I Bag Your Pardon: The Albertan ae/ɛ Vowel Shift as a Window into
Transcription
I Bag Your Pardon: The Albertan ae/ɛ Vowel Shift as a Window into
UNIVERSITY OF CALGARY I Bag Your Pardon: The Albertan ae/ɛ Vowel Shift as a Window into Community Grammars by Jacqueline Jones A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS GRADUATE PROGRAM IN LINGUISTICS CALGARY, ALBERTA DECEMBER, 2015 © Jacqueline Jones 2015 Abstract This thesis explores a vowel shift by speakers in Alberta in which [æ] is shifting before [g]. A production experiment was designed to examine the direction, extent, and sources of this change. I hypothesized that differing prompt modalities might elicit productions that could be used to support the existence of a triadic grammar, where productions are influenced by the community (auditory), the self (pictorial), and the standard (orthographic) grammars. I hypothesize a refinement of Ohala’s Active Listener hypothesis to include “super” and “inactive” listeners as other possible sources of sound change. The results show this is an in-progress merger by approximation and that changing the prompt modality altered listener productions. Auditory prompts had the greatest influence on production: Those most likely to merge the vowels [æ] and [ɛ] were more likely to mimic auditory cues. This indicates that super perceivers spread sound change to a greater degree. ii Acknowledgements Primarily, I would like to thank my supervisor, Dr. Stephen Winters, for his support and guidance throughout this process. This work would not exist without his advice and patience. I would also like to thank Dr. Darin Flynn and Dr. Robert Murray, whose work and teachings inspired many of the questions you find herein. Thank you to Dr. Flynn, and Dr. Hayashi for agreeing to be my examiners. I’d like to express my gratitude to my friends. My linguistics crews, both the original (Rein, Jen, Adrienne, and Jessi), and the new (Sarah, Kelly, Una, Saskia). The friendship and much-needed humour provided by the OLLC, the Phonetics Club, and the Cool Freaks Cabal cannot be underestimated. Extra special thanks are required for Sarah Greer, who not only served as support, friend, and fellow Old Lady, but provided the extensive time and effort to serve as the voice for my auditory stimuli. Thanks also to Emmett, a great friend and midnight-hour proofreader. Finally, Emmett (the other one) and Saskia deserve gratitude for delivering some tasty chicken curry at the time it was most needed. I wish to acknowledge my family: Andrew, for always being there with hugs, flattery, and the willingness to listen while I talked through my ideas before committing them to paper; My mom, for bringing me fresh-baked goods and help with the laundry; and Io, for being very soft and offering the support that only a cat sitting on the keyboard at 2:00am can provide. When I was a little girl, I stole a book from the school's library about a little girl who stole a book from the library. Diane Duane, through Kit and Nita, ignited in me a passion and lifelong love of the inner workings of language, and I would be totally remiss to not acknowledge her influence on my life. She taught me that you can open any door, change hearts, and even change yourself, just by finding the right words at the right time, and saying them with iii the right intentions. (I eventually returned the book, although to a different library. I didn’t quite know how those things worked). When I wrote to her, decades later, with the gushing fangirlism reserved for childhood heroes, she told me "I suspect you might have found your way to where you are without me, but I'm so pleased to have been of use on the journey. Go well." Dai’stiho, Diane. This thesis could not have been completed without the support of the University of Calgary, my peers in the Linguistics department (and later, the LLC), and the SSHRC. iv Dedication To Cynthia. Do you see that sassy language maven over there? The one with the quick wit? The church lady with the tricycle and the blue hair? …That’s you, that is. v Table of Contents Abstract ............................................................................................................................... ii Acknowledgements ............................................................................................................ iii Dedication ............................................................................................................................v List of Tables ................................................................................................................... viii List of Figures and Illustrations ......................................................................................... ix INTRODUCTION ..................................................................................1 BACKGROUND ...................................................................................4 2.1 Introduction. ...............................................................................................................4 2.2 Literature Review: Pre-Velar Raising........................................................................6 2.2.1 Previous Studies ................................................................................................6 2.2.2 Motivations for Pre-velar Raising .....................................................................8 2.3 A brief foray into classical Neogrammarian theories of sound change ...................13 2.4 Eckert’s Indexical Fields .........................................................................................16 2.5 Ohala’s Active Listeners ..........................................................................................20 RESEARCH OBJECTIVES AND HYPOTHESES ........................24 3.1 Introduction ..............................................................................................................24 3.2 The Three Grammars ...............................................................................................24 3.2.1 Introduction .....................................................................................................24 3.2.2 Self grammar ...................................................................................................25 3.2.3 Community Grammar ......................................................................................25 3.2.4 Standard grammar ...........................................................................................27 3.2.5 The hidden influence of the triadic grammar ..................................................27 3.3 Identifying Mergers .................................................................................................28 3.4 Pilot Study................................................................................................................30 3.4.1 Participants ......................................................................................................30 3.4.2 Stimuli .............................................................................................................31 3.4.3 Procedure .........................................................................................................31 3.4.4 Results of Pilot ................................................................................................33 3.4.5 Pilot Conclusions .............................................................................................36 METHODOLOGY ............................................................................38 4.1 Research Objectives .................................................................................................38 4.2 Participants...............................................................................................................40 4.3 Materials and Procedure ..........................................................................................41 4.3.1 Demographic and Personality Questionnaire ..................................................41 4.3.2 Calibration .......................................................................................................42 4.3.3 Modality Stimuli ..............................................................................................42 4.3.4 Procedure .........................................................................................................46 4.4 Analysis ...................................................................................................................48 4.4.1 Finding Mergers and Splitters .........................................................................48 4.4.2 Correlations With Personality Characteristics .................................................54 4.4.3 ANOVAs .........................................................................................................56 vi 4.4.3.1 Introduction to ANOVAs.......................................................................56 4.4.3.2 ANOVAs Without [œ] Auditory Stimuli...............................................59 4.4.3.3 ANOVAs Run On Auditory Stimuli. .....................................................73 4.4.3.4 Influence of Order of Presentation. ........................................................77 DISCUSSION AND CONCLUSION .................................................79 5.1 Confirming Mergers ................................................................................................79 5.2 Personality and Demographic Data .........................................................................80 5.3 The Grammar Triad’s Variable Influence on Production. .......................................81 5.4 Factors influencing production ................................................................................83 5.4.1 When the only targets are [æ] and [ɛ]..............................................................83 5.4.2 In the [œ]/Non-English vowel condition .........................................................86 5.5 Conclusion ...............................................................................................................87 REFERENCES ..................................................................................................................89 APPENDIX A: PILOT STUDY STIMULI .....................................................................100 APPENDIX B: SUBJECT QUESTIONAIRRE ..............................................................105 APPENDIX C: CALIBRATION VOWEL WORD LIST ...............................................107 APPENDIX D: EXPERIMENTAL STIMULI WORD LISTS .......................................108 APPENDIX E: PLOTS OF ANOVA RESULTS ............................................................113 APPENDIX F: POST HOC RESULTS ...........................................................................122 vii List of Tables Table 4-1: Significant Effects and Interactions in ANOVA on distance from calibration [æ]..... 59 Table 4-2: Significant effects and interactions in ANOVA on distance from calibration [ɛ] ...... 60 Table 4-3: Perceptual Possibilities for Auditory Stimuli .............................................................. 74 Table 4-4: Effects of experimental factors on participants' [æ] vowel: Significant results. ......... 75 Table 4-5: Effects of experimental factors on participants' [ɛ] vowel: Significant results. .......... 76 viii List of Figures and Illustrations Figure 2-1: Western Canadian Vowels. Lumber, 2008. Based on Labov et al., (2006). ................ 4 Figure 2-3: “The West” Dialect Region, from Labov et al. 2006 ................................................... 7 Figure 2-4: Liberman et al (1957), reproduced from Casserly & Pisoni (2010) .......................... 10 Figure 2-5: Frequency of second formant versus frequency of first formant for vowels spoken by men and children, which were classified unanimously by all listeners. (Peterson & Barney 1952) ......................................................................................................................... 12 Figure 2-6: taken from Eckert (2008) ........................................................................................... 17 Figure 3-1: Merger by Approximation (two subtypes). (Taken from Wassink, 2014)................. 28 Figure 3-2: Merger by Expansion ................................................................................................. 29 Figure 3-3: Merger by Transfer .................................................................................................... 30 Figure 3-4: Comparison of vowels [æ] (light) and [ɛ] (dark) before [g] and [k] .......................... 33 Figure 3-5: Productions by Modality: Pictorial (crosshatches), or Orthographic (shape) ............ 34 Figure 3-6: Combined Formants of Both Vowels and Contexts .................................................. 35 Figure 3-7: Gender differences in production of [æ] (green) and [ɛ] (blue) before [g] and [k] ... 36 Figure 4-1: Training slide from the Pictorial block ...................................................................... 44 Figure 4-2: Pictorial slide to prompt the nonword "bregg" .......................................................... 45 Figure 4-3: Labelled Spectrogram as prepared for each token by Praat and Perl Scripts. ........... 47 Figure 4-4: Comparison of Stimuli Vowels to Calibration Vowels, All Speakers ....................... 48 Figure 4-5: Production distances between calibration vowels and those before [g]..................... 50 Figure 4-6: A velar pinch. Image reproduced from Baker et al. (2007), green circle added by me. ......................................................................................................................................... 52 Figure 4-7: A Merger and a Splitter's formants of both vowels before [g] .................................. 54 Figure 4-8: Correlation Matrix for Personality data (“Merger” highlighted) ............................... 56 Figure 4-9: Normalized Comparison of Stimuli Vowels to Calibration Vowels. ......................... 58 Figure 4-10: Main Effect of: Vowel Target .................................................................................. 61 ix Figure 4-11: Main Effect of: Context ........................................................................................... 62 Figure 4-12: Main Effect of: Word ............................................................................................... 63 Figure 4-13: Main Effect of: Merge (only significant in regards to [æ])...................................... 63 Figure 4-14: Main Effect of: Stimtype (only significant in regards to [ɛ])................................... 64 Figure 4-15: Two-Way Interaction between Target and Context ................................................. 65 Figure 4-16: Two-Way Interaction between Target and Merge ................................................... 65 Figure 4-17: Two-Way Interactions between Context and Merge ............................................... 66 Figure 4-18: Two-Way Interaction between Stimulus Type and Word ....................................... 67 Figure 4-19: Two-Way Interaction between Target and Word .................................................... 68 Figure 4-20: Two-Way Interactions between Context and Word ................................................. 68 Figure 4-21: Three-Way Interaction between Stimtype, Target, and Context. ............................ 69 Figure 4-22: Three-way Interaction between StimType, Target, and Context with regards to Calibration [ɛ] ....................................................................................................................... 70 Figure 4-23: Three-Way Interaction between Stimtype, Target, and Word. ................................ 71 Figure 4-24: Three-way Interactions between Target, Context, and Merge ................................. 72 Figure 4-25: Three-way Interactions between Stimtype, Context, and Word. ............................. 72 Figure 4-26: Three-way Interactions between Target, Context, and Word. ................................. 73 Figure 4-27: Main Effect of: Presentation Order (With Target Included) .................................... 78 Figure 5-1: Normalized Plot of Mergers' and Splitters' [æg] and [ɛg] productions [_g], [_k], and Calibration contexts. ...................................................................................................... 79 x Introduction A WORD is dead When it is said, Some say. I say it just Begins to live That day. — Emily Dickinson (Part One: Life: 89, line 1-6, 1924) It might seem a little presumptuous to begin a Master’s thesis in Linguistics with a quote from a poet. After all, isn’t poetry the exception to the rules? Aren’t the greatest wordsmiths those that are considered to defy the expected and descriptive rules that linguists and educators so carefully attempt to control and harness and write manuscripts about? But each time someone speaks, even when reading a poem that has been read aloud thousands of times before, the sounds that person produces are never exactly like those produced previously. Even the same speaker reading the same poem on the same day does not produce exactly the same sounds. Sometimes, these minute variations bring about larger changes. For example, Dickinson spoke with a characteristic New England non-rhotic accent, as did most speakers throughout Maine, New England, and Rhode Island in the 19th century, dropping the [ɹ] from “word” to pronounce it as [wɜːd] (Miller, 2012:49). However, a speaker living in her home town of Amherst today would likely pronounce it [wɜɹd], as, according to Labov et al., (2005:48), since the non-rhotic dialect area has shrunk and now encompasses only the eastern coastline of Massachusetts. As Dickinson says, it is when words are spoken that they are given life enough to change, and this thesis concerns itself with the forces that shape and constrain that change: the speaker, the listener, and the environment (which I will later identify as the primary influence on the self, community, and standard grammars, resspectively). The specific sound change I am examining 1 in this manuscript is the possible vowel merger in Western Canada between the low front vowels /æ/ and /ɛ/ before /g/, using three modalities (visual/pictorial, auditory, and orthographic) to measure the influence of the three forces on the change. Chapter 2 (Background) gives a brief explanation of vowel mergers of this type in other areas of the world, and details the merging process and what can be expected in regards to experimental findings to support the existence and direction of these changes. Chapter Two also gives an overview of two different explanations for sound change from a social perspective: Ohala’s active listeners and Eckert’s indexical field. I then propose an alternate catalyst of sound change, combining the two previous theories to form what I call a “triadic grammar.” In the triadic grammar, indexical fields and active/inactive listeners work in tandem to initiate, spread, and resist sound change. It is via the triadic grammar that the intersection of these aforementioned forces may be studied. Chapter 3 (Research Objectives and Hypotheses) outlines in detail the specifics of the phenomenon under study (that is, the æ/ɛ merger), and the hypotheses being tested. It also gives an overview of a pilot study completed earlier to help focus and refine the current study. Chapter 4 (Methodology) provides the details of the testing materials, participants, procedure, and results of the primary study conducted for this thesis. Chapter 4 also explains the statistical methods used and explains why some methods and materials were chosen over others for this particular study. Chapter 5 (Analysis) presents the results of the statistical analyses of the experimental data. Chapter 6 (Discussion and Conclusion) examines these results in light of the theories presented in Chapter 2, and discusses whether the hypotheses were confirmed or disconfirmed. Potential shortcomings, unexpected windfalls, and areas of future research are also highlighted. 