j ^ s  5cm-  TB TR  10cm. I B  Ocm  RID G60  C5  Ah  M-Mode Ultrasound Image from mid Id  Raising Of T B I  Distance (cm) I 5cm - I M  Time of Current . 8-Mode Frame in M-Mode tracing  -  Lowering Of T M  -TR  O^RIO G60 C7 1: TONGUE  120  Backing of TR  fl2  0.5s  I' EG  Time (s)  15  1.0s  1.5s  Exact locations of the relevant lingual events varied across subjects, so the cursor's position was determined based on the location of movements resulting in constrictions in practice utterances observed on the monitor in real time, and in some cases frame-byframe. Once fixed, the position of the cursor was constant throughout the rest of the data collection session. This method highlights a significant departure from other methods such as lingual point tracking, in that tongue movement is measured at fixed constriction locations along the vocal tract rather than predetermined points on the surface of the tongue. This method is well suited to the observation of a segment with as much variation as Ixl in that information about the midline position of the rest of the tongue is available in addition to the measured area. This then allows us to judge basic tongue shape and identify places in which measurements of a relevant articulatory movement might otherwise be missed. Sweep speed (the rate at which the M-mode data is displayed/refreshed on the Ultrasound monitor), was set at the highest setting (1.5 seconds) in order to have the most detailed data possible available in the exported video. The range, the total real distance represented in the window on the screen, for both the B-mode and M-mode displays was set at 10cm. The 10cm of the M-mode display space was divided between the three cursors, meaning that each cursor could track movements over about 3.333cm. These settings allowed for the whole tongue surface to be imaged in B-mode and the full range of motion to be tracked in M-mode. This data was recorded to a DV cassette tape along with a synchronized audio signal (via a Yamaha 01V digital mixing console) and subsequently loaded onto a Macintosh G4 computer using Adobe Premiere 6.0. Single frames clearly showing the complete M-mode traces of each instance of Ixl to be analyzed were exported as PICT files and used for analysis in Adobe Photoshop 7.0.1. An example is given in Figure 2.4. Figure 2.4 Frame showing M-mode trace of Ixl UBC  ISRL  .  BSL  -  _ B  RIO  G60 CS  H  Y ,  .F 72Hz  x  TB gesture for Ixl  TR gesture for in  -RIO  G60  1 •TONGUE  c?  A;  120 f E D  16  3  ALOKA 27-JUL-04 14:02:10 9118 6.0 DUfl:100* MI = 0 . 3 9  Not all of the data collected was included in the analysis. For this study, only the tokens in the vocalic contexts Id and /a/ were selected for analysis. This decision was based on the way in which subjects produced Ixl and which vowel contexts were most suitable for observing and comparing all three gestures (i.e. T B , TR, Lip) across subjects. For these 9 speakers, movement associated with the T B gesture was visible in all three syllable contexts with the vowel /a/ and the TR gesture was similarly visible in the context of Id. Further, the timing and amount of Lip movement associated with Ixl could be observed with both these vowels because they are unrounded in Canadian English. The relative timing of the two lingual gestures could then be compared using the labial gesture as a reference point. Magnitude measures comparing the amount of movement across syllable positions were possible since the gestures were clearly visible for all subjects in all syllable contexts with these vowels.  2.4.2 Optotrak System An Optotrak 3020 system was used in conjunction with Collect (version 2.002), a Northern Digital Optotrak program which records the positions of 12 infrared-emitting diode Optotrak markers. As illustrated in Figure 2.5 (from Wilson, in preparation) markers 1-4 were placed on a modified pair of glasses worn by the subject and designed to track head position throughout the trials so that error and/or head correction could be calculated. Markers 5 and 6 were on the transducer (7cm and 14cm from the tip), in order to provide a stable line of reference in space. Marker 7 was mounted on a small piece of open cell foam and attached below the chin on the jawbone to provide information on jaw movement, and markers 8-11 were placed around the lips. The final marker (12) was placed on a clapper, which was used to synchronize the Optotrak with the Ultrasound (as described below).  17  Figure 2.5 Placement of Optotrak markers (from Wilson, in preparation)  Marker position data (x, y, & z coordinates with greater than 0.1mm accuracy) was recorded at 90Hz to a Micron Millennia X K U 333 computer. Data collected were monitored in real time for missing values (which could be due to an obstruction, a change in angle, and/or the marker falling off) during each 131s trial. In addition, prior to the first set and any head stabilization, a trial of 40 seconds was collected during which the subject rotated the head horizontally and then vertically and then fully rounded and stretched the lips in order to demonstrate the range of movement of both the head and lips. Raw data files containing information from each of the three cameras were converted within NDI Collect to reflect distances from a central point and these distances were extracted through a Northern Digital Data Analysis Package using the Data Display and Manipulation Program.  2.4.3 Audio/Synchronization of signals A Sennheiser M K H 416 P48 super-cardioid short shotgun condenser interference tube microphone was used to send the audio signal to a Yamaha 01V digital mixing console. Identical audio signals were then recorded synchronized with both Optotrak and Ultrasound data signals via the mixer. A hinged wooden 'clapper' with an Optotrak marker attached was used at the beginning and end of every trial to set a 0 point and an end point for the synchronization of the Optotrak and two Audio signals. The equipment synchronization system is (abstractly) represented in figure 2.6 below. The 0 point was 18  then used to calculate all time comparisons between the two visual signals. In addition, by comparing the time between the 0 point and the end point in the audio from the Ultrasound recording, audio from the Optotrak and the actual marker position Optotrak signal, it was possible to verify that no significant delay was introduced by the mixer. As discussed below, the Ultrasound video signal is assumed to be relatively well synchronized with the audio signal during the recording and subsequent transfer to computer via Adobe Premiere 6.0. Figure 2.6 Audio Synchronization Set-up Illustration  Video Signal  2.5 Analysis 2.5.1 Timing Measures and Calculations In order to combine two separate methods of data collection to inform a single question on the timing of articulatory movement/gestures, the synchronization of multiple signals is essential. In this case, there are four signals/sources of information concerned: (i) the Optotrak marker position information, (ii) the audio recorded to the Optotrak system, (iii) the audio recorded to the Ultrasound video, and (iv) the Ultrasound video itself. Figure 2.7 illustrates how a clapper with an Optotrak marker attached can be used to create a definable point in three of the four signals from which other measures of time can be taken. The observable interruption of each visual signal is represented, which demonstrates the method used to locate the 0 point in the signal.  19  Figure 2.7 Locating the 0 point in the four signals  0 point/clap Optotrak Audio  y  H i ' j Optotrak  Ultrasound Audio  H  Assume synch  M-mode Ultrasound i  mwmm Time  1  The clapper thus confirms that the data recorded by the external microphone is synchronized with the Optotrak data, a state that cannot simply be assumed, and allows synchronization with the Ultrasound audio. Because the clapper is not visible in the ultrasound signal, synchronization between the Ultrasound and Audio must be assumed for this experiment. This is not an optimal situation, and the degree to which this assumption is viable requires further examination in future studies. It is reported by Adobe online technical support that in Premiere, the video and audio should never have a difference of more than one third of a frame, and that the audio does not drift away from the video signal over time (Adobe Systems Incorporated, 2004), thus the maximum potential for error during this step is ± 11.12223ms. This, however, does not include the potential for error in the original capture/encoding to DV, which is not reported by the manufacturer.  20  2.5.1.1 Ultrasound Data In order to compare the relative timing of an event visible in the Ultrasound signal to an event captured in the Optotrak signal, it is necessary to know how far each is from a single point in real time. In this case, all timing measurements were based on the 0 point set by the clapper, a point which is not directly visible/identifiable within the Ultrasound signal. This being the case, three time intervals must be known for the Ultrasound in order to discover the time to the 0 point from any given event. These are, the time from a frame boundary before the clap up to the time of the clap (tl), the time between the initial frame boundary and one following the event (t2), and the time between the event and the following frame boundary (t3). The time from the frame to the clap (tl) and the time from the event to the final frame boundary (t3) can then be subtracted from the time between frame boundaries (tl). This method is summarized in Figure 2.8 and explained in more detail below. Figure 2.8 Ultrasound timing calculation illustration  In order to determine t l , the amount of time the clap time was off from the frame boundary at the start of the film, a few frames of audio from each video was exported as an Audio Interchange File Format (AIFF) file. These were opened in the audio analysis program Praat (Boersma and Weenink 2004, V4.2.12) and the time from 0 to the onset of the clap noise was measured. Frame numbers were recorded during extraction of the frames in which the Ixl gesture time was to be measured and, based on the number of frames multiplied by the duration of each frame (33.3667ms), the time from the absolute first frame boundary to the one visible in the M-mode signal following the Ixl (t2), was calculated. Frames were exported as PICT files and opened in Adobe Photoshop (v 7.0.1), which had been set up to automatically scale the images based on the Ultrasound's built-in timescale (visible along the M-mode window on the Ultrasound display). The time (t3) in milliseconds back from frame marker to the time when the identifiable movement associated with the gesture was completed was then measured using Photoshop's rectangular selection tool, as in Figure 2.9. 21  Figure 2.9 Sample measures of timing Time is measured from frame marker back to a)TB gesture and c)TR gesture. ALOKA UBC ISRL • F 3 27-JUL-04 13:59:26 6 72Hr 9118 6.0 DUfl:100k HI =0.39  Tune measure back to TR gesture from frame marker  i-RlO  G60 C7  1:T0NGUE  fi2  1  120 BEG  The point in the trajectory which was considered to be the point at which the gesture was 'completed' was the time at which it first reaches a point where it is within 5% of its peak level, based on the total range of movement of the articulator during speech.  2.5.1.2 Optotrak Data Files were exported from DAP in ASCII format and converted to Microsoft Excel (version 10.1.0) format for analysis. The dataset includes time in increments of 11.111s and vertical (x), horizontal (y), and depth (z) position data for each marker at each time. A sample of data collected from the upper and lower lip markers (#s 9 & 11) is given in figure 2.13 below. There were approximately 11790 samples in each dataset.  22  The Optotrak vertical (x) movement path of the marker on the clapper (marker #12) was used to identify the point at which the clapper hit, and this was set as time 0 for the data. Since the Optotrak data represents a sample every 11.111ms, the exact time at which the clapper hit would often not be one of the data points recorded. In addition, the clapper can bounce as much as 6mm, as established in a separate trial at 180Hz. Given these facts, the 0 point chosen was the first point after the clapper marker began its descent where there was a drop in its velocity of more than 5 mm/11.111ms (the velocity would be expected to continue increasing until the point when the clapper hits) and when the marker was within 6mm of its rest height. This selection method was supported by verification using the audio data recorded by the Optotrak data collection system. Figure 2.10 offers an example of the selection of a 0 point. Figure 2.10 Sample of Optotrak data illustrating selection of 0 point 162 X Lower Lip  Y  Lower Lip  163  -236.805  -99.389  164  -236.792  165  -2 3 6 . 7 9 ?  166  -2  167  -2  168  -2  169  -2  170  -2  171  -2  172  -2  173  -2  174  -2  175  -2  176  -2  177  -2  178  -2  179  -2 3 6 . 7 0 7 ;  180  -236.72^  Z Lower Lip  X  Clapper  Velocity  m m / l l m s  Trial  time  -2533.163  157.84  9,892  -99.388  -2533.18  146.328  11.512  1,8  -99.37R  2533,183  133.08  13,248  1.81111  2533,207  118,001  15,079  1,82222  2533.223  101.115  16,886  1.83333  2533.225  82.312  18.803  1,84444  2533.216  61,417  20.895  1.85556  X - vertical distance from cameral bar level Y - horizontal distance to the side from centre of camera bar Z - distance of depth from camera bar  2533,193  T , ' ,  52,001  ms  from  Opt 0  . . ,.  1.78889  9.416  1.86667  2533.183  51,444  11,11111!  0.557  1.87778  2533,195  50,562  22.22222  0.882  1.88889^  2533,181  50.046  33,33333  0.516  1,9  2533,188  50,024  44,44444  0,022  1.91111  2533.187  49.908  55.55555  0,116  1.92222  2533.168  49,878  66.66666  0.03  1.93333  2533,167  49.894  77.77777  -0.016  1.94444  2533,177  49.901  88,88888  -0.007  1.95556  2533.213  49.898  99.99999  0.003  1.96667  -99.25  -2533.241  49,888  111.1111  0.01  1.97778  -99.259  181  -236.715  -99.243  -2533.236  49,888  122.22221  0  1.98889  182  -236.693  -99,257 ^ - 2 5 3 3 . 2 3 3  49.879  133.33332  0.009  2  Audio 0 =1.866ms  Once the 0 point was set, it was possible to use the time calculated for Ixl from the Ultrasound signal to locate the approximate time of the Ixl in the Optotrak data. Based on this, the point of closest approximation of the lips was located (this process is described in more detail in section 2.4.5.2 below), and the time at which the lips came within 5% (based on the range of movement during speech) of the peak was recorded as the time the gesture was completed. See figure 2.11 below for an example of the Optotrak numbers which were used, and Figure 2.14 for a sample plot of the data points over time. The times obtained for the Lingual gestures were then subtracted from the corresponding times for the Lip gesture, thus giving a measure of the difference in timing between the Lip and T B , and the Lip and TR. These differences could then be compared as relative timing offsets from the Labial gesture indicating time differences between the T B and TR.  23  Figure 2.11 Selection of Optotrak Timing and Magnitude Measures D e g r e e o f a p p r o x Lip Peak T i m e ( m s ) Time Lingual Peak(s)  34,4997191 34,10 33,43 32,58 31.79 31.09 30. 