2 It then summarizes all previous chapters, giving an overview of the thesis as a whole, while encouraging the reader to determine whether the words contained herein give empirical support to Dickinson’s intimation. I believe they do. 3 Background 2.1 Introduction. Vowels are undergoing drastic changes across North America. The so-called “caught/cot” merger of the low back vowels has been well-documented in Minnesota (Arctander et al, 2009), Missouri (Majors, 2005), and Kentucky (Irons, 2007). The high back vowels [u] and [ʊ] are fronted in California (Clopper et al 2005) and, to some extent, in Winnipeg English (Hagiwara, 1997). The retraction of the front lax vowels [æ], [ɛ], and [ɪ], dubbed the Canadian Shift, has been documented in Montreal (Boberg, 2005), Ontario (Clark et al., 1995), Vancouver (SadlierBrown et al., 2008), and Winnipeg (Hagiwara, 2006). This thesis examines a potential vowel merger in Western Canada between the low front vowels /æ/ and /ɛ/ before /g/. Figure 2-1 highlights the vowels under study. The graph represents Figure 2-1: Western Canadian Vowels. Lumber, 2008. Based on Labov et al., (2006). the “vowel space” from a sampling of speakers taken by Labov in western Canada between 1991 and 1993 (Labov, 2006). The X and Y axes represent the first formant (F1) and the second 4 formant (F2) of the vowel, the “characteristic resonance regions” (Peterson et al. 1952:175) or frequency bands that distinguish vowels from each other. Vowel maps are traditionally presented in this way, as they both elucidate acoustic structure, and represent (loosely) the tongue’s position in the mouth at each point of articulation. The F1 corresponds to vowel ‘height,’ referring to the highest point of the tongue’s position in the vocal tract, while F2 refers to vowel 'backness,' the position of the highest point of the tongue either forward towards the lips or back towards the throat. As displayed in the figure, the “normal” vowel positions for speakers in Western Canada have [æ] produced lower and farther back than [ɛ]. I have informally observed, however, that many speakers in the Calgary area, myself included, appear to be merging these two vowels in certain contexts, among them before [g]. This makes “bag” ([bæg]) and “beg” ([bɛg]) homophones for these speakers. Zeller (1997) first documented the existence of this particular change in parts of the American Midwest in 1990, and identified it as an [æ] → [e] merger, taking place in the contexts of [æg], [æŋ], and [æŋk]. It should be noted that most of the literature confirms with Zeller’s observation that General American English produces [e] in these contexts, while Canadian English speakers are considered to produce a sound closer to (and transcribed as) [ɛ] (Labov, 2005). This phenomenon of low front vowels increasing in F2 before [g] is called pre-velar raising (Freeman, 2014). The experiment conducted for this thesis will, as its first goal, determine the nature of the æ/ɛ merger in Western Alberta. That is, it will seek to confirm or deny the change within the population studied. After confirming that a significant percentage of speakers merge these vowels in certain contexts, the next step is to determine the direction of this shift. While these possibilities will be examined in depth in Chapter 3, the primary goals of this analysis were to 5 determine: (1) Whether [æ] is rising and fronting to encroach on the area generally reserved for [ɛ], or vice-versa ([ɛ] backing and lowering); (2) Whether both vowels are moving towards a central “non-æ, non-ɛ” point (a neutralization), or, finally; (3) A shift in other directions indicating a chain shift or other phenomena more complex than a simple merger or neutralization. 2.2 Literature Review: Pre-Velar Raising 2.2.1 Previous Studies To examine pre-velar raising, and the [æ]/[ɛ] shift, it is important to situate Alberta within the greater linguistic landscape of Canada (and, as we will see, the Pacific Northwestern region of the USA). Labov, Ash, and Boberg’s 2006 edition of the Atlas of North American English consider most of Canada a homogenous dialect region, saying “a single type of English is spoken across the 3,000 miles (4,500 km) from Vancouver, British Columbia to Ottawa, Ontario” (p. 216) However, there is a growing body of evidence that supports dialectal variation within the 4,500km2 outlined by Labov et al. as homogenous. Boberg (2008) found that [ɛ] was produced fronter in the Prairies (a higher F2) than all other regions in Canada, while Hagiwara (2006) found that in Winnipeg [ɛ] does not appear to be lowering. Though just outside of Labov’s region, in an earlier study Boberg (2005) also found that in Montreal the [ɛ] is moving towards [ʌ] and [æ] is both lowering and retracting. Complicating the distinction between the current variation in Alberta and some of the similar mergers noted in the American Pacific Northwest is the ongoing Canadian Shift. The merger between [ɔ] and [ɑ] (making caught and cot homophones) created a vacuum in the Canadian vowel space. [æ] shifted lower and backer into this space, and [ɛ] followed suit (also 6 moving lower and backer) (Labov et al. 2006.): This is why Canadians produce [ɛ] in this context, while the American vowel is [e]. According to Labov et al. (2006, p.222) in General Canadian English “It is evident that /e/ is moving backward and downward in apparent time, and /æ/ is moving backward.” This evidence is based on recordings and acoustic analysis of speakers across Canada between the years of 1991 and 1993 (p. 20). The extent to which Canadian dialect regions have been studied pales in comparison to those in the United States, but it is prudent to mention here those studies carried out on a region assumed by the Atlas to be similarily monolithic: The West. This region of the United States encompasses all or portions of every state west and north of Texas (Figure 2-3, from pp. 138) Figure 2-2: “The West” Dialect Region, from Labov et al. 2006 Researchers of American dialectology have already identified distinctive regions within The West: most relevant to the current work is that group called the Pacific Northwest (PNW), 7 which encompasses the states of Washington and Oregon (and optionally Idaho, Montana, portions of northern California, and southern British Columbia) (Findley & Coates, 2002, p.2). The varying positions of the [æ] and [e] vowels were noted in this area as early as 1961, where Reed observes an infrequent raising of [æ] in “bag” (p. 561). Zeller (1997) documented the existence of this particular change in Wisconson in 1990 and identified it as an [æ] → [e] merger, taking place in the contexts of [æg], [æŋ], and [æŋk]. Zeller’s findings were later republished in the Atlas (p.182), but no further comment on this merger is made within. Wassink et al. (2009) found, in a survey of 30 PNW speakers, that prevelar raising affected both [æ] and [ɛ], and also showed a gender divide: Male participants merged [æ], [ɛ], and [e:] (the vowel found in the word “bake”), while females maintained separate [æ] and [ɛ] vowels (but did tend to merge [ɛ] and [e:]). More recently, Wassink (2015) confirmed this raising of [æ] before [g] in the Pacific Northwest, and linked it to a continuing trend of both vowels towards the diphthong [eɪ], though she find this is a Washington-area trend not shared by speakers in Vancouver, BC. A small-scale study of 4 speakers from Oregon and Washington conducted by Riebold (2012) showed a similar tendency in half of the speakers to merge [æ] and [ɛ], though no gender divisions were found. Conn (2002) found evidence of the Canadian Shift in Oregon but no evidence of merged vowels. Finally, Riebold (2015) found significant pre-velar raising in [ɛ] before [g] across all participants, they found a subsequent raising of [æ] before [g], causing the two vowels to maintain distinct areas within the vowel space for most speakers, and so did not confirm a [ɛ]/[æ] merge. 2.2.2 Motivations for Pre-velar Raising Shifts involving [æ], have been documented widely throughout history. Old English [æ] lowered to become [a] in Middle English, which then shifted back upwards to [æ] in Early 8 Modern English (Baker, Mielke, Archangel, 2008). It is possible that [æ], rather than cycling downwards again to [a], is continuing to rise. This causes it to encroach on the vowel space occupied by [ɛ]. Once this encroachment happens, speakers are forced to alter their vowel spaces to accommodate the change, either through collapse or expansion. The options available to these speakers in regards to these two vowels can be explained cohesively via Dispersion Theory. According to Dispersion Theory (Liljencrants & Lindblom, 1972; see also Burzio 2007) “insufficient distance” between vowels in the vowel space is viewed as a form of markedness. The vowels [ɛ] and [æ] become insufficiently distant from each other when [æ] shifts higher and fronter in the vowel space, and so one of two repair strategies applies—either the vowels are pushed further apart in the vowel space (dispersion), or they are collapsed into a single phoneme (neutralization). If [ɛ] and [æ] are becoming neutralized before before [g] and [k], the explanation for this neutralization may have its roots in coarticulation. It is well documented that velar articulations drive F2 and F3 together immediately before closure, an effect known as the classic “velar pinch” (Ladefoged 2006:193). Bringing the F2 up in anticipation of this pinch may be behind the neutralization of the two vowels before velars. Bauer & Parker (2008) conducted a series of production experiments using ultrasound and concluded that this coarticulatory motivation was responsible for the raising of [æ] in Wisconsin (though, unlike Zeller, they did not identify a merger between [æ] and [ɛ[ before [g]). Many vowel shifts have taken place in North American English dialects even in the relatively brief time since colonization. The reason for this frequency is due to the gradient nature of vowels within the vowel space, and the way humans perceive and categorize these incoming sounds. Liberman et al. (1961) demonstrated the phenomenon of categorical 9 perception using synthetic speech and the stop sounds [b], [d], and [g]. Acoustically, two of the distinguishing features that separate each of the sounds from the others is: (1) The frequency of the second formant at the onset, and; (2) the length of delay between the onset of the first formant relative to the second (and third) formants. Liberman et al. (1957) created stimuli that altered this time delay in 10ms increments (therefore altering its F2 transition cue), and asked subjects to label the sounds in an ABX task. Figure 2-4 illustrates their results. Figure 2-3: Liberman et al (1957), reproduced from Casserly & Pisoni (2010) The Y-axis shows the percentage of participants that labelled the sounds as [b], [d], or [g]. The X-axis shows which stimulus was being labeled by participants, with each “stimulus value” varying in its F2 by 120 cycles per second (or Hertz) (recall that this change in F2 is a product of the 10ms time delay). As shown, humans’ “discrimination between different tokens of the same category (analogous to two shades of red) is very close to chance. They are highly accurate at discriminating tokens spanning category boundaries, on the other hand.”(Casserly & Pisoni 2010:4) This discrete cataloguing of incoming gradient data, known as categorical perception, is not limited to speech perception but has also been found to affect other modalities, such as size, colour, and texture (Goldstone & Hendrickson, 2010). 10 Categorical perception relates to the vowel shift under study for two reasons: It, like speech perception in general, is found to be cross-modal (and I am examining the influence of different modalities on production, which are based on listener perception of input), and because vowels in natural speech have much wider categories: They can be produced with much more variation than stop sounds and still be reliably identified (Pisoni 1973; Peterson & Barney 1952). Physiologically, the sounds [b] and [d] differ in their place of articulation: [b] is produced with the lips and [d] is produced with the tongue against the alveolar ridge (Ladefoged, 2006:10-11). The amount of variation with which speakers can produce these stop categories is limited, because each corresponds to a relatively stable "quantal state" (Stevens & Keyser (2010), where a wide amount of articulatory variation yields more or less the same acoustic output. The acoustic characteristics of vowels, on the other hand are more sensitive to variation in articulation, but the identities of those vowels can be still be gleaned by the listeners so long as these productions are “close enough”, acoustically, to other members of the category. It has also been found that listeners may engage short-term memory more when perceiving vowels than consonants: Pisoni (1973), in studying perceptual categories in vowels, found that “steady-state vowels have been found to be perceived continuously, much like nonspeech sounds… differences in discrimination between consonants and vowels are primarily due to the differential availability of auditory short-term memory for the acoustic cues distinguishing these two classes of speech sounds.” (pp.1-2). Figure 2-5 shows the results of an experiment (Peterson & Barney, 1952) in which 75 people were asked to categorize the vowels of men and children, and it illustrates that vowels are produced with variable points, but that their (relative) meaning remains clear as long as the vowels do not stray too far from the designated perceptual boundaries. F1 is represented on the 11 x-axis and F2 is represented on the y-axis. (It should be noted that this paper predates the conventional way to chart vowel spaces, and rotating the following figure 180° will place it in the same orientation as the other vowel charts in this thesis.) Figure 2-4: Frequency of second formant versus frequency of first formant for vowels spoken by men and children, which were classified unanimously by all listeners. (Peterson & Barney 1952) The figure also illustrates, once again, the influence of people’s knowledge of community grammars in perception. Listeners know that children’s voices are generally higher in both F1 and F2 than either women’s or men’s voices, due to their relatively short vocal tracts. As illustrated in Figure 2-5, the pure acoustic properties of the child’s [ʌ] vowel (the central sound in the word “bud”) places it closer acoustically to the [æ] produced by adult men than the adult 12 male [ʌ]. However, indexical (non-linguistic) knowledge, such as pitch (or, in the case of the world outside of the experiment booth, the visual knowledge of the speaker) allow listeners to accommodate for these factors and extract the intended meaning (Nearey, 1989, see also Klatt, 1979 and Johnson, 2008). To put it another way, listeners unconsciously adjust their own perception to align with what they know about the world. It seems likely that if you played listeners a pitched-up male voice producing these vowels, while showing them a picture of a child, you could create a misperception, similar to that found by McGurk (1976), in which the listener “hears” the vowels they would expect to hear from a child, rather than the altered speech of an adult source. In this way, community grammars and the application of paralinguistic knowledge are potentially able to contribute to sound change. 