28,47  Approx. is within 0.66mm of max (5% of range = 0.66mm  2 6 ,.5635g 56 25  [~  • 24 7.-:Sx,>^  70655.54849 70666.6596 70677.77071 70688.88182 70699.99293 70711.10404 70722.21515 70733.32626 70744.43737 70755.54848 70766.65959  24.54357164 24.44278552  77.7707 ^88181  24.57944051  24.91^X^22 25.263 25 26 26 27  [644  Max approx. (mm)  28.60166331 29.60932059 30.55296328  70.65554849 70.6666596 70.67777071 70.68888182 93 T i m e of 04  a c h i e v e m e n t of fingual g e s t u r e (in s e c o n d s )  15 26 37  J 4 ':' 6665959 >. ^/777707 70T78888181 j  T  i  70.79999292 Gesture completion time (ms)  03 P14 25 f36  70855.54847 70866.65958 70877.77069 70888.8818 70899.99291  70.81110403 70.82221514 70.83332625 70.84443736 70.85554847 70.86665958 70.87777069 70.8888818 70.89999291  2.5.2 Magnitude Measures Magnitude was measured for both Lingual and Labial gestures. Similar methods to those used for determining timing were employed to extract information about the relative size of the gestures (from an arbitrary point) across syllable positions. Because only the relative magnitude across positions (positional reduction/augmentation) for a single gesture within a single vocalic context was of interest, no claim is made about the absolute magnitude of these gestures/constrictions.  2.5.2.1 Ultrasound The same Ultrasound frames that were used for the timing measures were used for magnitude measures although magnitude measures were taken at the time of the absolute peak (not at the time of the timing measure). Photoshop was used to scale the images so that actual distances, based on the Ultrasound display's scale, could be measured. Exactly the same scaling procedure was applied to all of the images for all of the subjects. The magnitude of gestures was measured as the distance up from the border of the window (an arbitrary but consistently identifiable point) to the peak of the  24  movement visible in the M-mode track for TB and TR, as in fig 2.12. The relative magnitude across positions could then be compared.  Figure 2.12 M-mode image of Ixl with sample magnitude measurement locations for a) TB raising and c) TR retraction. UBC ISRL . Y ALOKA ,F 3  ESL  72Hz  27-JUL-04  14:02:10  9118 6.0 DUPn 1 0 0 * HI =0.39  0 6 0 C5  RIG  Magnitude measure forTB gesture  Magnitude measure for TR gesture ywniilliiu rmiKlm  t-R10  ,  G60 C 7  1=T0HGUE  ft2  120 BEG  2.5.2.2 Optotrak Magnitude of the lip gesture was determined by finding the value for the extreme of approximation of the lips nearest the time of the Ixl (based on lingual timing data) in the Optotrak data. This is illustrated in figure 2.11 above and samples of vertical lip movement and degree of approximation during Ixl are given in figures 2.13 and 2.14 below. Degree of approximation was based on the Euclidean distance between the upper lip and lower lip markers (#9 & #11) as it more accurately reflects actual approximation than only vertical measures. This was calculated for the data using the vertical (x) and depth (z) dimension measures with the following equation: Euclidean Distance Equation:  d = V[(x -x ) + (z -z )M 2  UL  LL  UL  LL  25  Figure 2.13 Plot of Vertical Lip Movement during Ixl Lip H e i g h t O v e r T i m e - 1 8 5 -i  r  i  ;  —r—.  :  -195  E  1  2  3  4  5  6  7  8  9  10 11 12 13 14 15 16 17 18 19 2 0 21 22 23 2 4 25 2 6 2 7 28 2 9 30 31 32 33 3 4 35  Frame  Figure 2.14 Degree of Lip Approximation during Ixl (based on Euclidean distance) Lip A p p r o x i m a t i o n f o r / r /  12  I  : 1  2  3  4  5  6  7  8  9  10 11 12 13 14 15 16 17 18 19 20 21 22 23 2 4 2 5 2 6 2 7 2 8 2 9 3 0 31 32 3 3 3 4 35  Frame  26  2.5.3 Qualitative Observations 2.5.3.1 Tongue shape Although this was not the focus of the present study, because the data were available and tongue shape may impact on the measures taken, the data were examined and categorized (subjectively) for general tongue shape. Despite some evidence that there exists a continuum of tongue shapes possible for Ixl (Alwan et al, 1997; Westbury et al, 1998) for this study a simple classificatory paradigm was used. Following Hagiwara (1995), the designations 'tip up', 'blade up', and 'tip down' were used, as illustrated in figure 2.15 (modified from Hagiwara, 1995). Because this study makes use of Ultrasound data, which has limited application for imaging the tongue tip due to interference from air in the sublingual cavity, if either the whole tip of the tongue was visible and curled upward or if the last visible part of the tongue closest to the TT curved upward the token was classed as 'tip up'. If the tongue mid was not lowered significantly and/or the full anterior region of the tongue could be seen to be rounded and pointing down, the token was classed as 'tip down', whereas if there was a significant depression of the tongue mid and the anterior region was raised with the tip angled down, the token was classified 'blade up' Figure 2.15 Sagittal diagram of idealized tongue shapes for American Ixl (Modified from Hagiwara, 1995, p.97)  2.5.3.2 Glottal insertion in the Resyllabifiable context Tokens in the Resyllabifiable context were also examined for the insertion of a glottal stop between the Ixl and the following vowel. This was not controlled for in the experiment, other than by asking participants to speak naturally. Spectral analysis was 27  done using Praat (Boersma and Weenink 2004, V4.2.12). Three categories were defined, as illustrated in figure 2.16, (i) tokens with no evidence of a glottal stop, (ii) tokens which had some glottalization realized on the Ixl and/or the vowel but with continuous phonation, and (iii) tokens with a full glottal stop. Tokens with glottal insertion were not excluded from the analysis, and the position was designated 'Resyllabifiable' rather than 'Resyllabified' for this reason. Figure 2.16 Examples of (a) no glottal stop, (b) glottalization, (c) full glottal stop (speaker: BWG, utterance: 'hair A') a) 0.4991  -0.01424 -0.4931 5000Hz  OHzl 18.591445  18.591445  V i s i b l e part 0 . 7 1 5 1 7 0 s e c o n d s  19.30661  10.022705  b) 15.214272 0.4991  -0.0713 -0.498 5 0 0 0 Hz!  0 Hz  14.348104 1  14. 310 I U t  V i s i b l e part 0 . 8 6 6 1 6 8 s e c o n d s  5.214273  14.115048  C) 9.956245  9.013149  6.360059  28  3.0 Results In this chapter the results of the experiment are given and the statistical treatment of the data is described. Cross-subject and individual patterns are presented for both timing and magnitude differences, followed by qualitative observations about the variation in the speakers' productions of Ixl.  3.1 Timing This section deals with the effect of syllable position on the relative timing of gestures. Cross-subject statistics and some individual patterns are reported. A significance level of p < .05 applies throughout.  3.1.1 Cross-Subject Timing Results Data for all 9 subjects were combined and t-tests were calculated to test for differences in timing between the achievement of the T B and TR gestures relative to the Lip gesture in Initial (e.g. haw raw), Resyllabifiable (e.g. har awe), and Final positions (e.g. har haw). In addition, one group t-tests were calculated to test for significant differences between the Lip and lingual gestures in each position (using a hypothesized mean of 0 for Lip values). The overall order of gestures found is Lip > T B > TR (front to back) in Initial position, TR > Lip > T B in Resyllabifiable position, and T R / Lip > T B in Final position. Figure 3.1 illustrates the timing differences found and the results of the t-tests are given in Tables 3.1 and 3.2. One group t-tests are used to test for differences between the timing of the T B and Lip and the TR and Lip because the timing of these gestures was calculated based on a mean of 0 ms for Lip. Unpaired t-tests were used to test for differences between the T B and TR because in this case there are two mean numbers (each of which is relative to Lip) which must be compared. Finally, A N O V A s were used to compare the differences in timing between positions for each articulator. The timing of the TR gesture in Final and Resyllabifiable conditions pattern together, and both happen much earlier (relative to the lip gesture) than the TR gesture in Initial position, as shown in figure 3.2.  29  3.1.1.1 Cross-subject timing results by position Figure 3.1 Cross-subject timing of achievement of TB and TR gestures (relative to Lip), by position (Error bars indicate 95% confidence interval, 0 = time of Lip gesture)  Initial  TR  Resyl  TB  Final  -30  -20 -10  0  —i—  —i—  —i—  —i—  10  20  30  40  50  Time Difference (ms)  N.B.: Results shown in this figure are intended to allow visual comparison of means only; statistics are based on separate t-tests. As shown in figure 3.1 (and tables 3.1 and 3.2 below), in Initial position there was a significant difference in timing between the Lip and the TB (p < .0001) with the Lip gesture preceding the TB gesture by an average of 20.804 ms, as well as a significant difference between the Lips and TR (p < .0001) with the Lip preceding the TR by an average of 36.12 ms. Based on this, the TB preceded the TR by an average of 15.316 ms (p = .0016). In the Resyllabifiable position, there was a significant difference in timing between the Lip and the TB (p < .0001) with the Lip preceding the TB by an average of 28.287 ms. TR was also significantly different from Lip (p = .0043), preceding it by an average of 10.996 ms. TR therefore preceded TB by an average of 39.283 ms (p < .001). In contrast with the Initial position order (Lip > TB > TR), the TR precedes the Lip and TB in Resyllabifiable position, where the order is TR > Lip > TB. In the Final position there was a significant difference between the Lip and TB gestures (p < .0001) with the Lip preceding the TB by an average of 27.601 ms. The difference between the Lip and T R was not significant (p = .0911) in this position, but the difference between the TB and the TR was (p < .0001) with the TR preceding the TB by an average of 33.886ms. The order of the gestures in Final position, TR/Lip > TB, seems to be different from but bears a close resemblance to that of the Resyllabifiable position (TR > Lip > TB), although see discussion of this in section 3.1.1.2 below.  30  Table 3.1 One group t-tests for differences between lingual gestures and Lip (Lip = hypothesized mean of 0) (Significance level p < .05) MEAN 20.804 36.12 28.287 10.996 27.601 6.284  Initial Lip/TB Initial Lip/TR Resyl Lip/TB Resyl Lip/TR Final Lip/TB Final Lip/TR  P VALUE p < .0001 p < .0001 p < .0001 p = .0043 p < .0001 p = .0911  Table 3.2 Unpaired t-tests for differences between T B and TR by position (Significance level p < .05) TB M E A N 20.804 28.287 27.601  Initial TB/TR Resyl TB/TR: Final TB/TR  TR M E A N 36.120 -10.996 -6.284  M E A N DIFFERENCE 15.316 39.283 33.886  PVALUE p = .0016 p < .0001 < .0001  3.1.1.2 Cross-subject t i m i n g results by articulator The timing of the T B gesture relative to the Lip gesture is relatively constant across positions; the Lip always precedes the TB. The TR gesture, however, follows the Lip in Initial position and shows the reverse order, TR then Lip, in Final and Resyllabifiable positions, as shown in figure 3.2. Figure 3.2 Cross-subject differences in TR timing (relative to Lip) across positions. (Error bars indicate 95% confidence interval, 0 = time of Lip gesture)  I  -20  ,  ,  -10  ,  1  0  ,  !  10  ,  (  ,  20  !  30  Time Difference Cms)  31  ,  !  40  ,  1  50  Significant differences were found in an A N O V A of the differences in time of achievement of the TR gesture relative to Lip across positions [F (2, 290) = 54.614; p < .0001]. Fisher's PLSD post hoc analysis revealed significant differences between the Initial and Final conditions (p < .0001) and between the Initial and Resyllabifiable conditions (p < .0001) for the TR. Despite the above finding that the TR was significantly different from Lip in Resyllabifiable position but not in Final position, no significant difference was found between the Final and Resyllabifiable contexts (p = .3462) for the TR. (Recall that the insertion of glottal stops between the Ixl and the following vowel (which could prevent resyllabification) was not controlled for so the lack of resyllabification effect is likely due to this.) An A N O V A of the differences between the times of achievement of the T B gesture relative to Lip in the three positions did not provide significant overall results [F (2, 198)= 1.073, p = .3441].  3.1.2 Individual timing results This section summarizes individual timing patterns. As above, t-tests were carried out in order to ascertain the order of gestures (the timing differences for an articulator within a given syllable position) and A N O V A s were used to test for resyllabification effects (the difference between syllable positions for a given articulator). Data from two subjects (BWG and DSL) could not be examined for most individual timing differences due to the frequent lack of an apparent labial gesture in Final and Resyllabifiable positions in the context of the vowel /a/ (har#haw or har#awe). Initial position and the post-vocalic TR data for these subjects were included in the analysis.  3.1.2.1 Individual timing results by position The individual order of gestures by position is summarized in Table 3.3. Detailed numerical results are available in Appendix III. The frequency of each timing pattern in each position is given in Table 3.4, highlighting the amount of variability across individual speakers. Despite this variation, it should be noted that for all of the subjects the Lip precedes the TR in Initial position and the TR precedes the T B in both Resyllabifiable and Final positions.  32  Table 3.3 Summary o ' individual order of gestures by position SUBJECT Overall/Combined AGL ASN BWG CRT DCM DSL LIN MIY PTM  INITIAL Lip > T B > TR Lip > T B / T R Lip > T R / T B Lip > T B / TR Lip > T B > TR Lip > T B > TR Lip > T B > T R T B / Lip > TR T B > Lip / TR Lip > T R / T B  RESYL TR > Lip > T B TR > Lip > T B Lip / T R > T B Lip > T R TR > Lip > T B TR > Lip > T B Lip / T R TR > T B / Lip TR > Lip > T B Lip / T R / T B but TR > T B  FINAL TR / Lip > TB TR > Lip > T B TR > Lip > T B Lip > TR TR / Lip > TB TR / Lip > TB Lip > TR TR > Lip / TB TR > Lip / TB Lip > TR > T B  N.B.: The order in cases where there is not a significant difference between the timing of two gestures reflects the mean times of the gestures.  