2.3 A brief foray into classical Neogrammarian theories of sound change The study of sound change as a phenomenon is one of the earliest concerns of linguists (or their predecessors, the philologists). Many of its methods of approach and rules of description date back to the 19th century (Koener, 1989:15) and the Neogrammarians, a group of German linguists at Leipzig. Among these scholars was Karl Bruggmann, who described sound change thusly: “Every sound law must be conceived of as allowing no exceptions; everything that diverges from it must be assumed to be due to analogical formation.” (Koener, 1995:24) This insistence on regularity is a marked change from the viewpoint of previous scholars. For example, even Jacob Grimm allowed exceptions in his famous Grimm’s Laws, describing the Germanic Sound Shift (Robins 2013:207). This insistence on regularity or movement towards eventual regularity is justified by Robins (2013:207-208). “If sound change were not regular, if word forms were subject to random, inexplicable, and unmotivated variation in the course of time, [descriptions of sound change] would lose their validity, and linguistic relations could be 13 established historically only by extra-linguistic evidence.” All irregularities, regardless of origin or spread, were filed under “analogy” by the Neogrammarians and regarded as not worth a researcher’s time to examine. This is quite a useful approach if one seeks to examine long-term diachronic change (changes over long periods of time), such as the changes undergone by related languages like German and English from their common ancestor Proto-Indo-European. But if one seeks to explain the motivations for change to happen in the first place, or to study a change in progress to examine its virility and identify the motivators that cause its adoption and spread, one cannot dismiss everything out of hand as “analogy.” Labov’s seminal 1963 work “The Social Motivation of a Sound Change” outlines specifically how random variation across speakers can lead to a change across most (and eventually all) of a dialect region. In the initial stage, variations are introduced into a language randomly, motivated by expected speech errors such as coarticulation or assimilation, analogy (in the Neogrammarian sense), borrowing, or “any number of processes in which the language system interacts with the physiological or psychological characteristics of the individual.” (pp. 1-2). In the second stage, one of these countless random variations begins to gain prestige and be used more widely, and in the last stage only one of the possible options for a given sound (including the original, nonvaried productions) remains, and the sound change (or lack thereof) is said to be complete. Labov argues that it is the work of a linguist to identify not just the sound changes themselves, but also the social structures that influence a given variable’s elevation or rise in prestige that allows it to “triumph”(p.2) over other variables and become a completed sound change. So while the Neogrammarian concern with regular mechanical (physiologicallymotivated) sound change is a useful starting point, I take Labov’s view that psychologically14 motivated meaning, classically dismissed as analogy, plays a significant, measurable role in this change affecting the [æ]/[ɛ] vowels in southern Alberta. The [æ]/[ɛ] merger studied herein can be classically considered a “conditioned regular change” (Campbell, 2004:15). This is a change that, according to Campbell “takes place uniformly wherever the phonetic circumstances … are encountered.” (p.17) – in this case, the rise in F2 caused by the raising of the back of the tongue in forming the following velar consonants [g] (and [k]/[ŋ]). This study was designed to examine, at its most basic level, the existence of and direction of the change, and why it appears to be affecting vowels before [g] but not [k]. To use classical terminology, I sought to determine whether æ > ɛ / _g (to be read as “[æ] becomes [ɛ] when preceding [g]”), or ɛ > æ / _g (“[ɛ] becomes [æ] when preceding [g]”), or possibly some third option where both vowels become something else. It should be noted that the classical study of sound change concerns itself with the underlying representations of sounds, or phonemes, which are related to, but not the same as their “surface” phonetic representations. Since I examine a sound change in progress, affecting only some speakers of those studied, and only under certain conditions, I remain agnostic about the state of these sounds at the phonemic level. I use the classic terminology only as a means of explaining the phenomenon and not as an assertion that these changes have reached some underlying mental representation. Whether the speakers studied who merge these vowels on the surface have merged the phonemes underlyingly is not a question I am prepared to answer. For these reasons, I will concern myself only with the phonetic reality of the change, and leave the questions of underlying phonemic representation aside. 15 2.4 Eckert’s Indexical Fields Eckert (2008) identifies sound change as a product of speakers’ ongoing navigation of their social geography and linguistic landscape. Social geography refers to those categories which Labov (1963) first identified in New York and Martha’s Vineyard: variables such as social class hierarchy, attitudes to geographical location, gender, etc. The linguistic landscape refers to possible variations of a given sound, based on articulatory possibility (and ease), as well as the frequency of given variations. The interaction between social geography and linguistic landscape creates the indexical field, created by the “continual reconstrual of the indexical value of a variable… A constellation of meanings that are ideologically linked (p.464),” or alternatively, “the various variables that might have related indexical value (p.473).” On the next page is an example indexical field for the variation between [ɪŋ] and [ɪn] for the (ING) ending in English (p.466). In the centre of the indexical field are personality traits people associate with speakers of the [ɪŋ] variant (“working”): They are seen as educated, formal, and articulate, and to use this more formal variant is seen as “effortful” by listeners. On the edges of this central point are perceptions of the [ɪn] variant (“workin’”): Uneducated, easygoing, etc. Different listeners evaluate the nature of the same variant’s traits differently. For example, the [ɪn] variant may be seen as easygoing (a positive trait) by a friend or coworker who would see the [ɪŋ] variant as too pretentious. A boss, though, may judge the [ɪn] variant as lazy (a negative trait), or see its lack of formality negatively. When speakers are aware of and have both variations open to them, the relevant indexical field can affect both listener perception and speaker production. 16 Figure 2-5: taken from Eckert (2008) Social geography enters the indexical field in that listener judgements and speaker perceptions are not limited to the pure acoustic output of a given variant and the indexical meanings attached to it, though they are both an inherent aspect of most social geographies. (One can imagine situations where these things are not part of the social geography, for example in a nunnery where all members have taken a vow of silence, but for the most part they are inextricable parts of the whole). For example, the (ING) variant as a marker of speaker authenticity will vary in listener perceptions, depending on the race, location, and reputation of the speaker (Bucholtz 2003:410). In a similar fashion, the perception of the indexical field does not necessarily have a basis in objective reality—it is arguably more effortful for a speaker who doesn’t use [ɪn] regularly to produce the “easygoing” form, though perceptions of that form as easygoing may not change. Influences outside of the individual speakers and listeners also affect the indexical field. A speaker whose native style/dialect includes the “effortful” variant is still 17 likely to view it as “effortful” (or, at least, view the other production as “lazy”) based on the society-wide perception of these variations. The [ɪn] variant acts as a social marker above the phonetic level, marking its users as lazy and uneducated, due to its location outside of the standard language forms. That is, speakers who most often use the [ɪn] variant are considered “less than” by those in power. Therefore, the variant itself becomes undesirable to those seeking power due to its association with less powerful groups (Lippi-Green, 1997:214). The concept of the indexical field is useful in that it posits a perceptual whole made from three disparate parts: the speaker, the listener, and the social landscape. For my thesis, I take this triadic approach and apply it to grammar. The indexical field approaches sound change as something that “unfolds in the course of day-to-day exchange, and that exchange involves constant local reinterpretation and repositioning.” (Eckert 2008:472). I seek to examine sound change at the individual level, positing a speaker’s production of variation as the outcome of three grammatical influences: the self, the community, and the standard (that is, prescriptive knowledge). The grammar of the self is unique to each individual. Its analog in the indexical field is the speaker. The self grammar is also the product of a given person’s experience. Under an exemplar model, it is the entire collection of traces and weightings that the person utilizes when producing speech. Within my experimental framework, the grammar of the self is studied via the use of pictorial stimuli, as they contain no outside phonetic (acoustic or articulatory) influence on production, and therefore participants must rely solely on their internal linguistic representations or traces to produce a word. Community grammar is the analog to Eckert’s listener. It is an individual’s perception of the community’s standards, variants, and phonetic variations. Since, as in all perception, each 18 individual’s experiences are unique, so too are concepts of community grammar. However, there is a wide range of overlap in separate individuals’ concepts of community grammar, depending on how much shared experience overlaps between individuals. For example, Johnstone & Kiesling (2008) found that Baby Boomers from Pittsburgh felt they spoke with the local accent, “whether they did or not” (p.27), thus identifying with the local community. In that case, the community grammar could be defined as an individual Pittsburgher’s collection of exemplars of speakers from Pittsburgh, filtered through their perception of “what it means to be from Pittsburgh.” In its simplest form, community grammar is how “they” talk, and can be understood as the average productions and variations produced by a given group of people. Within this paper, the auditory stimuli stand in for the community grammar. A single voice is a poor substitute for an entire community, but it is meant to test the influence of direct outside sources on an individual’s choice of variants in speech production. The influence of other speakers on listener productions, known as convergence or phonetic imitation, has been extensively studied (for examples, see Goldinger 2005, Alan et al. 2015, and Meyerhoff (2011) p.74). The standard grammar refers to the prescriptive rules of grammar taught and internalized by most speakers as the “right” way to speak within a given community. It takes the same position as the “landscape” in Eckert’s indexical field, but differs slightly in that it is ultimately a fabrication created by those in power, based on stereotypes. For example, the same middle-aged Pittsburghers described above, who identified with the local community, rated the local dialect as both “charming” and “embarrassing,” indicating that the standard grammar and community grammar can be at odds within an individual and create dissonance in speech. The orthographic 19 stimuli in my experiment represent the standard grammar, and were used to draw standard productions out of my participants. 2.5 Ohala’s Active Listeners Ohala (2012) posits that listeners are active agents in sound change. The active listener can incorrectly reinterpret sounds they’ve heard when there’s an extenuating factor in the input. These factors include simple speech errors, as well as perturbations in the signal caused by coarticulation, environmental disturbances (such as noise or other people talking), or psychological disturbances (such as too much cognitive load or a simple case of “not listening”). These incorrect assumptions about the sound’s production cause the listener-turned-speaker to reproduce these effects, and thus sound change spreads. In an ideal relationship, Ohala’s active listener may be an infant acquiring speech for the first time, in which case the “incorrect” representations become part of the child’s knowledge of their language. However, this hypothesis can also extend to adults, by assuming a gradual shift in production. As an example (taken from Flemming (2006) and Clements (1991)), California’s fronted /u/ may have begun in certain coronal environments, such as in the word [dud] “dude”, and that fronting spread to other environments not necessarily via “new” speakers, but by lexical diffusion, where active listeners apply what they heard (the coarticulation caused by /u/ preceding coronals) to new words first with similar environments (e.g. [tun] “tune”) and then to new environments, that no longer share coronal articulation, such as [bum] “boom.” Sound change via lexical diffusion is also supported by Bybee (2002), who tracked wordfinal t/d deletion (in words such as “grand”) and found that high-frequency items underwent this deletion first. Bybee’s work is important to this thesis because it posits that even Neogrammarian 20 “regular” sound changes are observed to take place gradually via lexical diffusion, provided a researcher can catch a given change before it has spread throughout an entire language. My hypothesis breaks down these agents in sound change farther, identifying not just active listeners as participants in sound change, but two additional types: “super perceivers” and “inactive listeners.” Inactive listeners are less likely to spread an innovation in speech, because their personality is self-focused. An inactive listener is less likely to attend to outside productions, and any shifts in those productions, as they place more focus on their own stored productions when producing speech. Super perceivers are the opposite—they attend to changes and rebuild signals erroneously more often, because they are less likely to “check” their own or stored productions compared to the speaker they have just heard. Super perceivers are more likely to spread a sound change, and inactive listeners are more likely to resist it. However, under this hypothesis, inactive listeners are more likely to initiate sound change, as they do not place enough weight on the (‘standard’) productions of others to prevent their own speech from drifting naturally. It is worth noting that this suggests a difference between initiators and transmitters of sound change. Inactive listeners are initiators, not transmitters; super perceivers are transmitters, not initiators. Alan Yu (2015) found that individuals who ranked higher on an Autism Quotient test, indicating neurotypical individuals with more “ “autistic” traits” (p.2) attended more to the phonetic productions of speakers than to the word targets, as measured by lower rates of perceptual compensation: That is, those with more autistic traits were more likely to be super listeners. It should be noted here that the individuals studied were not diagnosed as on the autistic spectrum themselves, but merely exhibited some traits considered “autistic.” That is, while people with autism may not be considered “other-focused” (a requirement of super 21 listeners), some traits found in those with autism are also those traits that contribute to the likelihood of a person being a super-listener. While I hypothesize that there are other personality types that make some individuals more likely to be super perceivers or inactive listeners, I think that these are tendencies and all individuals are able to attend (or ignore) changes in the acoustic signal at different times. That is, these tendencies within individuals are not discrete, but exist on a continuum. In rough terms, sound change can be explained by its relationship to the three groups: Super Listeners Active Listeners Inactive Listeners 1. Super Listeners (or perceivers): Spread sound change to a great degree. Less likely to innovate their own changes, but this is difficult to observe experimentally. More likely to apply shifts in environments where there is no physiological grounding for sounds to be “misperceived.” 2. Active Listeners (or perceivers): Spread sound change to a mild-to-moderate degree. Resist great shifts in language or only spread change in places where physiologically sensible. 3. Inactive Listeners (or perceivers): Conservative speakers. May initiate sound change, but always resist changes not of their own invention. In the following chapter, I outline the particulars of the phenomenon under study, and explain how the experimental design was structured. One goal of the study was to determine whether each of my participants was a super, active, or inactive listener-turned-speaker, based on 22 how much they attended to (or did not attend to) stimuli with sources in the self, community, or standard grammar. I also investigated whether these personality types affected the extent to which the participants exhibited the [æ] to [ɛ] sound change in progress in southern Alberta. 23 Research Objectives and Hypotheses 3.1 Introduction This chapter outlines my experimental goals. It begins with my proposition for a theoretical framework of a triadic grammar within speakers. It then examines how possible results can be interpreted in terms of identifying the type of change the vowels are undergoing. Finally, I give the details and results of the pilot study done to narrow the scope of the research. I conclude with the specific hypotheses and research objectives for the main project. 3.2 The Three Grammars 3.2.1 Introduction When approaching any kind of sound change, either completed or in progress, determining the empirical facts about the sounds involved answers only half the question. Providing a satisfying answer to the explanation of why the sound change is occurring must be considered with equal attention as that afforded to the what or the how. I have explained above that the classical Neogrammarian view does not concern itself too deeply with these questions, assuming either regularity or analogy sufficient to explain the motivations behind all sound changes. As I leave this view behind, I turn now towards more contemporary theories of sound change. Greater researchers than I have laid the groundwork, and I devote some time ahead to outlining their theories which I find most applicable to the current research. I then add to those theories my own idea of a triadic grammar, the three parts of which coexist inside each individual speaker to inform their speech decisions and thus contribute to sound change. The concept of the three grammars refers to three separate sets of mental representations of sound within a speaker: The self-grammar, the community grammar, and the standard grammar. I take the view that all three grammars outlined are arranged in an exemplar model in 24 the broadest sense. That is, “The representation of a concept consists of separate descriptions of some of its exemplars.” (Smith & Medlin, 2002:208). 3.2.2 Self grammar The self grammar refers to an individual’s sum of experiences with a given sound. This is the broadest category, and includes, under an exemplar model (Pierrehumbert, 2001), all instances in which the individual heard the sound uttered, including the ones they produced themselves. It also includes related exemplars, such as similar sounds. For example, the word (and sounds of the word) “hag” or “bog” may be activated when the listener hears the word “bag.” Non-phonetic exemplars still considered related to the parent category can also be included, such as the orthography for the word “bag” or an actual physical bag, or a drawing of one. It should be noted that the self grammar refers only to representations of the category, and not the objects or sounds themselves. If I see a bag, my memory of that bag will be stored as an exemplar, but the bag itself does not become part of my self grammar. This is an important distinction to make as the separation between an object and the representation of that object leaves room for inaccurate perception (either misperception or assumption) to influence the mental representation. In regards to phonetic variation, this means a listener can “store” a representation that was unintended by the speaker (for example, a speaker produces [õ] but a listener stores [on].) All other categories outlined below are, in some sense, contained within the self grammar. 3.2.3 Community Grammar The community grammar refers to an individual’s perceptions about how the community speaks. This consists of exemplars of specific speech events (for example, if the mayor or 25 principal said “bag”), in addition to perceptions of the community. I may have never heard someone speak German with an Australian accent, but I have a concept in my head of what that would sound like, based on generalizations about my past experiences. That concept will possibly influence my interactions with German-Australians (just as my interactions with members of that community will alter my perceptions of it). The community grammar is the collection of all exemplars of an individual’s perception of that community, coloured by their idea of “what it means to be X.” People define what they mean by community and community grammars in different ways. You can draw a line geographically (as in Labov’s Atlas of North American English, 2005), by age (Johnstone & Kiesling, 2008), by social class (Lawson 2011), identity (Labov, 1972), personality (Yu, 2015), race (Sharman & Sankaran 2011), or generation (Maegaard et al. 2013). All of these have been found to contain trends unique to each definition. But when you can find significance seemingly wherever you carve the lines, what are we really talking about when we talk about community? Can you have a community of two, as in twin languages (Bishop & Bishop 1998)? Do communities require some geographical link, or does the internet age make it possible for community grammars to emerge without physical proximity, or even spoken communication at all (Bourlai & Herring, 2014)? Johnston & Kiesling (2008 p. 7) point out what they refer to as the “intentional fallacy”, which is a researcher’s assumption that a hearer’s interpretation matches a speaker’s intention: That is, does part of the definition of community require an understanding between speakers and listeners about the members, or can communities emerge from mismatched interpretations and intentions? These are broad questions which cannot be answered in the current study, but they are worth keeping in mind as any line carved out and labelled “community” reflects the views of the researcher just as much, if not 26 more so, than reality. For the purposes of this thesis, “community grammar” refers to the specific community that a given speaker feels they are themselves a part of. So, in the broadest sense, the community under study in this thesis was “speakers from Alberta.” 3.2.4 Standard grammar The standard grammar refers simply to an individual’s perception of the “right” way to speak. Sometimes called the prescriptive rules of language (Pinker, 1999), it is the way that grade school teachers have always taught us we “should” use language, but also the way none of us consistently do. In short, it is an idealization, the Platonic form of language. The standard grammar includes broad grammatical rules like “never end a sentence with a preposition” as well as social rules like “Never swear in front of a lady” and “to speak properly you must say ‘eating’ and never ‘eatin’.’” This is the grammar imposed by the system (or rather, a speaker’s impression of the system), and it exists in day-to-day life after grade school in the form of orthography, formal communication, and “polite company.” 3.2.5 The hidden influence of the triadic grammar In this thesis, I assume that this information and these categories are somewhat hidden from each speaker—that their choices are informed by their three grammars but that they have no access to the information or categories within. The nature of how this information stays hidden I leave to future research, but it is necessary to hypothesize because of the inevitability of sound change: Were people truly able to control their access to the hidden influences on speech, one would expect to find instances where sound change is resisted completely, and that is simply not borne out by history (though there are many examples of individuals resisting sound change, for examples see Meyerhoff (2011), p.74, and Stanford (2008)). The reason for hypothesizing sound change as an interplay of three grammars is based on the differing sources of influence, as is 27 evident in the names given to them. One should not assume they are anything more than convenient descriptive labels applied to the observable results of cognitive and social influences, and not specific cognitive acts or categories themselves. Though this experiment seeks to affirm the existence of some interplay between these three influences on speech behavior, I leave it to future researchers to determine whether there should be more, fewer, or different categories to explain sound change. Now, I will briefly explore two other theories of sound change and how they are incorporated into my concept of the grammar trinity. 3.3 Identifying Mergers Once the existence of a merger is confirmed, the second step is determining what kind of merger it is. In determining the direction of the shift, I three possible labels from Wassink (2014). Mergers by approximation (Figure 3-1), mergers by expansion (Figure 3-2), and mergers by transfer (Figure 3-3). Figure 3-1: Merger by Approximation (two subtypes). (Taken from Wassink, 2014). Mergers by approximation take one of two forms; either both sounds find some “middle ground” between them and the original distinction is lost (left), or one sound moves nearer towards the other until it encroaches on the space of the first, at which point a new vowel is formed (right). 28 Mergers by expansion are those in which the boundaries around both sounds grow, until each encroaches on the others’ space. Figure 3-2: Merger by Expansion Wassink notes that this type of merger that is “indistinguishable from merger by approximation mid-course” (pp.7). That is, a merger by expansion looks like a merger by approximation when it is in progress, and can only be categorized after the process is complete. However, I believe this is only the case if one assumes sound change affects an entire population to equal degrees at each point in time. My theory of different individuals perpetuating and adopting changes at different rates means we should expect to find different subjects at different stages of merging [æ] and [ɛ], making it possible to distinguish between a merger by expansion and a merger by approximation for an individual listener/speaker, even if the change is not yet complete. It is also possible to identify different strategies being employed by individuals: That is, one speaker may approach the merge via expansion, and another via approximation. The final type, merger by transfer, is a unidirectional shift as well. It differs from the unidirectional subtype of merger by approximation in two ways: First, the shift is abrupt, with no intermediate forms; and second, the first sound (A) is entirely subsumed into the second (B). This is the classic, abrupt sound change of the Neogrammarians. Most cases of deletion could be 29 argued to be merger by transfer, one example is [k] becoming [tʃ] in Sanskrit (McMahon, 1999, p.48). Figure 3-3: Merger by Transfer 3.4 Pilot Study A pilot study was conducted to determine more clearly the nature of the [æ]/[ɛ] vowel shift and refine the methodological structure and research questions before committing to a large-scale experiment. The results indicated that the two speakers tested in this pilot were both shifting their vowels, but employed different strategies in rearranging their vowel spaces to accommodate a shift in the production of [æ] before [g]. 3.4.1 Participants Participants were one female and one male speaker, who were born and had lived virtually all of their lives in Alberta (Calgary and Edmonton area). Both participants were phonetically trained native English speakers. Both speakers were aware of the [æ]/[ɛ] merger and neutralization, but were unaware that this was the purpose of the current investigation before participating. 30 3.4.2 Stimuli Stimuli consisted of PowerPoint slides containing 100 pictures and 100 orthographic words (see Appendix A: Pilot Stimuli for a complete list of stimuli), to determine if “self” and “standard” grammars showed any difference in production output. Pictures and words were presented in blocks, with the pictures first and words second. Stimuli consisted of three types of words: Target words, which contained vowels before /g/, such as Maggie and legacy; control words, which contained vowels before /k/, such as trachea and wax; and filler words, such as baby, which contained no vowels of interest but which were included to prevent participants from become aware of the task’s purpose. 3.4.3 Procedure The research took the form of a participant-driven elicitation task. A PowerPoint presentation of both visual and orthographic tokens was created, and participants were asked to either name the object in the picture or read the word, while being recorded to a .wav format sounds file on a Shure SM48 microphone with pop-filter. The recordings were made with Adobe Audition 2.0 in mono at 44100Hz in a sound-attenuated booth at the Phonetics Lab of the University of Calgary. A researcher controlled the recording levels and asked participants to repeat tokens when necessary. Participants were allowed to take a break whenever they wanted and were provided with water. Participants were asked to repeat tokens if they were pronounced incorrectly or if there was clipping in the recording. They were also asked to repeat tokens in case a variable pronunciation could be elicited (for example [ɛgz]-[ɛgs]-[ɛks] variation in words spelled with ex like exit), indicating an incomplete merger within tokens. Finally, they were asked to repeat tokens if the participant’s production contained creaky voice, breaks, and list or 31 rising intonation (‘uptalk’), as these can cause errors with the automatic scripts used to analyze the recordings. I used Praat to splice the recordings into files containing a single response. When multiple repetitions of a single token were asked for, each repetition was spliced into its own file (for example craig.wav, craig2.wav). This means that the number of words of each type in each condition was not balanced, as some words had more responses than others. Each file was checked for problems, such as clipping, incomplete tokens, noise, or tracking errors. They were then annotated in Praat to include information about the start and end of the vowel portion of each word. A combination of a Praat script and a Perl script created by Dr. Stephen Winters then processed these files to pull the Hz measurement for F1, F2, and F3 from the 25%, 50%, and 75% mark from each vowel and output them to a tab-delimited table. This process made 9 measurements per token. The first two tokens from each speaker were measured by hand in Praat and compared to the script’s output. No major deviations were found and so the values presented by the script were used in further calculations. 32 3.4.4 Results of Pilot Comparison of æ and ɛ formants before k 3000 3000 2500 2500 2000 2000 Hz Hz Comparison of æ and ɛ formants before g 1500 1500 1000 1000 500 500 0 0 25% 50% 75% 25% Measured point (average) 50% 75% Measured Point (average) Figure 3-4: Comparison of vowels [æ] (light) and [ɛ] (dark) before [g] and [k] The tables in Figure 3-4 compare all formant measurements across participants for [æ] and [ɛ] before [g] (left) and [æ] and [ɛ] before [k] (right). These results were graphed to explore the degree of neutralization between the vowels, and the quality of the vowel’s change. The most obvious difference in quality was in the 2nd formant between both groups. Both vowels before [g] show a greater rise in F2, much more than the vowels before [k], though both [k] and [g] have the acoustic marker of the “velar pinch,” where F2 rises and F3 drops immediately before the stop closure. The most drastic of these changes is the rise in the 75% point of [æ]’s F2, between the voiced and voiceless stop compared to these stops in the before [k] context. In [æg] sequences, F2 at 75% is 2117Hz, which is 380Hx greater than 1737Hz, the same measurement in [æk] sequences. Compared to the change from [ɛg] to [ɛk] sequences (2243Hz - 1958Hz, 285Hz), the difference between the two differences is almost 100Hz. This is evidence that the direction of the merger is upwards in the vowel space, causing the [æ] vowel to get nearer to the same space as [ɛ]. The significance of this result was not possible to test, due to the low number 33 of participants and data points, but provides motiivation to continue the research in a larger-scale study. COMPARISON OF VOWELS BY MODALITY, ɛg 3000 3000 2500 2500 2000 2000 HZ HZ COMPARISON OF VOWELS BY MODALITY, æg 1500 1500 1000 1000 500 500 0 0 1 25% 2 50% 1 025% 0% 300% 75% AVERAGE FORMANT AT EACH POINT 3 75% AVERAGE FORMANT AT EACH POINT 250% 00% Figure 3-5: Productions by Modality: Pictorial (crosshatches), or Orthographic (shape) The two graphs above in Figure 3-5 compare the average Hz measurements of each formant based on the type of stimuli presented. Though there was not enough data for a statistical analysis, the graph shows that results of both modalities are nearly identical. While some speakers show greater care in production in reading tasks compared to picture lists (Meyerhoff, 2011:31), the two subjects of the pilot did not. This may be a flag that the merger in progress is a sociolinguistic indicator, rather than a marker (that is, below the level of awareness, rather than above it; see Meyerhoff 2011, p.23). However, this result may also be due to the subjects’ awareness of the task (that is, the modality change rather than the vowel shift) or the small sample size. Since there was no obvious difference between orthographic and pictorial presentation of stimuli on the formant values of the vowels, all further results presented from the pilot do not include this distinction. However, this modality split was retained in the main study, 34 and auditory stimuli were added on, as it was felt that a three-way division better represented input in everyday life from three differing sources (the triadic grammar), and the influence of the order of presentation on vowel production could be measured as well. Figure 3-6 combines both vowels and consonants, and illustrates more obviously the change of quality in [æ] when appearing before [g] and [k]. Figure 3-6: Combined Formants of Both Vowels and Contexts Figure 3-7 (below) compares productions between the female speaker (left) and the male speaker (right) across vowels before [g] (upper quadrant) and [k] (lower quadrant). Though the [_g] trajectories of both speakers are more similar than those before [k], the female speaker appears to be adopting this merger to a greater degree, especially in the F2 (which was shown above to be the most prominent quality of the merger.) The role of women as progenitors and 35 earl adopters of language change is well documented (for examples, see Meyerhoff 2011:136), so this is an expected result. Vowel comparison of æ/ɛ before g, Female Vowel comparison of æ/ɛ before g, Male 3000 3000 2500 2500 2000 Hz Hz 2000 1500 1500 1000 1000 500 500 0 0 Vowel comparison of æ/ɛ before k, Female Vowel comparison of æ/ɛ before k, Male 3000 3000 2500 2500 2000 Hz Hz 2000 1500 1500 1000 1000 500 500 0 0 Figure 3-7: Gender differences in production of [æ] (green) and [ɛ] (blue) before [g] and [k] 3.4.5 Pilot Conclusions Recent research in speech perception and production has moved beyond the simple sex dichotomy of male/female. For example, Pierrehumbert et al. (2004) found that gay men’s vowel spaces were wider than their straight counterparts, while lesbian and bisexual women fronted vowels less than straight women. Davies et al. (2006) found significant differences in the speech of transgender individuals compared to cisgender individuals. These differences in speech are not limited to gender and sexual orientation, but also in terms of group identity. Labov’s seminal research on Martha’s Vineyard (1963), half a century ago, concluded that young people’s speech 36 was determined in part by their feelings towards their hometown: Those who wanted to stay in Martha’s Vineyard after graduation had a stronger local dialect than those whose eyes and aspirations were set elsewhere. Conversely, Rosen & Scriver (2015) determined that Mormons in Alberta patterned their speech more similarly to American Mormons than their geographic neighbours, non-Mormon Canadians. These results show the speakers display an alignment and solidarity that had nothing to do with plans to move somewhere else, but rather with how they identified with a non-local group. These findings, along with the differences in gender found in the pilot, inspired me to expand the scope of the demographic data gathered. While for the most part this vowel merger acts as an indicator (a way of speaking that operates below the level of speaker consciousness), there is evidence it may be a marker (something speakers are aware of) (Meyerhoff 2011:23). Both speakers recorded were unaware of the extent of their neutralization, even after being questioned after recording. However, it is important to note that both participants were linguistics students late in the program, and as such were aware of the possibility of a neutralization, and had discussed local dialects extensively throughout the program, even if they were unsure of their own participation in it. Testing on truly naïve speakers is important for determining the extent and nature of any social recognition that this possible marker has, so I decided to test linguistically naïve speakers for the main study. 