Table 3.4 Frequency of timing patterns by position Position Timing Frequency / 9 INITIAL Lip>TB >TR 3 Lip >TR/TB 4 TB/Lip > T R 1 TB > Lip/TR 1 RESYL TR > Lip > T B 4 TR > Lip/TB 1 Lip/TR > T B 1 Lip > T R 1 Lip/TR 1 Lip/TR/TB but T R > T B 1 FINAL TR/Lip > T B 2 TR > Lip >TB 2 Lip > T R 2 TR > Lip/TB 2 Lip > T R > T B 1  33  3.1.2.2 I n d i v i d u a l timing results by articulator There was very little evidence of resyllabification in the timing patterns for the TB and TR gestures individually. Full numerical results of A N O V A s testing for significant differences between Initial, Resyllabifiable, and Final positions for the T B and the TR individually are given in Appendix IV. For the TR (relative to Lip), 7 of the 9 subjects showed no significant difference in timing between the Resyllabifiable and Final conditions. Of the two subjects who did not fit this pattern, Fisher's PLSD post hoc analysis indicated that for P T M , T R in Resyllabifiable position was significantly different from both Initial (p = .0019) and Final (p = .0347) which were not significantly different from each other (p = .1885), and for B W G there was no significant difference in TR timing between Resyllabifiable and either Initial (p = .3019) or Final (p = .1075), although these were significantly different from each other (p = .0083) Information on resyllabification of the T B (relative to Lip) was more limited and variable. Due to the apparent lack of a Lip gesture in post-vocalic positions with /a/ (the context that T B was measured in) there was insufficient data to calculate timing differences for 2 subjects (BWG and DSL). In addition, individual A N O V A s (see Appendix IV) did not yield significant results for another three subjects (ASN, CPT, and PTM). This being the case, no reliable information is available for 5 of the 9 subjects. Two of the four remaining subjects (DCM and MIY) showed no significant difference in timing between the Resyllabifiable and Final conditions, similar to results for TR. One subject (AGL) had no significant difference between Initial and Resyllabifiable (p = .5221) and one (LIN) had no significant difference in T B timing between Resyllabifiable and either Initial (p = .2840) or Final (p = .1884), although these were significantly different from each other (p = .0139).  3.2 Magnitude This section presents results on the effect of syllable position on the magnitude of gestures. Individual and cross-subject statistics are reported. A significance level of p < .05 applies throughout.  3.2.1 Cross-Subject Magnitude Results Data for all 9 subjects were combined and A N O V A s were calculated (based on nonnormalized data) in order to determine if significant differences in the magnitude of gestures in the three syllable positions were present. For two subjects no Lip gesture was observable in the post-vocalic conditions but data for the Initial condition was included in the analysis. As shown in figures 3.3 - 3.5, a significant difference was found between Final/Resyllabifiable and Initial positions for all three articulators; however, no significant differences between the Final and Resyllabifiable conditions were evident.  34  In Initial position both the T B and Lip gestures have a greater magnitude than in Final or Resyllabifiable positions, while the TR gesture shows the opposite pattern and is reduced in Initial position, as compared to Final and Resyllabifiable position. Figure 3.3 Lip aperture across syllable positions (Error bars indicate 95% confidence interval) 30  Initial  Resyl  Final  Position  N.B.: For the Lip measures it is 'aperture' (the distance between the lip markers) which is represented, such that, in the above figure, the Initial condition has the most closure. As can be seen in figure 3.3, overall results indicate significant variances in the magnitude of the Lip gesture across positions [F (2, 539) = 105.042; p < .0001]. Fisher's PLSD post hoc analysis indicated that the degree of Lip approximation was significantly greater in the Initial condition than in the Final or the Resyllabifiable condition (p < .0001). No significant difference was observed between the Final and Resyllabifiable conditions (p = .4329). Figure 3.4 Magnitude of T B gesture across syllable positions (Error bars indicate 95% confidence interval) 20 n  1  Initial  Resyl  Final  Position  As can be seen in figure 3.4, overall results indicate significant variances in the magnitude of the T B gesture across positions [F (2, 276) = 4.963; p = .0076]. Fisher's PLSD post hoc analysis indicated that the TB gesture was significantly greater in Initial position than in the Final (p = .0020) or Resyllabifiable (p = .0461) conditions.  35  The mean for the T B gesture was slightly higher for the Resyllabifiable condition (Mean = 16.270mm) than the Final condition (Mean = 15.745mm) but the difference was not significant (p = .5978). Figure 3.5 Magnitude of TR gesture across syllable positions (Error bars indicate 95% confidence interval)  Initial  Resyl  Final  Position  Overall results indicate significant variances in the magnitude of the TR gesture across positions [F(2, 298) = 10.729; p < .0001 ]. Fisher's PLSD post hoc analysis indicated that the magnitude of the TR gesture was significantly less in Initial position than it was in Final (p < .0001) or Resyllabifiable (p = .0003) conditions. There was no significant difference between the magnitudes for Final and Resyllabifiable conditions (p = .5978), which had a mean difference of only -0.269mm.  3.2.2 I n d i v i d u a l magnitude results The data on magnitude were also examined for syllable position-based differences for each subject individually. A summary of the results for Lip, T B and TR are presented in figures 3.6 to 3.8. For the more detailed numerical results of individual ANOVAs, see Appendix V. Notably, the Lip and TR results are relatively consistent across subjects while T B results are more variable. Figure 3.6 Individual Lip aperture results (Error bars indicate 95% confidence interval) 35  n  Subject  36  As can be seen in figure 3.6, for all 9 subjects the mean Lip approximation was significantly greater in Initial than in either Final or Resyllabifiable position. Fisher's PLSD post hoc analysis indicated a significant difference in Lip approximation between Final and Resyllabifiable positions for only one speaker (AGL, p = .0002) who has a greater average closure for Resyllabifiable than for Final. Figure 3.7 Individual magnitude results forTB (Error bars indicate 95% confidence interval)  AGL  ASN  BWG  CPT  DCM  DSL  LIN  MIY  PTM  Subject  As can be seen in figure 3.7, six of the 9 subjects had a significantly greater TB gesture in Initial position, as compared to Final and Resyllabifiable contexts (see Appendix IV for full numerical results). Of the three subjects who do not display this pattern, the individual A N O V A for B W G was not significant overall [F (2, 26) = 1.814; p = .1830] and Fisher's PLSD post hoc analysis indicated that the other two (ASN and MIY) show the opposite effect. T B was significantly smaller in Initial position than in Final or Resyllabifiable for these subjects (ASN p < .0001, MIY Final p = .0166 Resyl p = .0020). Two subjects had significant differences between the Final and Resyllabifiable contexts, for A G L Resyllabifiable T B was larger than Final (p = .0134) and for PTM Final was larger than Resyllabifiable (p = .0432).  37  As can be seen in figure 3.8, of the 9 subjects, 7 had significantly smaller TR gesture in Initial position than in Final or Resyllabifiable contexts. For the two subjects who do not fit this pattern (ASN and BWG), the results of the individual A N O V A s were not significant overall (ASN [F (2, 32) = .237; p = .7901J and B W G [F (2, 26) = .4671; p = .4671 J) but the means for Initial position are slightly lower for both subjects than the means for Final or Resyl (ASN F = 10.308mm, I = 9.883mm, R = 10.164mm; BWG F = 17.664mm, I = 17.256mm, R = 17.956mm) which is consistent with the pattern for other subjects. Fisher's PLSD post hoc analysis showed that one subject had a difference between the Resyllabifiable and Final contexts: LIN had a larger TR gesture in Final position (p = .0360).  3.3 Production Variation (Qualitative Results) 3.3.1 Results for tongue shape As mentioned in the previous section, the focus of this study was not tongue shape. However, to provide a picture of qualitative variation across subjects, the data were categorized as one of three tongue shapes and compared across positions. Generally speaking, for these 9 subjects, the tongue shape for Ixl was composed of a raised tongue tip or blade, a depression of the mid section of the tongue (along the mid-sagittal plane), and backing of the TR. Very few tokens appeared to involve extreme retroflexion, but the tongue configuration was usually not what would be straightforwardly considered an American 'bunched' Ixl either. An illustration of the two most common tongue shapes found, tip up and blade up, is given in figure 3.9. Other researchers have found similar articulatory configurations for Ixl (e.g. Hagiwara, 1995; Westbury et al, 1995/8; Tiede et al, 2004) and have noted that Ixl production does not fall into two neat categories of 'retroflex' and 'bunched' (Delattre and Freeman, 1968).  38  Figure 3.9 Overlaid tracings of Ixl for AGL (contrasting vowels) (i) tip up (from /a#ra/), (ii) blade up (from /e#re/).  Two factors seemed to play a role in determining which variant was observed within a given context for a given speaker: vowel context and syllable position. An example of the variation across vowel context (/ra/ vs. /re/ for AGL) can be seen above in figure 3.9 and in figure 3.10 below an example from another speaker (MIY) illustrates the contrast in tongue shape between Initial and Final positions for the vowel /a/. Speakers were not always consistent in the tongue shape they produced in a given context but it was generally the case that most tokens (at least 75%), fell into a single category. Recall that qualitative tongue shape was judged impressionistically by the author and no measures were taken. Figure 3.10 Overlaid tracings of Ixl for MIY (contrasting position) (i) tip down (from /ar#ha/), (ii) tip up (from /a#ra/)  39  Qualitative observations of individual variation with regards to tongue shape category and the presence or absence of the Lip gesture post-vocalically (in Resyllabifiable or Final position) are summarized in table 3.5. It is also important to note that not every speaker consistently produced the same tongue shape within a position; exceptions to the shapes reported in table 3.5 were limited in number but were present. Contrary to the finding of Hagiwara (1995) the most stable tongue shape across positions was 'Blade up'. This was also the most common tongue shape, with eight of the nine subjects employing it in at least one vowel context. Six speakers had the 'tip up' variant in the context of /a/ and three of those are either more likely to, or only display this in Initial position. A cursory examination of the data from the other vowel contexts indicated that there is likely a similar tongue shape and similar variation for lot, and possibly occasionally l\xl as well. Only two speakers showed a 'tip down' tongue shape and this appeared to be both higher and rounder than in the idealized tongue shape for 'tip down' shown in figure 2.15 (modified from Hagiwara, 1995) in section 2.5.3. Table 3.5 Individual production variation: tongue shape and Lip gesture SUBJECT V INITIAL RESYL FINAL AGL  Id Id  ASN  Id Id Id Id Id Id Id Id Id Id  BWG CPT DCM DSL LIN MIY PTM  Id Id Id Id Id Id  Tip Up Blade Up Tip Up Blade Up Tip Up Blade Up Blade Up Tip Down Blade Up Blade Up Tip Up Blade Up Blade Up Blade Up Tip Up Tip Down Tip Up/Blade Up Blade Up  Tip Up/Blade Up Blade Up Tip Up Blade Up Tip Up Blade Up Blade Up Tip Down Blade Up Blade Up Tip Up Blade Up Blade Up Blade Up Tip Down Tip Down Blade Up Blade Up  Blade Up Blade Up Tip Up Blade Up Tip Up Blade Up Blade Up Tip Down Blade Up Blade Up Tip Up Blade Up Blade Up Blade Up Tip Down Tip Down Blade Up Blade Up  LIP POST-V Yes Yes Yes Yes No Yes Yes Yes Some Yes No Yes Yes Yes Yes Yes Some Yes  The loss of the labial gesture in post-vocalic allophones in Canadian English has been noted before in similar studies (Gick and Campbell, 2003) but in this case it was found to affect Ixl in this way only in the context of /a/, and only for four of the nine speakers, (two of whom lost the gesture only sporadically).  40  3.3.2 Results for glottal insertion in the Resyllabifiable context Table 3.6 summarizes the percentage of tokens from the Resyllabifiable (e.g. har awe) context that showed evidence of a glottal stop being inserted between the (lexically) word-final Ixl and the following vowel. All of the subjects included in the study showed some variation in their productions. The insertion of a full glottal stop happened more than 50% of the time for 7 of the nine subjects, and four of those subjects always produced some kind of glottal interruption in the signal. All of the subjects produced at least one token which had glottalization through the transition from the Ixl to the following vowel. Four subjects produced at least some tokens in which there was no glottal stop or glottalization, however only one subject never inserted a full glottal stop.  'able 3.6 Percentage of glottals inserted in Resyl abifiable position  AGL ASN  % with Glottal Stops 0 74  % with Glottalization 25 26  % with No Glottal 75 0  BWG  50  33  16  CPT  50  6  44  DCM  70  30  0  DSL  54  46  0  LIN  70  30  0  MIY  90  10  0  PTM  8  42  50  Average%  51.8  27.6  20.6  Subject  41  4.0 Discussion In this section the results presented in section 3.0 are further examined and the predictions of each of the proposals in section 1.3 is reviewed in light of what was found for Canadian English Ixl. A possible analysis of the data is presented and following this the qualitative tongue shape results are briefly discussed, as are directions for further study and possible problems with this experiment.  4.1 Summary of Overall Quantitative Results Since results for the Resyllabifiable condition were neither clearly distinguishable from results for the Final condition, nor clearly the same, as discussed below in section 4.6, this discussion focuses on the results for Initial (e.g. a#ra) and Final (e.g. ar#ha) positions. The position-based differences observed in the overall results were: 1. In Initial position: a. the timing was strictly front-to-back (Lip > T B > TR). b. the T R gesture showed reduction in magnitude. 