37 Methodology 4.1 Research Objectives My experimental goals and hypotheses for this research project take, broadly speaking, two forms: the empirical and the perceptual. That is, I seek to explain what is happening to [æ] and [ɛ] in a measurable way, and I seek to form an explanation for why these vowels may be undergoing these shifts. To that end, what follows are each of my hypotheses and a brief explanation of what can be expected from the experimental design: I. Albertan speakers are merging [æ] and [ɛ] before [g], producing a sound somewhere between the category boundaries for [æ] and [ɛ] in other contexts. If this is the case, I expect to find significant acoustic differences between the acoustic output of [æk] and [ɛk], but no significant differences between the acoustic outputs of [æg] and [ɛg]. As a consequence of this merger, I would also expect to find significant differences between [æk] and [æg], and [ɛk] and [ɛg]. II. This is a change-in-progress. Only some speakers are merging these vowels. As I hypothesize, based on my own informal observations and the results of the pilot study (where the female speaker was merging her vowels before [g] but the male speaker was not), that only some speakers in this area are shifting their vowels, I expect to find that only some speakers produce data that aligns with hypothesis (I). It is also expected that speakers may be undergoing the merger at varying rates, as in Bybee’s findings about lexical diffusion (2002). III. Certain personality types are more likely to innovate new changes (“inactive listeners”). Certain personality types are more likely to adopt changes made by others 38 (“super listeners”), and it is possible to measure these personality types’ effect on speaker production. IV. There exists some measurable influence of self, standard, and community grammars. This can be measured by examining the relationship between the influence of differing modalities on each speaker’s vowel productions and the personality characteristics of those speakers. In the previous chapter I outlined a continuum based on Ohala’s Active Listener hypothesis, in which I posit that some listeners-turned-speakers are more active than others. The possible existence of these so-called “Super Listeners” and their counterparts at the opposite end of the spectrum (the inactive listeners, or the innovators) can be confirmed experimentally in two ways: First, by measuring how closely a given listener-turned-speaker mimics the nonstandard speech of another; and, second, by finding correlative personality traits between subjects at similar stages of the merger in progress. The first measure is determined simply by introducing a target and measuring how closely each listener-turned-speaker mimics that target. The “targets” in this case refer to the words prompted by stimuli presented in different modalities and how those modalities align with the three grammars. Presenting subjects with pictorial stimuli seeks to engage their self grammar, free of outside influence (or as free as can be expected in any experimental setting). Asking subjects to read orthographic texts attempts to determine the extent to which they align with the standard grammar, as reading is an unnatural learned skill (compared to speaking), and the sound-letter correspondences are considered proper and correct speech. Presenting both words and nonwords was done for two reasons: First, Hay et al (2013) found that speakers tend to show greater distinctions when producing nonsense words compared to words in conditional mergers, 39 so they may provide finer grain data about the nature of this merge even among those whose productions of both [æ] and [ɛ] are identical. Second, presenting both words and nonwords tests the influence of the grammars even more directly, by forcing participants to generalize to a new production based on their existing exemplars (a person has probably said “bag” before, but has probably never said “gug” before). Finally, asking them to mimic nonstandard auditory speech seeks to quantify the degree of influence of the community grammar, as the voice presented acts as a representation of how “they” speak, where “they” represents the community speech that is sourced in the neither the self nor the standard grammars. By looking at correlations between personality traits and performance on this multifaceted task, I attempted to measure the relationship between the degree of influence of community grammars on a subject’s speech, and the extent to which they exhibited the merger in progress. I hypothesize that self-focused subjects (those that consider themselves introverted and leaders) are the innovators, as they are more likely to introduce variations in the speech signal, and less likely to reproduce the sounds of others. Similarly, other-focused subjects (those that consider themselves extroverted and followers) are more likely to reproduce changes that they have heard. 4.2 Participants Participants were 19 students at the University of Calgary between the ages of 19 and 26 years. They were compensated with research participation credit for the Linguistics 201 course at the University of Calgary. No participants reported significant speech or hearing disabilities. One participant’s data was removed due to recording error, leaving 7 male and 11 female participants. 40 4.3 Materials and Procedure Experimental materials consisted of a background questionnaire, calibration prompts, and three sets of prompts in different modalities. 4.3.1 Demographic and Personality Questionnaire Participants filled out a background questionnaire before any recording took place. This questionnaire consisted of two types of questions. The first type addressed basic demographic information: Age, gender, sexual orientation, highest level of education completed (and parents’ highest level of education completed), handedness, hometown, other places lived, mother tongue, and other languages spoken. The second set of questions asked participants to align themselves along the following “personality binaries”: Rural/Urban, Liberal/Conservative, Familyfocused/Friend-focused, Introverted/extroverted, Tech-savvy/tech-apathetic, Religious/Nonreligious, Leader/Follower, and Thinker/Feeler. The full questionnaire can be found in Appendix B: Subject Questionnaire. Interpreting social data is difficult, and other research (e.g. Alan et al. 2013, Dimov et al. 2013, Herrero 2008) has supported the validity of using results from tests designed for other spheres of study to determine the linguistic influence of broad personality traits. However, I used a short and focused personality questionnaire because no current existing psychometric tests could accurately answer the questions posed by my hypothesis while still allowing enough time for participants to complete the entire experiment without losing attention. Most tests of this type take upwards of an hour to complete, in addition to requiring too much of the researcher’s time to interpret and process responses to make it suitable for statistical analysis. Also, many personality tests use a Likert scale, but the personality questionnaire I designed forced 41 participants to choose between two extremes in order to exaggerate the underlying effects of broad personality types. 4.3.2 Calibration In the calibration portion of the experiment, participants produced 19 “baseline” vowels, providing recordings of all Canadian English vowels in [h_d] and [h_rd] contexts. The number is uneven because three instances of [ɑ] were collected; in “hawed”, “hod”, and “hard” – These vowels have already merged in most of Canada (Labov, 2005), and they were collected as separate tokens to identify nonlocal dialects. Since there is no word “hawrd” in English, the baseline vowels are uneven. All vowels were collected so that each participant’s vowel space could be fully defined and make the determination of possible vowel shifts within that space more accurate. Participants were presented with orthographic depictions of words in black text on a white computer screen, and were asked to repeat each word three times, with a pause in between each repetition. A researcher was present throughout the experiment and controlled the speed and presentation of the stimuli. If the recording was unclear (due to someone bumping the microphone or misspeaking on the participant’s part), participants were asked to repeat the word a maximum of two times before moving on to the next one. The wordlist used in collecting calibration vowels was taken from Peterson & Barney (1952) with the addition of the same vowels in the context of [h_rd]. The complete wordlist can be found in Appendix C: Calibration Vowel Word List. 4.3.3 Modality Stimuli The modality prompts were presented in three blocks based on the modality of presentation: Orthographic, Pictorial, and Auditory (explained in more detail below). In all three blocks, participants produced a single word or non-word. The order in which the three modalities 42 were presented was counterbalanced across participants: Three participants were presented with stimuli in the order of (A)uditory-(O)rthographic-(P)ictorial, three in APO, three in OAP, three in OPA, three in PAO, and three in POA. The full breakdown of all stimuli by word/nonword status and modality type can be found in Appendix D: Experimental Stimuli Word Lists. Each stimulus was intended to prompt a word/nonword that contained the sequences [æk], [ɛk], [æg], or [ɛg] in syllable-final position. The presentation of the prompts in the Orthographic block presentation was identical to the presentation of the prompts in the calibration block. Participants were presented with black written words on a white background on a computer screen. They were asked to read the word or non-word presented. There were 40 total prompts made up of 20 words and 20 non-words. The stimuli for the Pictorial block consisted of 40 slides to produce 40 recordings: 20 words and 20 non-words. The pictorial stimuli that prompted real word responses differed from the stimuli that prompted non-words. For words, participants were presented with a picture of an object (such as an egg) or a person and asked to identify the object or name the person. If participants didn’t know what an object or person was, they were given a specific clue (such as, “He was the president of the USA in the 1980s” for “Reagan,” or “She is going to have a baby because she is…” for “pregnant”). Only participants who were unsure or incorrect in identifying the images were given the clue, but all participants who received a clue received the same one. In cases where a participant did not know which target word to produce, after two prompts or hints the item was skipped and the experiment moved on to the next target. Because of this, the number of total productions per person varied. For non-word target prompts, participants were initially given a set of three tutorial slides that instructed them how to create the “words” from 43 the pictures presented. None of the tutorial slides contained words with the phoneme target sequences under study, [æg]/[ɛg] or [æk]/[ɛk]. Figure 4-1 shows one of the training slides. Figure 4-1: Training slide from the Pictorial block Participants were asked to take the onset (the “first sound”) from the first picture, and add it to the rhyme (the “rest of the word”) of the second. The training slides contained complex onsets in addition to the simple one displayed above, so participants were also tutored that, for example, shrimp and desk created shresk. While complex onsets are technically/phonetically more than one sound, saying “say the first sound of the first word and the rest of the second word” and giving examples with complex onsets produced the most consistent and correct responses from the (untrained) participants. As shown in Figure 4-2, the slides were identical to those presented in the training portion, except that they did not contain any orthographic “hints.” 44 Figure 4-2: Pictorial slide to prompt the nonword "bregg" All images used in the nonword pictorial slides were easily identifiable monosyllabic nouns, and all participants were able to identify every picture. No prompts from the researcher were necessary except for an occasional “try this again, please,” when the participant did not include the entire onset of the first image (for example, saying “beg” for the slide presented in 42). The Auditory block was the most complex set of stimuli, and contained over twice the number of tokens (90) as the other blocks. Participants heard a sound clip, produced by a trained phonetician, and were asked to repeat what they heard. The sounds presented took one of the following forms: 1. A “basic” word with the appropriate nonmerged vowel, e.g. egg [ɛg], bag [bæg]; 2. A “switched” word with the opposite nonmerged vowel than expected, e.g. freckle pronounced [fɹækl] instead of [fɹɛkl] or dragon pronounced [dɹɛgən] instead of [dɹægən]; 45 3. A word with a “nonvowel,” where the word/nonword was produced with the sound [œ], which is not found in Canadian English, e.g. [bœg] for bag or [gœk] for g_k (neither [gæk] nor [gɛk] are proper English words). If participants mispronounced a word, or pronounced a nonword in a manner different than expected, they were asked “Once more, please,” for a maximum of two times before moving onto the next target. The speaker who produced the auditory stimuli also produced the words in the calibration block of recording so that her baseline vowels could be compared to that of participants. This factor was held consistent with the intention of having a baseline by which to measure the degree to which her voice in the auditory portion influenced the speakers’ production. 4.3.4 Procedure Participants completed the personality questionnaire first, before entering the recording booth. Their responses were recorded to .wav files via a Shure SM48 microphone with a popfilter. The recordings were made with Adobe Audition 2.0 in mono at 44100Hz in a soundattenuated booth at the Phonetics Lab of the University of Calgary. A researcher controlled the recording levels and asked participants to repeat tokens when necessary. Participants were allowed to take a break whenever they wanted and were provided with water. Participants were asked to repeat tokens if they were pronounced incorrectly, if they created clipping in the recording, or if the participant’s production contained creaky voice, breaks, or list or rising intonation (‘uptalk’), as these can cause errors with the automatic scripts used to process the stimuli. The recordings were spliced into files containing responses to a single stimulus in Praat, then further into files containing a single word. Each file was annotated to include information 46 about the start and end of the vowel portion of each word, and during this annotation process each file was checked for problems (such as clipping, incomplete tokens, noise, or tracking errors). Figure 4-3 below is a labeled example of the values analysed. æ Figure 4-3: Labelled Spectrogram as prepared for each token by Praat and Perl Scripts. A combination of a Praat script and a Perl script created by Dr. Stephen Winters then processed these files to pull the Hz measurements for F1, F2, and F3 from the 25%, 50%, and 75% mark from each vowel and output them to a tab-delimited table (these are represented by the blue dots in Figure 4-3). This process made 9 measurements per token. Since this was the same procedure used to process the pilot data, and no deviations were found there between the programmed responses and the hand-measured ones, the values presented by the script were used in all statistical analyses. 47 4.4 Analysis 4.4.1 Finding Mergers and Splitters The first step in analysing the data was to determine whether subjects were undergoing a vowel shift at all. Figure 4-4 shows a graph of all vowels and environments under study, and is formatted as the other vowel spaces in this thesis, with F1 along the y-axis and F2 along the xaxis, and the values plotted in reverse order to more closely mirror the traditional vowel charts used throughout this thesis. 2200 2100 2000 F2 1800 1900 1700 1600 1500 1400 500 550 600 ɛg Average of all Calibration vowels ɛ 650 F1 700 æg 750 ɛk 800 æ 850 æk 900 950 Figure 4-4: Comparison of Stimuli Vowels to Calibration Vowels, All Speakers This illustrates a rough idea of the vowel space of the speakers tested, and confirms that in all contexts, [æ] is produced lower and farther back in the vowel space than [ɛ]. While this result is expected and confirms there is nothing horribly awry with the testing materials, it doesn’t tell anything about the way the individual speakers produce these vowels, nor does it allow us to determine with confidence the nature of this vowel shift. 48 Because a change in progress may not affect the entire population under study, it was necessary to examine each subject’s responses individually. Figure 4-5 (pp. 45) shows a comparison of all subjects’ average formant values of [æ] and [ɛ], both before [g] and in the “neutral” calibration environments (before [d] and before [ɹd]): Figure 4-5 is included merely as another way to visualize the data presented in Figure 4-4, with the results separated by individual participants rather than in aggregate. The x-axis lists the subject number, and the y-axis represents the normalized z-scores of each production (the normalization process will be explained later). Productions of both vowels before [g], for some speakers, appear to be more similar to each other than either the [æ] or the [ɛ] in the calibration block. However, this appearance must be justified statistically, and a uniform way of distinguishing those who are merging these vowels from those who are not must be determined. 49 Production distance of Average F1 and F2 from Calibrated æ 3 2.5 2 1.5 1 0.5 0 101 102 103 104 105 106 107 108 109 æd/ærd 110 æg 111 ɛd/ɛrd 112 113 114 115 116 117 118 119 118 119 ɛg Production distance of Average F1 and F2 from Calibrated ɛ 2.5 2 1.5 1 0.5 0 101 102 103 104 105 106 107 108 109 æd/ærd 110 æg 111 ɛd/ɛrd 112 113 114 115 116 117 ɛg Figure 4-5: Production distances between calibration vowels and those before [g] 50 When determining if two vowels are merging, it is appropriate to compare an individual’s vowel in one context with that same vowel in another context, as the individual’s vocal tract serves as the reference point—it does not change across contexts. However, to obtain a meaningful point of comparison across participants, some manipulation of the data is necessary, as one person’s vocal folds and physiology differ in size, pitch, frequency, etc. from all others. To this end, each participants’ vowels were normalized. Following the methods outlined in Lobanov (1971) and Cheswick & Miller (2005), a z-score was calculated for each vowel production by each participant, and a measure of Euclidean distance taken for each token compared to the sounds produced in the calibration step. The mean and standard deviation of F1, F2, and F3 measurements from each participant’s calibration vowels served as a base for the normalization process. Z-scores were obtained using the formula z = x – μ / σ, where x = the average F1 or F2 for each production. A Euclidean distance measure was then calculated for each production by each participant in a two-dimensional space: The first dimension was F1 and the second dimension was F2. For each token, the two points measured represent firstly a participant’s normalized sample production (s), and secondly the mean of that participant’s calibration vowels (avg). Therefore, the Euclidean distance for each measurement was deermined with the following equation: ((F1(s) – F1(avg))2 + (F2(s) – F2(avg))2)(½). I compared the acoustic properties of each subject’s productions of [æ] and [ɛ] before [g] to all the [æ] and [ɛ] vowels produced in the calibration block of recording (before [ɹd] and before [d]). I then repeated this comparison for each subjects productions of [æ] and [ɛ] before [k]. A series of paired sample t-tests was run to compare each subject’s F2 at the 50% mark in the conditions of vowel ([æ] and [ɛ]) and environment (before [g] or before [k]). That is, for each subject, I compared their productions of [æg] with their productions of [ɛg], and their 51 productions of [æk] with their production of [ɛk]. All participants maintained a significant (p ≤ 0.05) distinction in F2 between [æk] and [ɛk]. Participants who showed no significant difference in F2 (p > 0.05) between [æ] and [ɛ] before [g] were labelled “Mergers” (denoted by “M” in later graphs), while those who did show a significant difference between these vowels in this environment were labelled “Splitters” (denoted by “S” in later graphs). The 50% point of F2 was chosen as the data point of comparison to reduce levelling effects caused by the other two time points: The 25% point was likely to contain coarticulatory influences of the differing onsets, and the 75% point would likely be influenced by following [g]/[k], which causes the distinctive “velar pinch” towards the end of the vowel, as described in chapter 3 (Figure 4-6). Figure 4-6: A velar pinch. Image reproduced from Baker et al. (2007), green circle added by me. Since this procedure classified half of the subjects tested (9/18) as Mergers, it seemed possible that this classification scheme might be dividing up subjects by chance, since the expected result of taking the mean in the first place would place half of the subjects “above” the 52 mean and half “below.” To verify the validity of this classification, two further t-tests were run for each subject, comparing their production of [æ] with their production of [ɛ] in the calibration block of testing, which contained vowels in the environments [h_d] and [h_rd]. All subjects showed significant (p < 0.05) differences between the F2 at the 50% mark of [æ] and [ɛ] in the calibration block. That is, Splitters maintained a vowel distinction in both calibration and testing, while Mergers had a distinction in calibration that was lost during testing. Figure 4-7 shows the full vowel trajectories of [æ] and [ɛ] for a Merger and a Splitter. These two particular subjects were chosen after the t-tests determined their categorisation for having the clearest graphs. Comparison of a Splitter's [æ] and [ɛ] Vowels [_g] 3500 3000 2500 ɛg F1 2000 Hz ɛg F2 ɛg F3 1500 æg f1 1000 æg f2 æg f3 500 0 10 25 40 55 Relative Time (%) 53 70 85 Comparison of a Merger's [æ] and [ɛ] Vowels [_g] 3000 2500 ɛg f1 Hz 2000 ɛg f2 1500 ɛg f3 1000 æg F1 æg F2 500 æg F3 0 10 25 40 55 70 85 Relative Time (%) Figure 4-7: A Merger and a Splitter's formants of both vowels before [g] Frequency (Hz) is labeled on the y-axis and the relative time on the x-axis in Figure 4-7. Relative rather than absolute time is used because the frequency was measured at the 25%, 50%, and 75% mark of each token, regardless of how long that token was. Though the general trajectory of vowels before [g] remains similar across participants, the Mergers’ frequency values are almost identical regardless of the vowel being produced. 4.4.2 Correlations With Personality Characteristics The participants’ responses to the questionnaire were used to create a correlation matrix. Most responses were already binary by design, but the others were converted to binary responses as follows: Orientation was recorded as straight (0) or “not straight” (1), with “not straight” combining responses of “gay” or “bisexual”; Education was coded as “some university” (0) and “More than university” (1), as all participants, being pulled from the Intro to Linguistics student pool, had completed high school; Parents’ level of education was coded as “High school or below” (0) or “Some university and above” (1); Other native language was coded as either 54 “English only” (0) or “English and other languages” (1); Other places lived was recorded as “Alberta” (0) for those who had lived only in Alberta their entire lives, and “Outside of Alberta” (1), for those who had spent at least one year living outside of Alberta. Gender was recorded as a binary variable, as all participants responded either “male” or “female.” The correlation matrix can be found in Figure 4-8. Correlations between variables that did not reach significance (p ≥ 0.05) are crossed out, and the lower quadrant displays the degree of correlation, with red indicating negative correlation and blue indicating positive correlation. In all cases except those specified, the first response listed was coded as (0). For example, in the response “Leader/Follower,” “leader” was treated as (0) and “follower” was treated as (1). Only those correlations which reached significance will be discussed further. Handedness and the number of other places lived were negatively correlated to merging, while rurality had a positive correlation. That is, those who had spent at least a year outside of Alberta were more likely to merge, as were those who identified as “rural” on the Rural/Urban identity binary. Left-handed participants were also more likely to merge according to the tests. However, since only one respondent was left-handed, this correlation will not be explored further. 55 Figure 4-8: Correlation Matrix for Personality data (“Merger” highlighted) 4.4.3 ANOVAs 4.4.3.1 Introduction to ANOVAs The last section served to identify Mergers and non-Mergers based on how much each individual’s productions varied in the experimental contexts compared to their productions in the “neutral” contexts obtained in calibration. This section looks at all participants’ results and Chapter 6 interprets these results to outline a “hierarchy of influence” that different modalities have on production and repetition. A series of paired Five-way Analyses of Variance (ANOVAs) were run. The dependant measure in the first ANOVA of each pair was the normalized Euclidean distance of each vowel token from the combined F1 and F2 of the [æ] vowel obtained in the calibration portion of recording. The dependant measure in the second ANOVA of each pair was the normalized 56 Euclidean distance of each vowel token from the combined F1 and F2 of the calibration [ɛ] vowel. The independent measures were the same in each ANOVA and consisted of the following: 1. Target: Whether the participant’s target was [æ] or [ɛ]. In the case of words, this was determined by how the word is normally produced (by non-merging speakers). For nonwords this was determined by the letter presented (in the orthography block) or the sound usually produced when naming the second picture (in the pictorial block). 2. Context: Whether the target sound was produced before the voiceless velar stop [k] or the voiced velar stop [g]. 3. Word: Whether the target sound was produced within the context of a word (e.g. bag) or a nonword (e.g. kag). 4. Stimtype: The modality used to elicit the production. For these tests, this is either Pictorial (P), Orthographic (O), or Auditory (A). As the auditory block of the experiment contained more than two possible auditory targets (including [œ]), all auditory stimuli whose target was not [æ] or [ɛ] was removed from this series of tests, leaving 42 of the original 90 auditory tokens to be used in this analysis. Another ANOVA, outlined in the next section, examined the auditory stimuli alone. 5. Merge: Whether the speaker was a Merger or a Splitter, based on the criteria outlined in 4.3.1. 6. Order: The order in which the different stimulus types were presented. These were balanced across subjects: Auditory-Orthographic-Pictorial (AOP), Auditory-PictorialOrthographic (APO), OPA, OAP, PAO, and POA. 57 Merge and Order were between-subjects factors; the others were within-subject factors. Order was removed from the following ANOVAs after an initial test showed multiple 2-, 3-, and 4-way interactions with Order. An ANOVA testing this factor was run separately, and the results of this test can be found in section 4.3.3.4. Figure 4-9 shows a normalized comparison of all vowels produced by all speakers, in relation to their calibration vowels. It can be considered a companion chart to Figure 4-4, as it displays roughly the same information in a new, appropriate way. F1 and F2 have been converted to normalized z-scores, and the 0-0 point is the average F1 and F2 score of all speakers (and therefore, negative values are possible). F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.2 eg 0 0.2 0.4 Calibration e aeg ek 0.8 F1 0.6 1 1.2 1.4 aek 1.6 Calibration ae 1.8 Figure 4-9: Normalized Comparison of Stimuli Vowels to Calibration Vowels. Graphs of all the significant results and interactions can be found in Appendix E: Plots, though graphs of some of the more illuminating results will be reproduced within the following sections. 58 Non-significant results will not be reported unless relevant to understanding the significant results. 4.4.3.2 ANOVAs Without [œ] Auditory Stimuli. Two 5-way ANOVAs were run to test the effects of the above outlined conditions (except Order) on all speakers’ [æ] and [ɛ] vowels. Tables 1 and 2 list the results of these tests, while the rest of this section reports on the post-hoc testing. Table 4-1: Significant Effects and Interactions in ANOVA on distance from calibration [æ] Factor Degrees of Freedom Residual DoF F-Value Pr(>F) Target 1 2079 364 < 0.01 Context 1 2079 889 < 0.01 Word 1 2079 36 < 0.01 Merge 1 2079 259 < 0.01 Target:Context 1 2079 17 < 0.01 Target:Merge 1 2079 35 < 0.01 Context:Merge 1 2079 27 < 0.01 Stimtype:Word 2 2079 10 < 0.01 Target:Context:StimType 2 2079 7 < 0.01 Target:Word:StimType 2 2079 7 < 0.01 Target:Context:Merge 1 2079 25 < 0.01 59 Table 4-2: Significant effects and interactions in ANOVA on distance from calibration [ɛ] Factor Degrees of Freedom Residual DoF F-Value Pr(>F) Target 1 2079 79 < 0.01 Context 1 2079 548 < 0.01 Word 1 2079 46 < 0.01 StimType 2 2079 11 < 0.01 Target:Context 1 2079 240 < 0.01 Target:Merge 1 2079 19 < 0.01 Target:Word 1 2079 33 < 0.01 Context:Word 1 2079 9 < 0.01 Target:Context:StimType 2 2079 3 < 0.01 Target:Word:StimType 2 2079 6 < 0.01 Target:Context:Merge 1 2079 7 0.01 Target:Context:Word 1 2079 7 0.01 Context:Word:Stimtype 2 2079 9 0.03 60 TukeyHSD post-hoc tests were used to determine the direction of significance in all cases. The results follow, with main effects first, then two-way interactions, then three-way interactions. While generally main effects are passed over when significant interactions are present, I choose to present a brief overview of main effects first. The reason I approach the data this way is that the interactions are complex, and exploring the simpler main effects and two-way interactions first allows me to more clearly and concisely explain the three-way interactions. Post Hoc results for all ANOVAs can be found in Appendix F: Post Hoc Results. Both ANOVAs showed significant main effects of Target, Context, and Word. The ANOVA run using the [æ] vowel additionally showed a significant main effect of Merge, while the ANOVA run using the [ɛ] vowel showed an additional significant main effect of StimType. F2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 ɛ 0.4 Calibration e 0.8 æ 1 F1 0.6 1.2 1.4 1.6 Calibration ae 1.8 Figure 4-10: Main Effect of: Vowel Target In general, all participants’ productions of [ɛ] in recording were closer to their productions of [ɛ] obtained in calibration than the [æ] vowels elicited in the experiment proper, as expected. However, participants’ productions of [æ] were significantly closer to their [ɛ] 61 calibration vowel than their [æ] calibration vowel (see Figure 4-10). This is likely due to the effects of the merging subjects, as the interactions show below. The main effect of context (Figure 4-11) shows that all participants’ productions of both [æ] and [ɛ] are closer to the [ɛ] produced in calibration when the vowel appears before [g], compared to the vowels produced before [k]. 0.8 0.7 0.6 0.5 F2 0.4 0.3 0.2 0.1 0 0 Before [g] 0.2 0.4 Calibration ɛ 0.8 F1 0.6 1 Before [k] 1.2 1.4 Calibration æ 1.6 1.8 Figure 4-11: Main Effect of: Context In regards to the main effect of words (whether the sound produced was a word or nonword of English), both words and nonwords are produced closer to [ɛ] than [æ] (Figure 4-12). Words and Nonwords do not vary significantly from each other in regards to their distance from [ɛ]. 62 F2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Calibration ɛ 0.2 Words 0.4 0.8 1 F1 0.6 Nonwords 1.2 1.4 Calibration æ 1.6 1.8 Figure 4-12: Main Effect of: Word Post hoc testing on the main effect of Merge determined that all vowels produced by Mergers were farther away from [æ] than the vowels produced by Splitters (Figure 4-13). This is expected, due to the way Mergers were determined, by examining a similarity in both vowels before [g]. F2 0.6 0.5 0.4 0.3 0.2 0.1 Mergers Splitters Calibration æ 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 F1 0.7 Figure 4-13: Main Effect of: Merge (only significant in regards to [æ]) 63 The main effect of StimType was only significant with respect to the [ɛ] calibration vowel. (Figure 4-14). The productions in response to the orthographic stimuli were significantly farther away from the calibration [ɛ] than responses to the two other stimulus modalities. 0.8 0.6 F2 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 F1 Orthographic Stimuli Pictorial Stimuli Calibration [ɛ] 0.6 0.7 Auditory Stimuli 0.8 0.9 Figure 4-14: Main Effect of: Stimtype (only significant in regards to [ɛ]) Significant two-way interactions were found for both vowels between Target and Context and Target and Merge. Additionally, the ANOVA run using the [æ] vowel found significant twoway interactions between Context and Merge and Stimtype and Word. The tests run with [ɛ] found additional significant interactions between Target and Word and Context and Word. The interaction between Target and Context shows that productions of both vowels when produced before [g] are more similar to participants’ calibrated [ɛ] than their production of [æ] in the calibration block (Figure 4-14). Participants produced [æg] closer to calibration [ɛ] when before [g], compared to before [k]. 64 F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.2 ɛg 0 0.2 0.4 æg 0.6 ɛk 0.8 F1 Calibration ɛ 1 1.2 1.4 æk 1.6 Calibration æ 1.8 Figure 4-15: Two-Way Interaction between Target and Context Post hoc testing of the interaction between Target and Merge continue the trend of [æ] productions generally being closer to the [ɛ] produced in calibration (figure 5-12). Splitters’ production of [æ] is closer to their calibration [æ] vowels than Mergers, which is again expected due to the way this division of subjects was calculated. 0.8 0.7 0.6 0.5 F2 0.4 0.3 0.2 0.1 0 0 Splitters' [ɛ] Mergers' [ɛ] 0.2 0.4 Calibration [ɛ] 0.8 Mergers' [æ] Splitters' [æ] 1 1.2 1.4 Calibration [æ] 1.6 1.8 Figure 4-16: Two-Way Interaction between Target and Merge 65 F1 0.6 In examining the significant interaction of Context and Merge, all vowels produced before [g] were closer to the calibration [ɛ] (further from calibration [æ]) for both Mergers and Splitters. However, it is important to note that in both contexts, Mergers’ and Splitters’ vowels were significantly different from each other. F2 0.9 0.8 0.7 Mergers' productions before [g] 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 Splitters' productions before [g] 0.4 0.8 Splitters' productions Mergers' productions before [k] 1 before [k] 1.2 F1 0.6 1.4 1.6 Calibration ae 1.8 Figure 4-17: Two-Way Interactions between Context and Merge The only significant difference revealed by post hoc testing of the significant interaction between stimulus type and word is that, in the orthographic block, both vowels produced in words were closer to the calibration vowel [æ] than those vowels produced in nonwords (Figure 4-18). This interaction was not significant when examining the distance from the [ɛ] vowel. 