2.  In Final position: a. the TR and Lip gestures preceded the T B gesture (TR/Lip > TB). b. the T B and Lip gestures showed a reduction in magnitude.  4.2 Comparison of results with predicted patterns In section 1.3 several proposals that make specific predictions about the timing and magnitude of the gestures of Ixl were discussed. Table 4.1 repeats the summarized information on these predictions given in table 1.1, with the addition of the results of this study. In this section those predictions will be evaluated based on the cross-subject results of this study. A general summary is followed by a more detailed examination of each of the theories.  42  Table 4.1 (modified from table 1.1) Summary of predicted categorization of gestures and predictions of relative timing by position (with results).  Sproat & Fujimura (1993) Browman & Goldstein (1995) Gick (2003) Carter (2002) Gick et al (2004b) Results:  LIP [vocalic]  TB [vocalic]  TR [vocalic]  narrower constriction (than TR) C-gesture  narrower constriction (than TR) C-gesture  wider constriction  ?  anterior  seemingly consonantal less anterior  ?  C-like  seemingly vocalic least anterior V-like  V-gesture  INITIAL All three simultaneous, All three reduced All three simultaneous. Lip / TB > TR, TR reduced Any order: dialect dependent All three simultaneous Lip > TB > TR  FINAL All three simultaneous, No reduction TR > Lip TR>TB Lip & TB reduced TR > Lip / TB Lip & TB reduced (Lip)/TR > TB/(Lip) TR > TB > Lip TR / Lip > TB  The observed front-to-back timing was not predicted by any of the proposals for Initial position, although Carter (2002) doesn't rule it out. Gick (2003) does predict a timing offset and that this will involve the T B preceding the TR, however, based on this proposal, a three-way distinction is not expected. Both studies that make explicit predictions regarding reduction in Initial position, Sproat and Fujimura (1993) and Gick (2003), predict that the TR will be reduced in this position, however, Sproat and Fujimura (1993) also predict that the Lip and T B will be reduced (compared to final position), which is not the case. In Final position all of the theories except for Sproat and Fujimura (1993) predict the attested timing relationship between the two lingual gestures (TR > TB), although only Carter (2002), and possibly Browman and Goldstein (1995), allow for the possibility that the Lip will pattern with the TR rather than the T B . Both Gick (2003) and Browman and Goldstein (1995) accurately predicted the reduction of the Lip and TB in Final position. Sproat and Fujimura's (1993) definition of [consonantal] and [vocalic] gestures was not an effective predictor of the behaviour of gestures with regards to timing and magnitude differences across syllable position for Canadian English Ixl. As observed in Gick (2003) and here, in a segment composed of multiple relatively wide constrictions one or more of those gestures will behave more like a [consonantal] gesture. Thus it does not appear to be the case that a phonological distinction is drawn, in terms of gestures, between 'extreme constrictions' and wider ones.  43  The similar proposal of Browman and Goldstein (1995), this time based on relative constriction width, fares slightly better. The reductions in magnitude seen in the TB and Lip gestures in Final position are predicted by this theory since these form the narrower constrictions, as is the phasing of the TR before the T B . It was predicted that the TB would precede the Lip in Final position, rather than these gestures occurring simultaneously, however given the significant reduction in magnitude observed in the Lip gesture (the constriction at the Lip is on average more than 5mm wider in Final position than in Initial position) and the relatively larger magnitude of the TR gesture in this position (it is on average 2mm narrower than in Initial position) it is not clear exactly what the relative widths of constriction are between these two gestures. While the simultaneous realization of all gestures Browman and Goldstein (1995) predict for Initial position is absent, the prediction was based on observations of the American English pattern and it is possible that Canadian English differs consistently in this respect from American English (recall that a timing offset in Initial position was also observed for III [Gick et al, 2004b]). It is therefore possible that constriction width could play a role in explaining the observed temporal organization, perhaps in both syllable positions, although stable differences between dialects must still be accounted for. Notably, unlike the methods used to distinguish types of gestures in other proposals, the width of a constriction relative to another can change across syllable positions via changes in gestural magnitude. The phonologically assigned categories of Gick (2003) do appear to be consistent with the results for magnitude, assuming the categories assigned in Table 1.1 and repeated in table 4.1. This theory accurately predicts the timing relationships between the two lingual gestures in both positions, however it cannot account for the three-way timing effect seen in Initial position or the Lip occurring with the TR in final position. The lack of a mechanism for a three-way timing distinction in syllable-initial position is a serious problem for this theory in explaining the patterns observed for Ixl, as is the method of categorizing gestures and the relative inflexibility of the categories. Within this theory, the category of a gesture is determined by both timing and magnitude patterns, however in this case the Lip gesture is not easily classified as either consonant-like or vowel-like. This is illustrated in figure 4.1 where the results are presented again, this time separated into timing results and magnitude results with each gesture assigned a category, based roughly on the definitions of these categories given in Gick (2003). Notice that in Final position the Lip, which is otherwise assigned C-gesture status, displays a V-gesture-like timing pattern. Designating a single category is therefore not possible in the case of the Lip gesture for Ixl because its relative timing in Final position suggests that it is a Vgesture, but it also undergoes final reduction like a C-gesture. Figure 4.1 Illustration of predicted categories of gestures, based on the results Timing: Initial  Final  Magnitude: C C V Lip > TB > TR  C C V no reduction of Lip or TB; reduction of TR  V V C TR / Lip > TB  C C V reduction of Lip and TB; no reduction of TR 44  The theory of Carter (2002) doesn't rule out the timing differences found, so long as the Lip can be considered a more vowel-like gesture, since vowel-like gestures must precede consonant-like gestures in final position and Lip precedes T B . Unfortunately his theory is not well developed with regards to Initial position. In particular, it is not clear how a stable timing pattern (of two or more distinct phases) would be realized in syllable-initial position within a given dialect. He does recognize the possibility of phasing seemingly vocalic elements before seemingly consonantal elements in syllableinitial position, so it is not impossible to consider the Lip vowel-like. The result of this, according to Carter, would be a 'dark' variant/allophone of the segment in syllableinitial position, which he predicts will be the case for Canadian English Ixl (based on the dark syllable-final lateral liquid in this dialect). This does seem to be at odds with the later timing of the TR in this position, and with the results for magnitude, however no testable predictions were made regarding the relative magnitude of the gestures across positions in this study. The predictions of Carter (2002) were the only ones that are consistent with the combined overall results, although this is likely due to the fact that most were relatively vague predictions. In addition, the one clear prediction, that the timing offset should be categorical in Final position with more consonant-like gestures following more vowel-like gesture, while borne out by the overall results, are not so clear when individual results are examined. There was a reasonable amount of variability in final position across subjects, as seen in table 3.4 where 3 of 7 subjects have a three-way distinction in gestural timing for Final position. This amount of variability is not expected if the pattern is based on a categorical (structural) phonological effect. The combined perceptual and biomechanical explanation offered by Gick et al (2004b) also fails to account for the observed pattern in Ixl. In fact, neither of the predictions for increasing perceptual recoverability through gestural timing is consistent with the data. In syllable-initial position (where this was expected to play a bigger role) none of the gestures display the simultaneous timing expected, and in syllable-final position instead of a 'back-to-front' ordering we see the most anterior gesture and the least anterior gesture occurring first. That said, it is possible that, since this prediction is based on closures and since the Lip gesture is not a closure, a different outcome may still be consistent with an increase in perceptual recoverability. The phasing of the Lip with the TR may be an example of this as it has been observed that rounding/protrusion of the lips can enhance frequency lowering, an acoustic cue of Ixl (Docherty and Foulkes, 2001), or cooperate with a dorsal gesture to enhance the perceptibility of the vowel lul (Perkell, Matthies, Svirsky, & Jordan, 1993). The theory that the jaw cycle affects intergestural timing is consistent with the 'frontto-back' results for syllable-initial position, assuming that anteriority straightforwardly correlates to dependence on/coupling with the jaw. It cannot account for the timing of the gestures of Ixl in Final position, which it predicts will be back-to-front, unless the lips and TR are similarly independent of jaw closure. Thus it is not possible that coupling with the jaw is the determining factor for intergestural timing in both positions as the Lip could not logically appear both closest to the TR (as in syllable-final position) as well as farthest from it (as in syllable-initial position). Recall also that Gick 45  et al (2004b) predict that the jaw-cycle dependent pattern is more likely to occur syllable-finally. The fact that there is a three-way distinction in the timing pattern for Initial position does lend support to the notion that the organization is based on phonetic factors (rather than a binary phonological categorization). However if some combination of perceptual recoverability and the jaw cycle, as formulated in Gick et al (2004b), are to account for the results of this study, the opposite pattern to that of Gick et al (2004b) would have to be invoked, namely that perceptual recoverability plays a larger role in determining timing in syllable-final position while the jaw cycle controls timing in syllable-initial position. It emerges from this discussion that there are three aspects of the results which seem to pose interesting problems for previous analyses: (i) the 3-way front-to-back distinction in timing in Initial position, (ii) the apparent V-like timing of the Lip in Final position, and (iii) the observation that different dialects may show different organizational patterns. These issues are further explored in the following section, which presents a proposal to account for the results of this experiment.  4.3 Proposal Several points which have emerged as a result of this study must be considered when constructing an analysis of the data. As mentioned above, the behaviour of the Lip gesture requires that magnitude and timing be considered related but separate events, a binary phonological explanation of timing patterns is unlikely in light of the three-way timing results for Initial position, and the system must be flexible enough to account for both dialect differences in organization and the apparent category-shifting of the Lip gesture. It is proposed here that extending the concept of timing based on relative constriction width, from Browman and Goldstein (1995), such that it applies to all positions, may meet these requirements. Since the width of two or more gestures, relative to one another, can change based on syllable position (due to positiondependent reductions in magnitude) the timing of the gestures relative to one another can also change. I therefore propose that the relative timing of gestures is dependent on the magnitude of gestures. The limitations of previous proposals with regards to the present study were explored in section 4.2 and a major issue for most of them is the pattern seen in Final position. It is not possible for an articulator to vary its 'degree of anteriority' relative to the other articulators across syllable positions, and it would not be expected that a gesture could arbitrarily change its phonological category, and so these generalizations simply cannot capture the variable behaviour of the Lip gesture. As noted above, the results of this and other studies regarding positional reduction indicate that it is possible for a single gesture to have different constriction widths in different positions. It is therefore possible that since the width of constriction can be syllable-position dependent, the timing of gestures could consistently be dependent on the relative widths while maintaining the possibility of the relationship between the gestures changing. 46  Browman and Goldstein (1995) only apply the generalization about constriction width to syllable-final position, claiming that in this position wider constrictions always precede narrower constrictions. It is proposed here that this should be extended such that, where a timing offset is found, the wider a constriction is, the closer (temporally) it will be realized to the vocalic peak of the syllable, with respect to other gestures present. This idea, to a certain extent mirrors phonological generalizations about segments and the sonority hierarchy. Thus for Ixl if the Lip, as it is realized in syllableinitial position, is the narrowest of the three gestures and there is a timing offset in initial position in the dialect, it would be expected to precede the other two gestures. The T B would follow, as the next narrowest constriction, and the TR would be realized closest to the vocalic element in the syllable, exactly the pattern observed in this study. It might be expected that in Final position the reverse order should be seen, however recall that while no data on lingual constriction width, per se, is available in this study, the constriction at the Lip was, on average, over 5mm wider in Final position. This, combined with the extra 2mm of backing of the TR in Final position means that the difference in the widths between the Lip and TR is likely at least 7mm less in Final position. Given that all the gestures are relatively wide to begin with, this difference could mean that the constrictions at the Lip and TR in Final position are of approximately equal width, and would therefore be expected to have a similar timing. The T B gesture was an average of 2mm smaller in Final position than it was in Initial position, a small enough difference that the reduction of the Lip gesture could reverse these gestures in terms of relative width. This mechanism also allows for a certain amount of variation in timing patterns across dialects and individuals, as is attested, while still allowing that each individual and dialect can be following the same principles of temporal organization. In almost all of the studies comparing the timing of two gestures which are mentioned in section 1.1.2, one of the gestures is a closure or near-closure (tongue tip for III, oral closure for nasals) while the other is not (tongue dorsum for III, velum opening for nasals). This being the case, it would not be expected that a change in the magnitude of the gestures would be large enough to change their width relative to one another and so the results are generally consistent with this proposal. Recall that the current proposal deals only with cases where there is a timing offset, so that the results of Browman and Goldstein (1995) where the gestures were found to be simultaneous in syllable-initial position is not accounted for. No explanation is readily available for this, although according to Carter (2002) there are a wide variety of attested timing patterns for initial position across dialects, so it should not be unexpected. The other study which is of interest in this respect is Gick (2003), in which a greater timing offset was found in Initial position for /w/, a result which was considered anomalous when compared to patterns for nasals or American /I/. The observation that the Lip gesture in /w/ was reduced in syllable-final position and patterned more closely with the tongue dorsum gesture (although the Lip gesture did follow the T D gesture) is consistent with the results for Ixl in the present study, and would be predicted by this proposal, assuming that the reduction of the Lip gesture in final position brings it closer to the width of the tongue dorsum gesture. 47  A possible reason for narrower constrictions to be found at syllable edges in syllablefinal position could be perceptual recoverability factors, as discussed above. It could be the case that narrowed constrictions obscure the acoustic results of wider constrictions, in much the same way more anterior constrictions are proposed to obscure less anterior ones when produced simultaneously (Kochetov, in press). It is not clear whether the greater offset in timing in initial position observed for Ixl is a characteristic of segments with multiple 'wider' gestures, which may enhance the perceptibility of segments like Ixl and Iwl (Gick, 2003), or whether it is a feature of Canadian English, since both Ixl and III (Gick et al, 2004b) show this pattern for speakers of this dialect, or both. As for what might drive positional reduction, it should first be noted that this is an effect not unique to liquids in English - It/, Inl, Ikl and /p/ have also been observed to undergo reduction in syllable-final position (Browman and Goldstein, 1995). Browman and Goldstein (1995) suggest that the reduction of gestures syllable-finally is due to the relatively 'weaker' syllable position and they pose the question of whether multiple articulators are responsible for the apparent reduction or if it could be the effect of just one. They use the jaw as an example of a single articulator which could affect the magnitude of the lingual and labial constrictions. The gestures which undergo final reduction in this study, the Lip and the T B , are the gestures which were predicted to be affected more by the movement of the jaw, so it is feasible that it is these two gestures which reduce because of a lower jaw position in Final position. Preliminary analysis of jaw height data which was collected in this study suggests that it is in fact the case that the jaw is higher in syllable-initial position. Whether this effect is sufficient to account for the difference observed in Lip magnitude, however, will require future investigation One shortfall of a strictly jaw height-based explanation for positional variation in gestural magnitude is that it is not able to account for why there is reduction of the TR gesture in syllable-initial position. A possible suggestion is that this can be seen as augmentation (Fougeron and Keating, 1995) in syllable-final position, perhaps as a compensatory strategy. Additional data on the jaw will help to provide a more adequate theory in the future.  4.4. Discussion of qualitative results for tongue shape Potential dialect differences between Canadian and American English were noted in terms of the relative timing of gestures in section 4.2 above, and it appears that there may also be dialect differences between American and Canadian English in terms of articulatory configuration during Ixl. For speakers of Canadian English the most common tongue shape observed was the 'blade up' configuration (see figure 2.15), while Hagiwara (1995) found that 'tip up' was the most common for southern Californian speakers. It should however be kept in mind that that the methods used in his study may have influenced the results: sustained productions of Ixl were used and subjects were aware of the nature of the study so they may have produced a 'correct' rather than a natural Ixl.  48  Hagiwara (1995) raises the question of whether 'blade up' can be viewed as a variation of the 'tip up' configuration, and comes to the conclusion that it is more closely related to the 'tip down' class of tongue shapes. In his study subjects who used the 'tip up' variant did so consistently across syllable positions, while others tended to alternate and produce 'tip down' in syllable-final position and 'blade-up' in syllable-initial position. This is quite different from the findings of the present study, in which two subjects employed only 'blade up', and the others used some combination of two, most commonly 'tip up' with the vowel /a/ and 'blade up' with the vowel Id. None of the subjects used all three tongue shape classes, and only one speaker had 'blade up' alternating with 'tip down'. Thus the results of this study indicate that blade up is more likely to alternate with 'tip up' and therefore if it is to be considered a variant of one of the other two classes, it is more likely to be 'tip up'. It should be noted that even within a given context for a single subject there can be some variation present in the production of Ixl, so there appears to be a certain amount of 'free variation' present for this segment. Within the Canadian subjects, dialect (or region) did not correspond to any particular characteristics. As noted in section 3.3.1, vowel context and syllable position are the two factors which seem to affect the tongue shape used to produce Ixl for a given subject. It is interesting to note that in both Hagiwara (1995) and this study, of the two tongue shapes used by any one speaker the variant with the anterior part of the tongue raised higher is generally used in Initial position. This is potentially related to Initial being the 'stronger' syllable position and the reduction of the T B gesture in Final position, although Final reduction is seen in subjects who consistently use the same tongue shape across positions (e.g. DCM).  4.5. Implications for Articulatory Phonology One of the key benefits of Articulatory Phonology, according to Browman and Goldstein (1992) is its ability to consider a variety of phonological phenomena in terms of a more universal explanation. Several generalizations have come out of this study which fit easily within this framework, however at least one may also pose a challenge for the theory. With regards to the domain in which timing is organized, it is clear that a binary categorical (phonological) distinction between two types of gestures (consonantlike and vowel-like) is not appropriate, Articulatory Phonology is not incompatible with this finding, nor with the fact that a completely gradient (phonetic) notion of anteriority does not account for the results either. Articulatory phonology allows for differences in the organization of onsets and codas, as has been observed in this and other studies, and the possible role of the jaw in determining relative magnitude could also be easily incorporated into the theory. The one result of this study which does not appear to fit easily within an Articulatory Phonology framework is the dependence of timing on magnitude. If both of these effects depended (separately) on another factor, for example the jaw, this would not be a problem, however if this were the case no mechanism for the unexpected order of gestures in Final position is apparent. As it stands, some aspect of Articulatory Phonology will have to be modified if the current interpretation of these results is correct, since the gestures are not phased relative to one another first, but rather based on the effects of a process (reduction) which affects them.  49  4.6 Potential problems with the experiment There are several factors which must be considered when evaluating the reliability of the results of this study and how they should be interpreted. These include the potential for error introduced while calculating timing differences across signals, the lack of a reliable pattern in the Resyllabifiable context, and that due to the variability in /r/ productions there is the possibility that the measures from M-mode data do not always capture the relevant tongue movement or position. There is the potential for the experimental set-up to introduce a certain amount of 'error' in measurements of time and distance. Based on the precautions taken to avoid this, tests matching times and durations of observable events, and the results themselves, it is unlikely that any false effects were introduced. For timing measures, if the equipment or software were introducing a significant error, either a consistent effect on the data or a lot of 'noise' would be expected. In the case of a consistent effect, the Lip gesture would be offset in the same direction from both lingual gestures across all positions. Because we see variation in the timing of the Lip gesture relative to the lingual gestures this is unlikely to be the case. That said, because the Lip position measures are sampled once every 11.111ms and the use of audio to synchronize this with the Ultrasound data may introduce an error of up to one Optotrak sample (±11.11 lms). In addition, there is an unknown amount of potential error introduced during the digitization process, including the potential for 11.122ms drift between the audio and the ultrasound signals, as mentioned in section 2.5.1. However, this error should be distributed randomly throughout the dataset, and can therefore be interpreted as increased noise in the data. For distance measures, head movement and tissue compression at the transducer are the biggest concerns. As noted in chapter 2, a number of methods were used to minimize head movement (appropriate presentation of the data, head stabilizing equipment, and a consistent intonation/stress pattern throughout the randomized stimuli) but this movement was not actually measured. It is therefore possible, however unlikely (see Gick et al 2004a), that subjects could have systematically lifted or lowered the head during tokens of one context and not the other. It should also be noted that claims based on the Resyllabifiable context are limited, as resyllabification was inconsistent. As noted in table 3.6, many subjects inserted glottal stops between the final Ixl and the following vowel in Resyllabifiable context, and some glottalized the transition between the two. This does explain the tendency of the Ixl to pattern as Final, however the two conditions could not be combined because in some cases there were differences from the Final condition. The major difference of interest in this study would have been the apparent difference in overall timing, where there is a three-way distinction for Resyllabifiable and a two-way distinction for Final. This difference is the result of the TR being significantly different from Lip in the Resyllabifiable condition but not in the Final condition. It should be noted however that the difference between the TR across the two positions is not significant, and in any case the amount of potential error in timing calculations between labial and lingual 50  gestures exceeds the timing difference observed between the TR and the Lip in the Resyllabifiable context, so it is not clear whether the TR (in either or both of these positions) occurs simultaneously with the Lip or before it. A final point to consider is that the variability in tongue shape likely influenced, in certain cases, the measures of the T B gesture. For example, the fact that MIY tended to have a 'tip up' configuration in Initial position and 'tip down' in Final position, given the method of measuring tongue movement along a set trajectory, may explain why MIY appears to have a smaller T B gesture in initial position. This illustrates a benefit of combined B/M mode ultrasound and a disadvantage of using only M-mode data for measures of variable segments. In figure 3.10 it is clear that it is not the tongue tip that is being measured but rather the tongue blade, which is higher in the 'tip down' configuration for this speaker. B-mode data allows the researcher to go back and reconstruct the events around the time of the measurement for the whole tongue so that this measure can be explained. M-mode data is limited in that the line(s) must be set at the time of data collection and cannot be moved when an event of interest occurs elsewhere on the screen, and therefore in this case they were not able to capture the variation in a segment like Ixl for all speakers. M-mode ultrasound, however, remains the only method currently available which can image the tongue root at a sufficiently high temporal resolution to allow these timing comparisons to be made.  4.7 Suggestions for future work While the proposal put forth in this thesis appears to account for the data and have independent motivation, it is not conclusive, since this study provides no data on the actual size of constrictions. Future work on Ixl and other composite segments should take this into consideration, although it may not currently be possible to gather such data on all of the constrictions for English Ixl in different syllable positions. Advancements in MRI technology may soon make this possible. A similar study of American English Ixl to that conducted on Canadian English Ixl could further our understanding of dialect variation, in addition to representing another step toward understanding the principles behind the articulatory organization of syllables cross-linguistically. Such a study would perhaps benefit by including syllabic Ixl and intervocalic Ixl. The possibility of a dependence relationship between positional reduction (magnitude of gestures) and timing is a theoretical question for Articulatory Phonology which may prove a productive area of exploration. Lastly, while discussion of this matter is beyond the scope of this thesis, future work will want to consider where the line is to be drawn between phonetics and phonology, particularly with regards to gestural descriptions and position-based effects.  