66 F2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 Calibration [ɛ] Auditory Nonwords 0.4 Orthographic Words Pictorial Words Auditory Words Pictorial Nonwords Orthographic Nonwords 0.6 F1 0.8 1 1.2 1.4 Calibration ae 1.6 1.8 Figure 4-18: Two-Way Interaction between Stimulus Type and Word The last two two-way interactions are between Target and Word, and Context and Word. Vowels produced within words were closer to the calibration targets than vowels produced within nonwords (Figure 4-19). For both words and nonwords, [ɛ] was produced significantly closer to calibration [ɛ] than [æ] was produced to calibration [æ]. 67 F2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Words with [ɛ] Nonwords with [ɛ] Calibration [ɛ] 0.2 0.4 0.6 Words with [æ] F1 0.8 Nonwords with [æ] 1 1.2 1.4 1.6 Calibration [æ] 1.8 Figure 4-19: Two-Way Interaction between Target and Word Post hoc testing of the interaction between Context and Word (Figure 4-20) determined that, regardless of vowel, words and nonwords that ended in [g] were produced significantly closer to calibration [ɛ] than words and nonwords that ended in [k]. Nonwords were produced significantly closer to [ɛ] than all other Word/Context pairs. 0.7 0.6 0.5 F2 0.4 0.3 0.2 0.1 0 0 Words that end in g Nonwords that end in g Calibration e 0.2 0.4 0.6 0.8 1 Nonwords that end in k 1.2 Words that end in k 1.4 Calibration ae 1.6 1.8 Figure 4-20: Two-Way Interactions between Context and Word 68 F1 0.8 Significant three-way interactions for both tests were: Target, Context, and StimType; Target, Word, and StimType; and Target, Context, and Merge. The ANOVA run using the [ɛ] vowel found additional three-way interactions between Target, Context, and Word; and Context, Word, and StimType. In examining the three-way interactions, post hoc testing was only done on those interactions that shared 2 of the 3 factors being tested. The interactions between Stimtype, Target, and Context showed a few general trends, with some deviation (Figure 4-21). F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.5 Orthographic [ɛg] Auditory [ɛg] 0 Pictorial [ɛg] Auditory [æg] Calibration ɛ 0.5 Pictorial [æg] Orthographic [æg] F1 Auditory [ɛk] Orthographic [ɛk] 1 Pictorial [ɛk] Orthographic [æk] 1.5 Auditory [æk] Pictorial [æk] Calibration æ 2 Figure 4-21: Three-Way Interaction between Stimtype, Target, and Context. There was a general trend of the same vowel and context being produced in a cluster among all three modalities (Auditory, Pictorial, and Orthographic), with no significant difference in distance from calibration between differing stimulus types of the same target+context cluster, with the following exceptions: First, [æk] target+contexts, produced after hearing the auditory stimuli, were closer to the [ɛ] calibration vowel than those produced after seeing the pictorial 69 stimuli. Second, sounds produced in the [ɛk] target+context pair were significantly closer to the calibration [ɛ] when participants were responding after hearing the auditory stimuli, compared to when they were reading the orthographic stimuli. Another notable difference becomes apparent when graphing the z-scores of all three stimulus types and two targets in the [g] context with regards to the calibration [ɛ] (see Figure 4-22). While there is no significant difference between [ɛg] productions in any of the stimulus types, all three are significantly farther away from the calibration [ɛ] when compared to the same stimulus types in the [æ] context (with the exception of the orthographic [æg]). When these are charted, we see that the three [ɛg] productions are all further from calibration [ɛ] in the opposite direction (higher and fronter) than we find [æ] in the vowel space. 0.9 0.85 0.8 F2 0.7 0.75 0.65 0.6 0.55 0.5 -0.2 -0.1 Auditory_eg 0 Orthographic_eg 0.1 0.2 0.3 Auditory_aeg Calibration e Pictorial_aeg Orthographic_aeg F1 Pictorial_eg 0.4 0.5 0.6 0.7 Figure 4-22: Three-way Interaction between StimType, Target, and Context with regards to Calibration [ɛ] Both ANOVAs showed significant three-way interactions between StimType, Target, and Word (Figure 4-23). For orthographic stimuli with an [æ] target, words were produced 70 significantly closer to the calibration [æ] than nonwords were. There were no significant differences between this pair with respect to the [ɛ] calibration vowel. 0.7 0.6 F2 0.4 0.5 0.3 0.2 0.2 Orthographic W [ɛ] Orthographic NW [ɛ] Calibration ɛ 0.4 0.6 Auditory NonWord [æ] Pictorial Word [æ] Pictorial Nonword [æ] 0 0 Auditory W [ɛ] Auditory NW [ɛe] Pictorial W [ɛ] Pictorial NW [ɛ] 0.1 Orthographic Word [æ] Auditory Word [æ] Orthographic Nonword [æ] 0.8 F1 0.8 1 1.2 1.4 1.6 1.8 Calibration ae Figure 4-23: Three-Way Interaction between Stimtype, Target, and Word. This pattern holds true for pictorial stimuli, but no significant interaction with target was found for the pictorial stimuli. Nonwords with an [æ] target were produced significantly closer to the [ɛ] calibration vowel in the auditory block, compared to both other stimulus types, but the auditory and orthographic stimulus types were the only two [æ] nonwords that were significantly different in distance from each other. For the interaction of Target, Context, and Merge (Figure 4-24), in all targets and contexts, with the exception of [æk], Mergers’ and Splitters’ vowels are significantly different from each other. As Figure 4-23 shows, the greatest distance is between Mergers’ and Splitters’ [æg] productions, which fall on either side of the [ɛ] found in calibration. 71 F2 0.9 0.8 ɛgM 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.2 ɛgS aegM 0.3 Calibration ɛ ægS 0.8 ɛkS F1 ɛkM aekM 1.3 aekS 1.8 Calibration ae 2.3 Figure 4-24: Three-way Interactions between Target, Context, and Merge In the interaction between StimType, Context, and Word, all productions from all stimuli prompts before [g] were produced significantly closer to the [ɛ] calibration vowel than those produced before [k] (Figure 4-25). F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Auditory Words [_g] Orthographic Words [_g] Pictorial Words [_g] Auditory NonWords [_g] 0.2 0.4 Calibration ɛ Pictorial Nonwords [_g] 0.6 F1 0.8 Auditory Nonwords 1 [_k] Orthographic Words Pictorial Nonwords [_k] 1.2 [_k] Pictorial Words [_k] 1.4 Auditory Words [_k] Orthographic Nonwords [_k] 1.6 1.8 Calibration ae Figure 4-25: Three-way Interactions between Stimtype, Context, and Word. 72 Pictorial nonwords were produced significantly closer to calibration [æ] than pictorial words when produced before [k]. Similar results were found in the three-way interaction between Target, Context, and Word (Figure 4-26). 0.9 0.8 0.7 0.6 0.5 F2 0.4 0.3 0.2 0.1 0 -0.1 -0.5 Word [ɛg] 0 Nonwords [ɛg] Nonwords [ɛk] Words [æg] 0.5 Calibration ɛ F1 Nonwords [æg] 1 Words [ɛk] Words [æk] 1.5 Nonwords [æk] Calibration æ 2 Figure 4-26: Three-way Interactions between Target, Context, and Word. Generally, targets before [k] are produced significantly farther from calibration [ɛ] than those same targets before [g]. In sounds produced in the same target+context combination, for both [ɛk] and [ɛg], words were produced significantly farther from calibration [ɛ] than nonwords. 4.4.3.3 ANOVAs Run On Auditory Stimuli. As mentioned in Chapter 5, statistical analysis revealed a confound in the auditory block of testing after the experiment had been run. Participants were asked to repeat productions of one 73 of three vowels: [æ], [ɛ], or [œ]. The introduction of this third vowel, not found naturally in Canadian English, was included with the intention of giving subjects a somewhat neutral sound, so that I could test which vowel they “chose” as a replacement for it. However, this third option opened up many more perceptual possibilities, especially in regards to the influence of word formation. These 8 possibilities are listed in Table 4-3. Table 4-3: Perceptual Possibilities for Auditory Stimuli Auditory Possible Word Formation Name Label Prompt Target(s) Options [bæg] 1: [æ] 1: [æ] “bag” [æ] [æ] [bɛg] 1: [ɛ] 1: [ɛ] “beg” [ɛ] [ɛ] [blɛk] 2: [æ], [ɛ] 1: [æ] “black” Switched [ɛ] Ɛ [kæg] 2: [æ], [ɛ] 1: [ɛ] “keg” Switched [æ] Æ [dœg] 3: [æ], [ɛ], [œ] 0 Neutral _ vowel [blœk] 3: [æ], [ɛ], [œ] 1: [æ] “black” Neutral [æ] _æ [œg] 3: [æ], [ɛ], [œ] 1: [ɛg] “egg” Neutral [ɛ] _ɛ [bœg] 3: [æ], [ɛ], [œ] 2: [æ] “bag” Both _b [ɛ] “beg” In regards to possible targets, it was felt that the non-English vowel would only become available as a target immediately after it was presented in the experiment, and would not exist among possible choices except as a short-term mimicry. Since this possible confound was noticed only after testing was complete, the experimental materials are not balanced for all possibilities. The 90 stimuli presented in the Auditory Block have the following distribution: 21 74 [æ], 21 [ɛ], 18 Neutral [œ], 11 Switched [ɛ], 8 Both, 6 Switched [æ], 3 Neutral [æ], and 2 Neutral [ɛ]. Thus, a pair of two-way repeated measures ANOVAs were run, using the same factors as those described in 5.1.3.2, with the following changes: First, a third level was added to the “target” factor, making the possible options [æ], [ɛ], and [œ]. Second, the “StimType” factor was removed (as all tokens were obtained from the auditory block), and the “Label” factor was added (with the eight different levels from the table above). As described above, these labels referred to the perceptual possibilities and possible word formations for each token. The first ANOVA was run to compare the influence of the factors with regards to the normalized Euclidean distance from the [æ] obtained in calibration. The second ANOVA used the distance from [ɛ]. Tables 4-4 and 4-5 report the significant main effects and interactions of these tests for [æ] and [ɛ], respectively. Table 4-4: Effects of experimental factors on participants' [æ] vowel: Significant results. Factor Degrees of Freedom Residual DoF F-Value Pr(>F) Target 2 1528 288 < 0.01 Context 1 1528 336 < 0.01 Merge 1 1528 138 < 0.01 Label 5 1528 12 < 0.01 Target:Context 2 1528 44 < 0.01 Context:Word 1 1528 13 < 0.01 75 Target:Merge 2 1528 10 < 0.01 Context:Merge 1 1528 19 < 0.01 Context:Label 4 1528 6 < 0.01 Table 4-5: Effects of experimental factors on participants' [ɛ] vowel: Significant results. Factor Degrees of Freedom Residual DoF F-Value Pr(>F) Target 2 1528 32 < 0.01 Context 1 1528 112 < 0.01 Word 1 1528 11 < 0.01 Label 5 1528 8 < 0.01 Target:Context 2 1528 156 < 0.01 Context:Label 4 1528 11 < 0.01 Generally, the results follow the same trends as those found when other stimuli were tested: vowels before [k] are produced significantly closer to [æ] than vowels produced before [g]; [æk] was significantly closer to calibration [æ] than [æg]. Mergers produce all vowels generally closer to [ɛ] than Splitters do; and vowels produced in words are closer to their calibration targets than vowels in nonwords, indicating the influence of the standard grammar in orthography encouraging production targets to remain closer to the “standard” vowel production. With respect to results specific to this test, the three-level target factor, and the additional “Label” factor, showed the influence of mimicry and the “neutral” vowel on participant production. 76 The neutral vowel _ was produced significantly farther from both calibration vowels than either [æ] or [ɛ]. Switched vowels of both types (Æ and Ɛ) were produced closer to calibration [æ], though there was no significant difference in the distance between _æ and _ɛ. Æ vowels were produced closer to calibration targets than any neutral vowel with possible word formation (_æ, _ɛ, or _b). However, they were further from calibration targets than Ɛ vowels were. For example, [mɛgpaɪ] was produced with a vowel closer to [æ] than [mœgpaɪ], and [mægə] was closer to [æ] than [mœgæ]. In general, whether a word formation was possible had less influence on the vowel produced than the vowel target. Specifically, in regards to the tests using the calibration [ɛ], words were produced higher and fronter than nonwords. To put this in other terms, the influences of the community and self grammars are stronger than the influence of the standard grammar. 4.4.3.4 Influence of Order of Presentation. Two initial repeated measures ANOVAs, testing the effects of all factors listed in the last section (Target, Context, Word, StimType, Merge, and Order) on Euclidean distance from calibration [æ] and calibration [ɛ] showed significant main effects and 2-, 3-, and 4-way interactions of Order for both dependant measures. Thus, a simplified pair of two-way ANOVAs were run to more closely examine the influence of the between-subjects factors, Order and Merge, on participant production. It was felt that by determining the overall influence of the Order of Presentations factor first, it could be removed from further regressions, and the post hoc testing for the ANOVAs described above made marginally more manageable. It was found that participants who completed the experiment in the order PictorialOrthographic-Auditory (POA) produced all vowels closer to the calibration targets. Participants who completed the experimental sections with the pictorial component last produced vowels 77 farthest from those obtained in calibration, regardless of the order that the previous two blocks were presented in. There was no significant difference between the remaining orders of presentation (PAO, OPA, APO), except that all three were significantly distanced from both the POA order and the two Pictorial-last orders (AOP and OAP). Mergers show the greatest divergence from their calibration vowels when presented with stimuli in POA order, and the least divergence in the PAO order (See figure 4.27: This figure includes Target vowels for clarity). F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 OPA - ɛ AOP - ɛ OAP - ɛ PAO - ɛ APO - ɛ 0.2 POA - ɛ 0.4 OPA - æ Calibration e 0.6 PAO - æ AOP - æ APO - æ 0.8 F1 POA - æ 1 OAP - æ 1.2 1.4 1.6 Calibration ae Figure 4-27: Main Effect of: Presentation Order (With Target Included) The implications of these findings will be discussed further in the following chapter. 78 1.8 Discussion and Conclusion 5.1 Confirming Mergers The analysis of the participants’ vowel productions confirms Hypotheses 1 and 2. About half of the participants tested showed no significant differences in formant frequencies between [æ] and [ɛ] before [g], while they maintained this difference in other contexts (before [k], before [d], and before [ɹd]). The comparison of normalized productions (Figure 5-1, reproduced from Chapter 5), in particular the distances between Splitters’ and Mergers’ production of [æg], reveals the type and direction of the merge. 0.9 0.8 0.7 0.6 0.5 F2 0.4 0.3 0.2 0.1 0 -0.1 -0.5 Mergers' [ɛg] Splitters' [ɛg] 0 Mergers' [æg] 0.5 Splitters' [æg] Splitters' [ɛk] Merger's [ɛk] 1 F1 Calibration [ɛ] Splitters' [æk] 1.5 Mergers' [æk] Calibration [æ] 2 2.5 Figure 5-1: Normalized Plot of Mergers' and Splitters' [æg] and [ɛg] productions [_g], [_k], and Calibration contexts. 79 Only half of participants produced [æg] in a way statistically indistinguishable from their productions of [ɛg]. I can conclude that [æ] before [g] in Alberta is currently undergoing a merger by approximation of the unidirectional subtype. The division between vowels before [k], and between speakers, means we can discount a merger by expansion and the “middle ground” subtype of merger by approximation. It is not possible with just the Merge/Split data to determine whether we are seeing a merger by transfer or the unidirectional subtype of a merger by approximation. However, the variable productions noted when speakers were given differing modality prompts, which caused variable productions, indicates this change is neither abrupt nor complete, the two indicators of a merger by transfer. Recall Wassink’s 2015 research that identified significant raising of [æ] before [g] in the Pacific Northwest that was not found in Vancouver, BC (17). The reason for this is uncertain, though. Examining the other vowels of Alberta English as they compare to the Northwestern US and BC Canada may explain why this change seems to be “skipping” the western coast of Canada. If the data from 19 subjects can be extrapolated to the entire population, with half merging and half maintaining distinct vowels in all consonantal environments, it seems likely this merger is at somewhat of an impasse. A re-examination of this phenomenon at a future time will determine, assuming a triadic grammar, whether the merger will be completed due to perpetuation by active and super listeners, or whether the inactive listeners will cause the vowels to drift. 5.2 Personality and Demographic Data Geography provided the only relevant personality/demographic influence on vowel production. Participants who identified as “rural” and those who had lived outside of Alberta for at least a year were both more likely to merge vowels. It is possible this can be attributed to a 80 more locally-focused identity, as established in Labov’s seminal study in Martha’s Vineyard (1972). That is, those who travel away and then return while still under 30 are more likely to retain local dialect features. However, I hesitate to attribute meaning to any of these results due to the low number of data points. Of the 18 respondents, only 2 identified as “rural,” and 7 had spent time outside of Alberta (two of those 7 were the “rural” responders). Hypothesis 3 – that innovators and perpetuators of sound change can be identified by personality - cannot be confirmed based on the results of this experiment. 