51  5.0 Conclusion The goal of the research presented in this thesis was to answer the following questions: what are the differences in timing and magnitude for each of the three gestures of kl across syllable positions? is this predicted by previous explanations of syllable-based allophonic variation? and are timing offsets and magnitude reduction linked or independent phenomena? This study has examined the relative timing and magnitude of English kl across syllable positions. Differences in intergestural timing and magnitude were observed for all subjects and across subjects. Overall, the timing was observed to proceed from front-toback in syllable-initial position and in syllable-final position the TR and Lip preceded the TB. There was syllable-final reduction of the two more anterior gestures (TB and Lip) and syllable-initial reduction of the least anterior gesture (TR). These findings are not entirely consistent with any of the theories examined which attempted to explain syllable-based allophonic variation based on observation of two gestures. It is possible however that the generalization for syllable-final position put forward by Browman and Goldstein (1995) that constriction width can predict gestural timing patterns may play a role in both syllable positions. It is proposed that, perhaps due to a reduction in jaw height, the Lip and TB gestures are reduced in syllable-initial position, and that the significant reduction of the Lip gesture results in it patterning with the wider TR constriction. The results of this study would then indicate that the relative timing of gestures is dependent on the magnitude. A second conclusion of this study is that the three-way distinction in timing between the Lip, T B , and TR gestures in Initial position means that a straightforward binary phonological categorization of gestures is not sufficient to explain the attested pattern. It also appears that both the tongue shape for kl and the gestural organization of kl gestures across syllable positions can be quite variable. This holds both within and across speakers and dialects (for tongue shape). This being the case, the principles governing this organization must be sufficiently flexible to allow for a certain amount of variation, although whether that is to be laid out in the phonology or left to the phonetics must be set aside as a topic for future study.  52  References  Adobe online technical support: Premiere. (2004). Using locked and unlocked DV audio in premiere. Retrieved August, 2004, from http://www.adobe.com/support/techdocs/322499.html Alwan, A., Narayanan, S., & Haker, K. (1997). Toward articulatory-acoustic models for liquid approximants based on MRI and EPG data. Part ii. The rhotics. Journal of the Acoustical Society of America, 101(2), 1078-1089. Boersma, P., & Weenink, D. (1992-2004). Praat (Version 4.2.12). Browman, C. P., & Goldstein, L. (1992). Articulatory phonology: An overview. Phonetica, 49, 155-180. Browman, C. P., & Goldstein, L. (1995). Gestural syllable position effects in American English. In F. R. Bell-Berti, Lawrence J. (Ed.), Producing speech: Contemporary issues. For Katherine Safford Harris. Woodbury, NY: American Institute of Physics. Campbell, F., Gick, B., & Namdaran, N. (2004). The gestural organization of English Ixl: A study of timing and magnitude. Poster presented at From Sound to Sense: Fifty + Years of Discoveries in Speech Communication. MIT, Cambridge, M A . Carter, P. (1999). Ahstractness in phonology and extrinsic phonetic interpretation: The case of liquids in English. In Proceedings of the XlVth International Congress of Phonetic Sciences. San Francisco, C A . Carter, P. (2002). Structured variation in British English liquids: The role of resonance. Unpublished PhD Dissertation, University of York. Chitoran, I., Goldstein, L., & Byrd, D. (2002). Gestural overlap and recoverability: Articulatory evidence from Georgian. In C. Gussenhoven & N. Warner (Eds.), Papers in laboratory phonology V7/(pp. 419-447). Berlin, New York: Mouton de Gruyter. Delattre, P., & Freeman, D. (1968). A dialect study of American r's by x-ray motion picture. Linguistics, 44, 29-68. Docherty, G., & Foulkes, P. (2001). Variability in (r) production-instrumental perspectives. Etudes & Travaux, 4(Dec), 173-184. Espy-Wilson, C. (2004). Articulatory strategies, speech acoustics and variability. Proceedings of From Sound to Sense, 50+ years of discoveries in speech communication, Cambridge M A : MIT.  In  Fougeron, C , & Keating, P. A . (1995). Demarcating prosodic groups with articulation. Journal of the Acoustical Society of America, 97(5), 3384. 53  Gick, B. (1999). A gesture-based account of intrusive consonants in English. Phonology, 16, 29-54. Gick, B. (2002). The use of ultrasound for linguistic phonetic fieldwork. Journal of the International Phonetics Association, 32(2), 113-122. Gick, B. (2003). Articulatory correlates of ambisyllabicity in English glides and liquids. In J. Local, R. Ogden & R. Temple (Eds.), Labphon VI: Constraints on phonetic interpretation. Cambridge: Cambridge University Press. Gick, B., Bird, S., & Wilson, I. (2004a). Techniques for field application of lingual ultrasound imaging. Clinical Linguistics and Phonetics, Under Review. Gick, B., & Campbell, F. (2003). Intergestural timing in English Irl. In Proceedings of the XVth International Congress of Phonetic Sciences, Barcelona: Universitat Autonoma de Barcelona. Gick, B., Campbell, F., Oh, S., & Tamburri-Watt, L. (2004b). Toward universals in the gestural organization of syllables: A cross-linguistic study of liquids. Under review. Gick, B., & Goldstein, L. (2002). Relative timing of the three gestures of north American English Irl. Journal of the Acoustical Society of America, 111(5), Pt. 2,2481. Gick, B., Min Kang, A., & Whalen, D. H. (2002). MRI and x-ray evidence for commonality in the dorsal articulations of English vowels and liquids. Journal of Phonetics, 30, 357-371. Guenther, F., Espy-Wilson, C , Boyce, S., Matthies, M., Zandipour, M., & Perkell, J. (1999). Articulatory tradeoffs reduce acoustic variability during American English Irl production. Journal of the Acoustical Society of America, 105, 28542865. Hagiwara, R. (1995). Acoustic realizations of American Ixl as produced by women and men. UCLA Working Papers in Phonetics, 90, 1-187. Hashi, M., Honda, K., & Westbury, J. (2003). Time-varying acoustic and articulatory characteristics of American English [a]: A cross-speaker study. Journal of Phonetics, 31, 3-22. Keating, P. (1983). Comments on the jaw and syllable structure. Journal of 17,401-406.  Phonetics,  Kelso, J., Saltzman, E., & Tuller, B. (1986). The dynamic perspective on speech production: Data and theory. Journal of Phonetics, 14, 171-196.  54  Kochetov, A. (2002). Production, perception, and emergent phonotactic patterns: A case of contrastive palatalization. New York & London: Routledge. Kochetov, A. (in press). Syllable position effects and gestural organization: Articulatory evidence from Russian. In Papers in laboratory phonology viii: Varieties of phonological competence. Berlin, New York: Mouton de Gruyter. Krakow, R. A. (1989). The articulatory organization of syllables: A kinematic analysis of labial and velar gestures. PhD dissertation, Yale University. Krakow, R. A. (1999). Physiological organization of syllables: A review. Journal of Phonetics, 27, 23-54. Lindau, M . (1985). The story of Irl. In V. Fromkin, A. (Ed.), Phonetic linguistics: Essays in honor of Peter Ladefoged (pp. 157-168). London: Academic Press. Lindblom, B. (1983). Economy of speech gestures. In P. F. MacNeilage (Ed.), The production of speech (pp. 217-246). New York: Springer-Verlag. MacNeilage, P. F. (1998). The frame/content theory of evolution of speech production. Brain and Behavioral Science, 27(449-546). Mattingly, I. G. (1981). Phonetic representation and speech synthesis by rule. In J. L. T. Myers & J. Anderson (Eds.), The cognitive representation of speech (pp. 415420). Amsterdam: North Holland. Perkell, J. S., Matthies, M . L., Svirsky, M. A., & Jordan, M . I. (1993). Trading relations between tongue-body raising and lip rounding in production of the vowel lul: A pilot "motor equivalence" study. Journal of the Acoustical Society of America, 93(5), 2948-2961. Silverman, D. (1995). Phasing and recoverability. PhD dissertation, U C L A . Garland Publishing, 1997. Sproat, R., & Fujimura, O. (1993). Allophonic variation in English III and its implications for phonetic implementation. Journal of Phonetics, 21, 291-311. 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 . (2004). A fuzzy line: Understanding tongue features from ultrasound images. Journal of Clinical Linguistics and Phonetics, Under Review. Tiede, M . , Boyce, S., Holland, C. and Choe, A. (2004), A New Taxonomy of American th  English Irl using MRI and ultrasound, Poster presented at the 747 Meeting of the Acoustical Society of America, http://scitation.aip.Org/confst/ASA/data/l/5pSC37.pdf 55  Uldall, E. (1958) 'American "molar" R and "flapped" 17 Revista do Laboratorio de Fonetica Experimental, Universidad de Coimbra 4. 103-6. Westbury, J. R., Hashi, M., & Lindstrom, M. J. (1998). Differences among speakers in articulation of American English Ixl. Speech Communication, 26, 203-226. Wilson, I. (in preparation). Phonetic economy in the articulatory settings of bilingual speakers. PhD dissertation, UBC. Zawadzki, P. A., & Kuehn, D. P. (1980). A cineradiographic study of static and dynamic aspects of American English Ixl. Phonetica, 37, 253-266.  56  Appendix I: Individual subject information SUBJECT AGL ASN BWG CPT DCM DSL LIN MIY PTM *TBY  SEX F M F M F F F M M M  AGE 28 22 25 25 22 22 36 27 24 34  ORIGIN Winnipeg, M B Vancouver, B C B C Interior Vancouver, B C Vancouver, B C Vancouver, B C Southern Ontario Richmond, B C Sudbury, ON Vancouver, B C  OTHER L A N G U A G E S ? Some French, (Spanish?) Cantonese Some French Some French -  * This subject was not included in the analysis due to poor ultrasound image quality.  57  Appendix II: List of stimuli by vowel context Ixl  hee ree hear ee hear hee  Id  who roo who're oo who're. hoo  Id  hay ray hair A hair hay  lol  hoe roe hoar owe hoar hoe  Id  haw raw har awe har haw  58  Appendix III: Individual results for Timing AGL Illustration of Timing Differences for TB and TR gestures relative to Lip: Interaction Bar Plot for Some of 0/1 - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final * Some of TB/TR Error Bars: 9 5 % Confidence Interval  Resyl  TB Initial  TR  Final  30 - 2 0 - 1 0  l 0  r 10 20 30 4 0 50 60 70 Cell Mean  Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0)  Initial Initial Resyl Resyl Final Final  Lip/TB Lip/TR Lip/TB Lip/TR Lip/TB Lip/TR  Mean 12.103 14.649 17.774 -17.237 43.875 -8.418  p value .0015 <0001 .0309 .0012 .0005 .0421  Unpaired t-tests for differences between TB and T R by position (Significance level p < .05) TB Mean TR Mean p value Initial T B / T R 12.103 14.649 .5126 Resyl TB/TR: 17.774 -17.237 .0002 Final TB/TR 43.875 -8.418 <.0001 Initial: L i p > T B / TR Resyl: TR > Lip > T B Final: TR > Lip > T B  59  ASN Illustration of Timing Differences for TB and TR gestures relative to Lip: Interaction Bar Plot for Some of 0/T - U/S(ms) x -1 Effect: Initial/Resyl/Final * TR/TB Error Bars: 95% Confidence Interval  Resyl  TB  Initial  TR  Final I  40  1  1  1  r  -20  -i—|—,—i—i—i—r-  0  20 Cell Mean  40  60  Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0) (Significance level p < .05) Mean p value Initial Lip/TB 48.032 <.0001 31.994 Initial Lip/TR .0018 Resyl Lip/TB 24.084 .0035 Resyl L i p / T R .093 .9919 Final Lip/TB 33.549 .0007 Final Lip/TR -17.346 .0255  Unpaired t-tests for differences between T B and TR by position (Significance level p < .05) TB Mean TR Mean p value Initial T B / T R 48.032 31.994 .1399 Resyl TB/TR: 24.084 .093 .0393 Final TB/TR 33.549 ' -17.346 <0001 Initial: Lip > T R / T B Resyl: Lip / T R > T B Final: TR > Lip > T B  60  BWG Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0) (Significance level p < .05) Mean p value Initial Lip/TB 55.445 <0001 Initial Lip/TR 66.547 <0001 Resyl Lip/TB NA NA Resyl Lip/TR 57.882 <0001 Final Lip/TB NA NA Final Lip/TR 44.760 <.0001  Unpaired t-tests for differences between TB and T R by position (Significance level p < .05) TB Mean TR Mean p value Initial TB/TR 66.547 55.445 .2108 Resyl TB/TR: NA NA NA Final TB/TR NA NA NA Initial: Lip > TB/TR Resyl: Lip > T R Final: Lip > T R  61  CPT lustration of Timing Differences for TB and TR gestures relative to Lip: Interaction Bar Plot for Some of 0/1 - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final * Some of TB/TR Error Bars: 95% Confidence Interval  Resyl  TB Initial 1  TR  Final i  -40  -20  0  20 40 Cell Mean  1  60  r 80  100  Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0)  Initial Initial Resyl Resyl Final  Lip/TB Lip/TR Lip/TB Lip/TR Lip/TB  Final L i p / T R  Mean 31.133 69.154 29.732 -15.963 16.957  p value <.0001 <0001 .0120 .0173 .0174  -9.667  .1870  Unpaired t-tests for differences between T B and TR by position (Significance level p < .05) TB Mean TR Mean p value Initial TB/TR 31.133 69.154 <0001 Resyl TB/TR: 29.732 -15.963 .0003 Final TB/TR 16.957 -9.667 .0080  Initial: Lip > TB > T R Resyl: TR > Lip > T B Final: TR / Lip > T B 62  DCM Illustration of Timing Differences for TB and TR gestures relative to Lip: Interaction Bar Plot for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final * Some of TB/TR Error Bars: 95% Confidence Interval  Resyl  TB Initial  TR  Final 1 i— —i— —i— —i—i1  •80  -60 -40  1  •20  1  0  20  40  60  80  Cell Mean  Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0)  Initial Lip/TB Initial Lip/TR Resyl Lip/TB Resyl Lip/TR Final Lip/TB Final Lip/TR  Mean 17.111 44.787 43.333 -47.631 48.433 -29.174  p value .0019 <0001 .0075 <0001 .0087 .0534  Unpaired t-tests for differences between T B and TR by position (Significance level p < .05) T B Mean TR Mean p value Initial TB/TR 17.111 44.787 .0003 Resyl TB/TR: 43.333 -47.631 <.0001 Final TB/TR -29.174 48.433 .0051  Initial: Lip > TB > TR Resyl: TR > Lip > T B Final: TR / Lip > T B  63  DSL Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0) (Significance level p < .05) Mean p value Initial Lip/TB 23.948 <.0001 Initial Lip/TR 43.563 <0001 Resyl Lip/TB NA NA Resyl L i p / T R 1.286 .7601 Final Lip/TB NA NA .0094 Final Lip/TR 11.932  Unpaired t-tests for differences between T B and TR by position (Significance level p < .05) T B Mean TR Mean p value <.0001 Initial TB/TR 23.948 43.563 NA NA NA Resyl TB/TR: NA NA Final TB/TR NA Initial: Lip > T B > T R Resyl: Lip/TR Final: Lip > TR  64  LIN Illustration of Timing Differences for T B and TR gestures relative to Lip: Interaction Bar Plot for Some of 0 / T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final * Some of TB/TR Error Bars: 9 5 % Confidence Interval  Resyl "  Initial 1  Final  -60  -40  -20 0 Cell Mean  20  40  Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0) (Significance level p < .05) Mean p value  Initial Lip/TB  -16.980  .0619  Initial Lip/TR  25.983  .0016  Resyl Lip/TB  -3.225  .7803  Resyl Lip/TR  -37.786  .0002  Final Lip/TB  14.155  .0955  Final Lip/TR  -29.582  .0003  Unpaired t-tests for differences between T B and TR by position (Significance level p < .05) TB Mean TR Mean p value Initial TB/TR -16.980 25.983 .0004 Resyl TB/TR: -3.225 -37.786 .0116 Final TB/TR 14.155 -29.582 .0001  Initial: T B / Lip > TR Resyl: TR > T B / Lip Final: T R > L i p / T B  65  MIY Illustration of Timing Differences for TB and T R gestures relative to Lip: Interaction Bar Plot for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final * Some of TB/TR Error Bars: 95% Confidence Interval  Resyl  TB  Initial 1  TR  Final "i  •100  1  1  -60  1  1  1  1  -20 20 Cell Mean  r  60  100  Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0)  Initial Lip/TB  Mean -55.848  p value .0080  Initial L i p / T R  -.169  .9925  Resyl Lip/TB Resyl Lip/TR  41.281 -41.259  .0499 .0053  Final L i p / T B  9.091  .6323  Final Lip/TR  -52.469  .0004  Unpaired t-tests for differences between T B and T R by position (Significance level p < .05) p value TR Mean TB Mean .0421 -.169 -55.848 Initial TB/TR -41.259 .0015 Resyl TB/TR: 41.281 .0058 -52.469 9.091 Final TB/TR  Initial: T B > Lip / T R Resyl: TR > Lip > T B Final: TR > L i p / T B 66  PTM Illustration of Timing Differences for TB and TR gestures relative to Lip: Interaction Bar Plot for Some of 0 / T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final * Some of TB/TR Error Bars: 9 5 % Confidence Interval  Resyl  TB Initial  TR  Final  20  40  60 80 Cell Mean  100  120 140  Results for Timing Differences: One group t-tests for differences between lingual gestures and Lip (Lip = Hypothesized Mean = 0)  Initial Lip/TB Initial Lip/TR  Mean 48.689 43.591  p value <0001 <.0001  Resyl Lip/TB Resyl Lip/TR  55.977 11.471  .0655 .1808  Final Lip/TB Final Lip/TR  67.875 31.450  .0450 .0009  Unpaired t-tests for differences between TB and TR by position (Significance level p < .05) TB Mean TR Mean p value  Initial TB/TR  48.689  43.591  .3160  Resyl TB/TR: Final TB/TR  55.977 67.875  11.471 31.450  .0245 .0390  Initial: Lip > T R / T B Resyl: Lip / T R / T B ; but TR > T B (?) Final: L i p > T R > T B  67  Appendix IV: Individual A N O V A s testing for ^syllabification effects AGLTB ANOVA Table for Some of 0 / T - U/S(ms) x -1 Some of Initial/Resyl/Final Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  5893.716  2946.858  7.352  .0027  14.705  .923  28  11222.441  400.801  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  Final  10  43.875  25.849  8.174  Initial  11  12.103  9.267  2.794  Resyl  10  17.774  21.985  6.952  S t d . Err.  Fisher's PLSD for Some of 0 / T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value  Final, Initial  31.772  17.918  .0011  Final, Resyl  26.101  18.340  .0069  Initial, Resyl  -5.671  17.918  .5221  68  A G L TR ANOVA Table for Some of 0/T - U/S(ms) x -1 Some of  Initial/Resyl/Final  Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  8053.145  4026.572  23.196  <.0001  46.393  1.000  41  7117.059  173.587  Means Table for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  15  -8.418  14.577  3.764  Initial  16  14.649  10.116  2.529  Resyl  13  -17.237  14.739  4.088  Interaction Bar Plot for Some of O/l - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval Resyl  Initial  Final  -30  -20  -10  0  10  20  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value <.0001  Final, Initial  -23.067  9.563  Final, Resyl  8.819  10.083  .0848  31.886  9.935  <.0001  Initial, Resyl  69  ASN T B - NOT SIGNIFICANT ANOVA Table for Some of 0/T - U/S(ms) x -1 DF Some of  Initial/Resyl/Final  Residual  Sum of Squares 2954.261  1477.131  30  15054.405  501.814  Means Table for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  12  33.549  25.014  7.221  Initial  9  48.032  17.891  5.964  Resyl  12  24.084  22.585  6.520  Final  Fisher's PLSD for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Diff.  P-Value  Initial  Mean Diff. -14.483  20.174  .1530  Final, Resyl  9.465  18.677  .3090  23.948  20.174  .0216  Final,  Initial, Resyl  Mean Square  2  Crit.  70  F-Value  P-Value  Lambda  Power  2.944  .0680  5.887  .522  A S N TR ANOVA Table for Some of 0/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  15001.168  7500.584  10.558  .0003  21.116  .988  32  22733.092  710.409  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  12  -17.346  23.275  6.719  Initial  12  31.994  27.019  7.800  Resyl  11  .093  29.569  8.916  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval  - 4 0 - 3 0 - 2 0 -1 0  0  10  20  30  40  50  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value  Final, Initial  -49.340  22.164  <.0001  Final, Resyl  -17.439  22.663  .1268  31.901  22.663  .0073  Initial, R e s y l  S  S  71  60  BWGTR ANOVA Table for Some of 0/T - U/S(ms) x -1 Some of  Initial/Resyl/Final  Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  2412.456  1206.228  4.215  .0265  8.430  .685  25  7154.201  286.168  Means Table for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  11  44.760  19.575  5.902  Initial  9  66.547  12.578  4.193  Resyl  8  57.882  17.141  6.060  Final  S t d . Err.  Fisher's PLSD for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Diff.  P-Value  Final, Initial  Mean Diff. -21.787  15.660  .0083  Final, Resyl  -13.122  16.189  .1075  8.664  16.929  .3019  Initial  Resyl  Crit.  72  CPT T B - NOT SIGNIFICANT  ANOVA Table for Some of O/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  1189.667  594.834  1.878  .1739  3.755  .342  25  7919.569  316.783  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  Final  10  16.957  18.446  5.833  Initial  10  31.133  7.326  2.317  Resyl  8  29.732  24.998  8.838  S t d . Err.  Interaction Bar Plot for Some of 0/1 - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval  0  10  20  30  40  50  60  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff. Final,  Diff.  P-Value  16.393  .0871  -12.775  17.388  .1428  1.401  17.388  .8695  Initial  -14.177  Final, Resyl Initial, Resyl  Crit.  73  CPTTR ANOVA Table for Some of 0/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  44991.789  22495.894  74.234  <.0001  148.468  1.000  27  8182.081  303.040  Means Table for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  10  -9.667  21.405  6.769  Initial  10  69.154  12.245  3.872  Resyl  10  -15.963  17.349  5.486  Interaction Bar Plot for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval  -40  -20  0  20  40  60  80  Cell Mean  Fisher's PLSD for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  -78.822  15.974  <.0001  Final, Resyl  6.296  15.974  .4257  85.118  15.974  <.0001  Initial, Resyl  74  100  DCM TB ANOVA Table for Some of O/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  4195.272  2097.636  5.263  .0159  10.525  .769  18  7174.807  398.600  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  4  48.433  15.789  7.894  Final Initial  10  17.111  12.454  3.938  Resyl  7  43.333  28.957  10.945  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval Resyl  Initial  Final  30  40  50  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % r  Final, Initial Final, Resyl Initial, Resyl  Mean Diff.  Crit. Diff.  P-Value  31.322  24.815  .0162  5.101  26.290  .6884  -26.221  20.671  .0158  75  D C M TB ANOVA Table for Some of 0/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  55608.833  27804.416  32.074  <.0001  64.148  1.000  31  26873.422  866.885  Means Table for Some of O/l - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  11  -29.174  44.207  13.329  Initial  12  44.787  16.704  4.822  Resyl  11  -47.631  20.644  6.225  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval  -80  -60  -40  -20  0  20  40  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value <.0001  Final, Initial  -73.961  25.066  Final, Resyl  18.457  25.605  .1516  Initial, Resyl  92.418  25.066  <.0001  76  60  DSLTR ANOVA Table for Some of O/T - U/S(ms) x -1 DF S o m e of  Initial/Resyl/Final  Residual  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  9718.047  4859.023  33.768  <.0001  67.536  1.000  28  4029.060  143.895  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  11  11.932  12.352  3.724  Initial  10  43.563  10.546  3.335  Resyl  10  1.286  12.920  4.086  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval  0  10  20  30  40  50  60  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  -31.632  10.736  <.0001  Final, Resyl  10.645  10.736  .0518  Initial, Resyl  42.277  10.989  <.0001  77  LIN TB ANOVA Table for Some of 0/T - U/S(ms) x -1 DF S o m e of  Initial/Resyl/Final  Residual  Sum of Squares  Mean Square  2  4596.974  2298.487  23  14888.843  647.341  Means Table for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  9  14.155  22.475  Initial  10  -16.980  25.197  7.968  Resyl  7  -3.225  29.252  11.056  Final  Std.  Err.  7.492  Interaction Bar Plot for Some of O/T • U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval  i 40  -30  1  —i -20  1  —i -10  1  —I— —i— —r 1  1  0  10  20  30  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  31.135  24.183  .0139  Final,  17.381  26.524  .1884  -13.754  25.938  .2840  Initial,  Resyl Resyl  78  40  F-Value 3.551  P-Value .0453  Lambda 7.101  Power .596  LIN T R  ANOVA Table for Some of O/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  23132.920  1 1566.460  38.133  <.0001  76.265  1.000  26  7886.351  303.321  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  11  -29.582  17.988  5.424  Initial  10  25.983  18.447  5.834  Resyl  8  -37.786  15.061  5.325  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval Resyl  Initial  Final  -20  0 Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Final, Initial Final, Resyl Initial, Resyl  Mean Diff.  Crit. Diff.  P-Value  -55.565  15.642  <.0001  8.203  16.635  .3201  63.768  16.981  <.0001  S  S  79  MIY T B ANOVA Table for Some of 0/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  35473.208  17736.604  6.952  .0048  13.904  .893  21  53576.987  2551.285  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  8  9.091  51.404  18.174  Initial  6  -55.848  32.109  13.108  Resyl  10  41.281  57.663  18.235  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval Resyl  Initial  Final  -100  -60  -20  20  60  100  Cell Mean  Fisher's PLSD for Some of 0/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Diff.  P-Value  Final, Initial  64.939  Crit.  56.729  .0268  Final, Resyl  -32.190  49.826  .1934  Initial, Resyl  -97.129  54.243  .0013  80  MIY TR ANOVA Table for Some of O/T - U/S(ms) x -1 DF S o m e of  Initial/Resyl/Final  Residual  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  15509.497  7754.748  4.418  .0219  8.835  .712  27  47395.981  1755.407  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  11  -52.469  33.918  10.227  Initial  10  -.169  55.167  17.445  Resyl  9  -41.259  32.598  10.866  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Diff.  P-Value  Final, Initial  Mean Diff. -52.300  37.562  .0081  Final, Resyl  -11.210  38.639  .5566  41.090  39.499  .0420  Initial, Resyl  Crit.  81  PTM T B - NOT SIGNIFICANT ANOVA Table for Some of 0/T - U/S(ms) x -1 S o m e of  Initial/Resyl/Final  Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  874.496  437.248  .819  .4609  1.638  .158  14  7474.842  533.917  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  3  67.875  25.833  14.914  Initial  10  48.689  12.866  4.068  Resyl  4  55.977  39.372  19.686  Final  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval Resyl  Initial  Final  0  20  40  60  80  100  120  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Diff.  P-Value  Initial  19.186  32.624  .2278  Final, Resyl  11.898  37.851  .5112  Initial, R e s y l  -7.288  29.319  .6023  Final,  Crit.  82  140  PTM TR ANOVA Table for Some of O/T - U/S(ms) x -1 DF S o m e of  Initial/Resyl/Final  Residual  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  5273.942  2636.971  5.948  .0068  11.896  .852  29  12857.276  443.354  Means Table for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Count  Mean  S t d . Dev.  S t d . Err.  Final  12  31.450  24.349  7.029  Initial  10  43.591  8.871  2.805  Resyl  10  1 1.471  25.005  7.907  Interaction Bar Plot for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Error Bars: 95% Confidence Interval  0  10  20  30  40  50  60  Cell Mean  Fisher's PLSD for Some of O/T - U/S(ms) x -1 Effect: Some of Initial/Resyl/Final Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value  Final, Initial  -1 2 . 1 4 1  18.439  .1885  Final, Resyl  19.980  18.439  .0347  Initial  32.121  19.259  .0019  Resyl  83  Appendix V : Individual results for Magnitude A G L Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  458.636  229.318  169.191  <.0001  338.381  1.000  74  100.298  1.355 |  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  27  20.813  .997  .192  Initial  27  15.220  .687  .132  Resyl  23  19.519  1.682  .351  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 22.5 20 17.5 c  1  5  | 12.5 % u  10 7.5 5 2.5 0  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  5.592  .631  <.0001  Final, Resyl  1.293  .658  .0002  -4.299  .658  <.0001  Initial, Resyl  84  A G L TB ANOVA Table for Some of Dist DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  216.326  108.163  34.828  <.0001  69.656  1.000  27  83.853  3.106  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  12  11.783  1.476  .426  Initial  8  18.450  1.007  .356  Resyl  10  13.780  2.422  .766  Final  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value  Final, Initial  -6.667  1.650  <.0001  Final, Resyl  -1.997  1.548  .0134  4.670  1.715  <.0001  Initial, Resyl  85  AGLTR ANOVA Table for Some of Dist DF S o m e of  Final/lnitial/Resyl  Residual  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  30.142  15.071  5.956  .0054  11.911  .865  41  103.750  2.530  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count Final  15^  Initial Resyl  Mean  S t d . Dev.  S t d . Err.  10.733  1.605  .414  16  9.019  1.831  .458  13  10.746  1.204  .334  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 9596 Confidence Interval  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  C r i t . Diff.  P-Value  Initial  1.715  1.155  .0046  Final, R e s y l  -.013  1.217  .9831  -1.727  1.200  .0058  Final,  Initial, R e s y l  S  S  86  A S N Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  404.087  202.043  131.532  <.0001  263.064  1.000  65  99.845  1.536  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  24  31.079  S t d . Dev. 1.307  S t d . Err. .267  Initial  21  25.624  .936  .204  Resyl  23  30.691  1.399  .292  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  5.455  .740  <.0001  Final, Resyl  .388  .722  .2872  -5.067  .747  <.0001  Initial, Resyl  87  ASNTB ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  62.705  31.352  47.977  <.0001  95.954  1.000  30  19.605  .653  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Final  Count  Mean  S t d . Dev.  S t d . Err.  12  16.892  .334  .096  14.1 1 1  .697  .232  1.148  .331  Initial  9  Resyl  12  17.433  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean D i f f . . C r i t . Final, Initial Final, Resyl Initial, R e s y l  Diff.  P-Value  2.781  .728  <.0001  -.542  .674  .1112  -3.322  .728  <.0001  S  S  88  A S N TR - NOT SIGNIFICANT ANOVA Table for Some of Dist DF S o m e of F i n a l / l n i t i a l / R e s y l Residual  S u m of Squares  Lambda  Power  .237  .7901  .475  .083  1.118  .559  75.411  2.357  Count  Mean  12  10.308  1.759  .508  Initial  12  9.883  1.199  .346  Resyl  11  10.164  1.598  .482  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 12  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Crit.  Diff.  P-Value  Final, Initial  .425  1.277  .5026  Final, Resyl  .145  1.305  .8228  -.280  1.305  .6647  Initial, R e s y l  P-Value  2  Final  Mean Diff.  F-Value  32  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl  Final  Mean Square  89  BWG Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  133.720  66.860  52.918  <.0001  105.836  1.000  42  53.066  1.263  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  18  26.032  1.107  .261  Initial  18  22.709  1.228  .289  Resyl  9  26.555  .907  .302  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 3 0 -i  '  1  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Diff.  P-Value  Final, Initial  3.322  .756  <.0001  Final, Resyl  -.524  .926  .2601  -3.846  .926  <.0001  Initial, Resyl  Crit.  S  S  90  BWG T B - NOT SIGNIFICANT ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  8.626  4.313  1.814  .1830  3.629  .333  26  61.807  2.377  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  10  17.510  1.222  .386  Initial  9  18.689  2.011  .670  Resyl  10  17.510  1.335  .422  Final  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 22.5 1  '  '  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit.  Diff.  Final, Initial  -1.179  1.456  Final, Resyl  0.000  1.417  Initial, R e s y l  1.179  1.456  P-Value .1081  • .1081  91  BWG TR - NOT SIGNIFICANT ANOVA Table for Some of Dist DF S o m e of F i n a l / l n i t i a l / R e s y l Residual  S u m of Squares  2  2.228  1.114  26  36.950  1.421  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  11  17.664  1.346  .406  Initial  9  17.256  1.164  .388  Resyl  9  17.956  .999  .333  Final  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff,  C r i t . Diff.  P-Value  .408  1.101  .4532  Final, R e s y l  -.292  1.101  .5905  Initial, R e s y l  -.700  1.155  .2240  Final,  Initial  r  Mean Square  92  F-Value  P-Value  Lambda  Power  .784  .4671  1.568  .164  CPT Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  308.573  154.286  216.881  <.0001  433.762  1.000  55  39.126  .711  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  20  21.628  .689  .154  Initial  20  16.809  .877  .196  Resyl  18  21.698  .954  .225  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  4.819  .535  <.0001  Final, Resyl  -.070  .549  .7990  -4.889  .549  <.0001  Initial, Resyl  93  CPT T B ANOVA Table for Some of Dist S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  12.103  6.052  3.470  .0462  6.940  .592  26  45.345  1.744  DF S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  10  19.260  1.354  .428  Initial  10  20.250  1.344  .425  Resyl  9  18.678  1.254  .418  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 22.5 1  '  '  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  C r i t . Diff.  P-Value  Final, Initial  -.990  1 214  .1057  Final, Resyl  .582  1 247  .3461  1.572  1 247  .01 5 5  Initial, Resyl  94  CRT TR ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  132.529  66.264  21.108  <.0001  42.216  1.000  27  84.761  3.139  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  10  17.880  1.457  .461  Initial  10  13.270  1.700  .537  10  17.560  2.099  .664  Resyl  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  C r i t . Diff.  Final, Initial  4.610  1.626  <.0001  Final, Resyl  .320  1.626  .6895  -4.290  1.626  <.0001  Initial, Resyl  P-Value  95  D C M Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist  DF  S u m of Squares  Mean Square  2  428.852  214.426  64  83.460  1.304  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  23  24.004  S t d . Dev. 1.503  S t d . Err. .313  Initial  22  18.570  .790  .168  Resyl  22  23.907  .992  .211  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff. Final, Initial Final, Resyl Initial, Resyl  Diff.  P-Value  5.434  Crit.  .680  <.0001  .097  .680  .7763  -5.337  .688  <.0001  96  F-Value  P-Value  Lambda  Power  1 6 4 . 4 2 8 I <.0001  328.857  1.000  j  D C M TB ANOVA Table for Some of Dist DF  S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  9.853  4.926  4.842  .0148  9.685  .764  31  31.538  1.017  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  13  9.531  S t d . Dev. 1.339  S t d . Err. .371  Initial  10  10.830  .634  .201  Resyl  11  9.891  .801  .241  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 12 n  •  1  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Diff.  P-Value  Initial  -1.299  .865  .0045  Final, R e s y l  -.360  .843  .3901  .939  .899  .0411  Final,  Initial, R e s y l  Crit.  S  S  97  D C M TR ANOVA Table for Some of Dist S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  16.720  8.360  6.561  .0041  13.123  .892  32  40.772  1.274  DF S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  12  10.358  S t d . Dev. 1.267  S t d . Err. .366  Initial  12  8.975  .794  .229  Resyl  11  10.500  1.272  .384  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Diff.  P-Value  Final, Initial  1.383  .939  .0052  Final, Resyl  -.142  .960  .7656  -1.525  .960  .0028  Mean Diff.  Initial  Resyl  Crit.  98  DSL Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  286.501  143.251  123.465  <.0001  246.930  1.000  46  53.372  1.160  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  13  17.896  .457  .127  Initial  20  13.364  .465  .104  Resyl  16  18.551  1.766  .441  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  4.532  .772  <.0001  Final, Resyl  -.655  .810  .1105  -5.187  .727  <.0001  Initial, Resyl  99  DSL TB ANOVA Table for Some of Dist DF S o m e of F i n a l / l n i t i a l / R e s y l Residual  S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  53.313  26.656  23.156  <.0001  46.311  1.000  27  31.082  1.151  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  10  17.860  1.085  .343  Initial  10  21.010  1.344  .425  Resyl  10  18.690  .684  .216  S t d . Dev.  S t d . Err.  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Diff.  P-Value  Final, Initial  Mean Diff. -3.150  Crit.  .985  <.0001  Final, Resyl  -.830  .985  .0951  Initial, R e s y l  2.320  .985  <.0001  100  DSLTR ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  19.710  9.855  19.983  <.0001  39.966  1.000  28  13.809  .493  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  11  15.391  .667  .201  Initial  10  13.540  .619  .196  Resyl  10  15.020  .811  .256  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Diff.  P-Value  Initial  Mean Diff. 1.851  .629  <.0001  Final, R e s y l  .371  .629  .2369  -1.480  .643  <.0001  Final,  Initial, R e s y l  Crit.  S  S  101  LIN Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  463.083  231.542  145.162  <.0001  290.324  1.000  55  87.728  1.595  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  21  23.479  S t d . Dev. 1.212  S t d . Err.  Initial  20  17.714  .907  .203  Resyl  17  23.861  1.633  .396  .265  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 30 1  '  '  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Final, Initial Final, Resyl Initial, Resyl  Mean Diff.  Crit. Diff.  P-Value  5.765  .791  <.0001  -.382  .826  .3585  -6.147  .835  <.0001  S  S  102  LIN TB ANOVA Table for Some of Dist DF  S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  5.966  2.983  5.057  .0136  10.114  .777  27  15.926  .590  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  10  17.550  .864  .273  Initial  10  18.580  .639  .202  Resyl  10  17.750  .784  .248  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff. Final,  Diff.  P-Value  .705  .0058  -.200  .705  .5652  .830  .705  .0227  Initial  -1.030  Final, R e s y l Initial, R e s y l  Crit.  S  S  103  LIN TR ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  24.707  12.353  20.138  <.0001  40.276  1.000  27  16.563  .613  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  11  19.255  .701  .211  Initial  10  17.100  .707  .224  Resyl  9  18.478  .946  .31 5  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 22.5 n  '  '  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff. Final, Initial Final, Resyl Initial, R e s y l  Diff.  P-Value  2.155  Crit.  .702  <.0001  S  .777  .722  .0360  S  -1.378  .738  .0007  S  104  MIY Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist DF  S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  19.918  9.959  8.588  .0006  17.175  .970  56  64.942  1.160  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  21  24.481  1.046  .228  Initial  19  23.079  1.232  .283  Resyl  19  23.965  .935  .215  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 30  i  '  '  25  c  2 0  ro QJ  =  1 5  O 10 5  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Diff.  P-Value  Initial  Mean Diff. 1.403  .683  .0001  Final, R e s y l  .516  .683  .1356  -.886  .700  .0140  Final,  Initial, R e s y l  Crit.  S  S  105  MIY T B ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  S u m of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  47.622  23.811  6.279  .0060  12.559  .868  26  98.590  3.792  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  10  17.290  1.083  .342  Initial  9  15.000  3.209  1.070  Resyl  10  18.070  .793  .251  Final  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value  Final, Initial  2.290  1.839  .0166  Final, Resyl  -.780  1.790  .3786  -3.070  1.839  .0020  Initial, R e s y l  106  MIY TR ANOVA Table for Some of Dist DF  S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  90.613  45.306  26.253  <.0001  52.506  1.000  28  48.321  1.726  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  11  13.200  1.373  .414  Initial  10  9.230  1.299  .411  Resyl  10  12.400  1.261  .399  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit.  Diff.  P-Value  Final, Initial  3.970  1.176  <.0001  Final, Resyl  .800  1.176  .1744  -3.170  1.203  <.0001  Initial, Resyl  107  PTM Lip Aperture (Recall that the distance measured is the distance between the lips) ANOVA Table for Some of Dist DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  513.867  256.934  278.391  <.0001  556.783  1.000  58  53.530  .923  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  21  26.984  1.108  .242  Initial  20  20.708  .373  .083  Resyl  20  26.787  1.177  .263  S t d . Dev.  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval 3 0 "j  '  '  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff.  Crit. Diff.  P-Value  Final, Initial  6.277  .601  <.0001  Final, Resyl  .197  .601  .5135  -6.079  .608  <.0001  Initial, Resyl  108  PTM TB ANOVA Table for Some of Dist S o m e of F i n a l / l n i t i a l / R e s y l Residual  DF  S u m o f Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  41.516  20.758  20.387  <.0001  40.774  1.000  32  32.582  1.018  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  Final  13  16.292  S t d . Dev. 1.028  .285  Initial  10  18.160  1.066  .337  Resyl  12  15.442  .939  .271  S t d . Err.  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 9596 Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Diff.  P-Value  Final, Initial  Mean Diff. -1.868  .865  .0001  S  Final, Resyl  .851  .823  .0432  S  2.718  .880  <.0001  S  Initial,  Resyl  Crit.  109  PTM TR ANOVA Table for Some of Dist DF  Sum of Squares  Mean Square  F-Value  P-Value  Lambda  Power  2  25.502  12.751  8.075  .0014  16.150  .951  33  52.110  1.579  S o m e of F i n a l / l n i t i a l / R e s y l Residual  Means Table for Some of Dist Effect: Some of Final/lnitial/Resyl Count  Mean  S t d . Dev.  S t d . Err.  Final  12  17.783  .964  .278  Initial  12  15.783  1.364  .394  Resyl  12  17.217  1.395  .403  Interaction Bar Plot for Some of Dist Effect: Some of Final/lnitial/Resyl Error Bars: 95% Confidence Interval  Final  Initial Cell  Resyl  Fisher's PLSD for Some of Dist Effect: Some of Final/lnitial/Resyl Significance Level: 5 % Mean Diff. Final, Initial Final, Resyl Initial, R e s y l  2.000  Crit.  Diff.  P-Value  1.044  .0004  .567  1.044  .2773  -1.433  1.044  .0086  110  

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.831.1-0091851/manifest

Comment

Related Items