5.3 The Grammar Triad’s Variable Influence on Production. Balancing the order of presentation is not only good practice in experimentation, but also provided an unexpected windfall in the analysis of these experimental results. The three modalities were intended to act as stand-ins for the three grammars that form the keystone of the theory presented: auditory for community grammar, orthographic for the standard grammar, and pictorial for the self-grammar. Finding significant results based on the order of the presented modalities allows me to further refine the grammar triad, but also to propose a hierarchy of influence. Recall that participants who completed the experiment in the order PictorialOrthographic-Auditory produced vowels closest to the ones produced in the calibration phase. This is because POA presents stimuli in an order in which participants are initially given the least external information that can be linked to specific phonetic data. They must first rely on the selfgrammar, then the standard, and finally the community. That is to say, in the auditory portion, participants receive direct phonetic data which allows them to be susceptible to the process of spontaneous phonetic imitation (Goldinger, 2005) as well as engage their short-term memory (Pisoni, 1973). 81 When presented with orthographic prompts, there is no phonetic signal to imitate or accommodate to. However, orthographic symbols are well known to influence readers’ production of words via sound symbolism (Ehri, 1980). Seeing the letter “a” or “e” may have removed the ambiguity as to what the target sound was for the production, and may have therefore been the reason the speakers in my experiment produced vowels closer to the sounds (and standard grammar) symbolized by those letters ([æ] and [ɛ], respectively) and counteracted the merging influence of [g]. The pictorial prompts contain the least amount of ‘bootstrapping’ data of all, as they depend on speakers’ internal knowledge of the production of the word, their self grammar. Therefore, when presented with the pictorial stimuli first (after the calibration wordlist), participants are producing words in a state as free from external phonetic (or indirectly phonetic) influences as possible in a lab setting. The nature of the pictorial task also lends itself to more spontaneous (non-imitating) productions due to its gamification: The use of game mechanics and approaches to solving non-game problems (Zichermann & Cunningham, 2011). As listeners are trying to guess the word represented by the picture, this task took a much less formal tone than that of the other two stimuli types. Participants are somewhat distracted by the effort required to guess the word, so they are less careful about the actual sounds produced. This is dissimilar to the other two tasks, which encourage more attention: The auditory block asked participants to “repeat what you hear,” and the orthographic reading task is inherently formal. For these reasons, participants given the pictorial block first are most likely to produce those vowels more similarly to the baseline vowels they produced in calibration phase. In the following blocks, participants had their own recent productions to draw on in order to potentially counteract the external influences of the orthographic and auditory stimuli. 82 Participants who underwent the experimental sections with the pictorial component last produced vowels farthest from those obtained in calibration, regardless of the order in which the previous two blocks were presented. This is explained via the same mechanisms described above: External influences in the auditory and orthographic blocks drive participant productions farther from their “self” vowel production, and continue to influence this production through the pictorial block due to their recency effects. Each block took 15 minutes or less to complete, with only a few seconds between stimuli and each token reinforcing a drive away from “self” productions, so it is possible participants’ short-term memory would still be engaged. The middling position of the other blocks (PAO, OPA, APO) can be explained if one assumes that the stimuli types present two sorts of influencing factors at odds with each other: The pictorial block encourages productions truer to a participant’s self-grammar, their baseline vowels. The auditory and orthographic blocks, on the other hand, discourage self-grammar productions and drive participants towards a more imitative, “standard” production. Of these two blocks, direct phonetic information – i.e. the community grammar – has a stronger influence on participant vowels than the indirect sound symbolism of the standard grammar, represented in the orthographic stimuli. 5.4 Factors influencing production 5.4.1 When the only targets are [æ] and [ɛ] In general, participants’ production of [æ] before both velars was closer to their calibration [ɛ], but not significantly so in all cases. This result has the possible implications that the actual target for [æ] is drifting higher and fronter, and the variation within [ɛ] productions indicate this may be a merger by expansion. However, the significant result of target on productions, when compared to both vowels (for all but the Mergers’ before [g] context), suggests these are still distinct 83 sounds in listeners’ minds, even if their productions before [g] do not bear that out. Future research may incorporate a perceptual experiment to determine the exact nature of these sounds within listeners’ minds. That the modality of the stimulus affected production supports the hypothesized hierarchy of influence of the triadic grammar. In all cases, the orthographic stimuli prompted participants to produce vowels closer to the calibration target. Recall that the orthographic stimuli serve as a stand-in for the standard grammar. As explained above, the letters presented in the orthographic block have an influence on speaker production, encouraging participants to produce what they “see,” rather than what they “know.” Regardless of whether a speaker was labelled a “Merger” or a “Splitter,” their production of both vowels before [g] was significantly closer to their baseline [ɛ] than vowels produced before [k]. This cannot be fully explained via the articulatory side effect of raising the velum to create a stop closure (since both [g] and [k] share this). However, the-called “Canadian Shift” in which front lax vowels ([æ], [ɛ], and [ɪ]) retract, may be the culprit, causing the vowels to shift before [k] independently of what they are doing before [g]. Recall that Boberg (2005) found that [ɛ] moving towards [ʌ] (the vowel in “bug”) and [æ] is both lowering and retracting, in Montreal; a region just outside of the homogeneous “Canadian English” zone outlined by Labov et al. 2006 in the Atlas of North American English. The Albertan speakers studied in my experiment, despite being closer geographically to those in Winnipeg, seem to be adopting something closer to Boberg’s version of the shift, with both vowels retracting and lowering before [k]. There were no significant interactions between the Stimulus type and Word factors on production distances from [ɛ]. In nonwords, orthographic and pictorial productions were aligned to calibration vowels less closely than stimuli recorded in the auditory block, but the auditory 84 and orthographic stimulus types were the only two values that varied significantly in distance from each other. This result may be a false positive, caused partially by the nature of the analysis—only a portion of nonwords were tested (those that were produced with non-ambiguous versions of either [æ] or [ɛ], and that did not form a word with either vowel). For this reason, the tests lack the ability to determine the influence of the removed vowels that were presented alongside the tested ones, both in words and nonwords. Still, these tests still indicate that participants, when asked to (re)produce clear vowels, will do so, even in the ambiguous before[g] context. The Word/Nonword status of productions had significant effects on both vowels (Figure 4-26). In most cases, words were produced closer to calibration targets than nonwords. In English, vowels inside words that end with voiced consonants are longer in duration than voiceless consonants, and it has been established by past research that listeners use these durational cues in identifying the consonant (Raphael, 1972). The results of my experiment may support the possibility of this perceptual bootstrapping happening in the opposite direction: Listeners may be using the voicing status of the final consonant to help determine the vowel. When presented with a nonword, listeners are unable to use this secondary cue to help determine the quality of the vowel, because the voicing status of the final consonant is unknown. This may have an effect on sound change: When inactive listeners hear a merged vowel, they can use the voicing cues to determine the intended word spoken (or a near-merged vowel, as in the case of [ɛ] and [æ] before [g]), and thus rebuild the vowel (resisting this particular sound change). However, when a super listener hears the merged vowel, they may pay more attention to the actual acoustic output than the secondary cues, and thus add a “not quite [æ]” to their (community) grammar, which may be reproduced at a later time. To put it a different way, 85 inactive listeners hear the word being said, and assume the (corresponding) acoustic output, while super listeners hear the sounds, and infer the word from the sounds. Thus, the differing results when testing with words and non-words is caused by the differing perceptual approaches of the two types of listeners. 5.4.2 In the [œ]/Non-English vowel condition For the most part, the results from examining the data containing [œ] mirror those of the previous tests, which were limited to [æ] and [ɛ]. Mergers produce all vowels closer to [ɛ] than Splitters, regardless of what sound was actually produced by the stimuli voice in the auditory block. This indicates that Mergers are inactive listeners—or, at least, that they are more inactive than Splitters within the experimental context. Recall that, as outlined in Chapter 2.5, inactive listeners may initiate sound changes, but resist sound changes not of their own invention. Though the question of how the sound change was initially adopted by the Mergers remains, this result indicates that its perpetuation is caused by the inactive listening of Mergers. That is, the Mergers are resisting the influence of the community grammar represented in the experiment by the auditory voice (and therefore resisting the influence of certain parts of the community that happen to not be merging). The reasons why one “chooses” to merge (or rather, why one’s method of perception encourages merging) in the first place cannot be answered with the data available. However, the significant differences in response to auditory stimuli between the two populations – Mergers producing all vowels closer to [ɛ] and Splitters producing all vowels closer to [æ] – indicates that there is a difference in the way these two groups listen, in addition to the way they speak. Switched vowels of both types were produced closer to the baseline [æ], compared to the neutral vowel [œ]. This may indicate that participants are more willing to merge two known 86 vowels than produce non-English vowels inside known vowel spaces. The influence of word formation was only significant with respect to the [ɛ] vowel, which neither supports nor disproves the bootstrapping posited in the last section. When the neutral vowel [œ] was the target, in all cases it was produced significantly far away from each calibration vowel. As the auditory modality served as the stand-in for the community grammar, this result indicates that the community has the greatest influence on production (or at least, has the greatest influence on production when research participants are explicitly asked to repeat what they hear). That is, the sounds we encounter in our day-to-day lives influence our own speech more than our internal representations or our perception of the standard grammar, at least in the short-term. 5.5 Conclusion I have confirmed the existence of a vowel shift among [æ]/[ɛ] in Alberta. I have determined this is a merger by approximation of the second subtype, in which [æ] is rising and fronting to the space of [ɛ]. I have further identified it as a merger in progress based on two distinct populations that treat [æ] differently when presented in the specific context of preceding a voiced velar stop [g], and that process information differently when it is presented auditorily. Finally, I have gathered evidence that there exists a hierarchy of influence of stimulus modality on speaker production. I hypothesize that these modality differences are caused by a triadic grammar within each individual: the self grammar, the standard grammar, and the community grammar. Just as we are products of our experiences, our speech cannot be exclusively our own. We present ourselves in the way we dress, our mannerisms and affectations, and who we choose to associate ourselves with. The way we speak is not divorced from these performances. As 87 Emily Dickinson says in the poem quoted in Chapter One, a word’s life begins once it is spoken. Our experiences make the sounds of each word individual, and words live on after the sound fades away, via their influence on listeners. If speaking gives words life, then the three grammars are their parents. 88 References Alan, C. L., Carissa Abrego-Collier, & Morgan Sonderegger. (2013). Phonetic imitation from an individual-difference perspective: Subjective attitude, personality and “autistic” traits. PloS one 8.9 Arctander, K., Kinney, H., & Esposito, C. "Low‐back vowel merger in Minnesotan English." The Journal of the Acoustical Society of America 126 (2009): 2162. Baker, A., Mielke, J., & Archangeli, D. (2008) "More velar than /g/: Consonant coarticulation as a cause of diphthongization." Proceedings of the 26th West Coast Conference on Formal Linguistics. Baker, A., Archangeli, D., & Mielke, J. (2011). Variability in American English s-retraction suggests a solution to the actuation problem. Language Variation and Change, 23: 347-374. Bauer, M., & Parker, F. (2008). /æ/-raising in Wisconsin English. American Speech, 83:4: 403431. Becker, K. (2009). /r/ and the construction of place identity on New York City's Lower East Side. Journal of Sociolinguistics, 13:5:634-658. Becker, K. (2010). Regional dialect features on the Lower East Side of New York City: Sociophonetics, ethnicity, and identity (Doctoral dissertation, New York University). Beddor, P. S. (2012) Perception grammars and sound change. In M-J Solé, & D. Recasens (Eds.) The initiation of sounds change: Perception, production, and social factors. 37-56 Philidelphia: John Benjamins Publishing Company. Bell-Berti, F., & Krakow, R. A. (1991). Anticipatory velar lowering: A coproduction account. The Journal of the Acoustical Society of America, 90.1: 112-123. 89 Bishop, D. V. M., & Bishop, S. J. (1998). Twin Language: A Risk Factor for Language Impairment?. Journal of Speech, Language, and Hearing Research, 41.1, 150-160. Brandth, B. (1995). Rural masculinity in transition: gender images in tractor advertisements. Journal of rural studies, 11.2, 123-133. Boberg, C. (2005) "The Canadian Shift in Montreal." Language Variation and Change 17.2: 133-154. Bourlai, E., & Herring, S. C. (2014). Multimodal communication on tumblr: i have so many feels!. In Proceedings of the 2014 ACM conference on Web science. 171-175. ACM. Bucholtz, M. (2003). Sociolinguistic nostalgia and the authentication of identity. Journal of sociolinguistics, 7.3: 398-416. Burzio, L. (2007) "Phonology and phonetics of English stress and vowel reduction." Language Sciences 29.2: 154-176. Bybee, J. (2002). Word frequency and context of use in the lexical diffusion of phonetically conditioned sound change. Language variation and change, 14.3: 261-290 Bybee, J. (2012) Patterns of lexical diffusion and articulatory motivation for sound change. In M-J Solé, & D. Recasens (Eds.) The initiation of sounds change: Perception, production, and social factors. 37-56. Philidelphia: John Benjamins Publishing Company. Campbell, L. (2004). Historical Linguistics: An Introduction. MIT Press. Casserly, E. D., & Pisoni, D. B. (2010). Speech perception and production. Wiley Interdisciplinary Reviews: Cognitive Science, 1.5: 629-647. Chiswick, B. R., & Miller, P. W. (2005). Linguistic distance: A quantitative measure of the distance between English and other languages. Journal of Multilingual and Multicultural Development, 26.1: 1-11. 90 Clarke, S., Elms, F., & Youssef, A. (1995) "The third dialect of English: Some Canadian evidence." Language Variation and Change 7.02: 209-228. Clements, G. N. (1991). Place of articulation in consonants and vowels: a unified theory. Working papers of the Cornell phonetics laboratory, 5: 77-123. Clopper, C., Pisoni, D., & De Jong, K. (2005). "Acoustic characteristics of the vowel systems of six regional varieties of American English." The Journal of the Acoustical Society of America 118:1661. Conn, J. (2002). It’s not all Rain and Coffee: An Investigation into the Western Dialect of Portland, Oregon. Presented at NWAV31, Stanford, California. Retrieved from http://www.pds.pdx.edu/. Davies, S., & Goldberg, J. M. (2006). Clinical aspects of transgender speech feminization and masculinization. International Journal of Transgenderism, 9.3-4: 167-196. Dickinson, Emily. 1924. “89. A word is dead.” The complete poems of Emily Dickinson. Boston: Little, Brown, 1924; Bartleby.com, 2000. www.bartleby.com/113/. Dimov, S., Katseff, S., and Johnson, K.. (2012) Social and personality variables in compensation for altered auditory feedback. In M-J Solé, & D. Recasens (Eds.) The initiation of sounds change: Perception, production, and social factors. 185-210. Philidelphia: John Benjamins Publishing Company. Drager, K. (2010). Sensitivity to grammatical and sociophonetic variability in perception. Journal of Laboratory Phonology, 1: 93-120. Eckert, P. (2008). Variation and the indexical field. Journal of Sociolinguistics. 12.4: 453-476. Ehri, L. C., & Snowling, M. J. (2004). Developmental variation in word recognition. Handbook of language and literacy: Development and disorders. 433-460. 91 Esling, J., & Warkentyne, H. (1993) "Retracting of/æ/in Vancouver English." Focus on Canada. Philadelphia: John Benjamins: 229-246. Evans, B. (2004). The role of social network in the acquisition of local dialect norms by Appalachian migrants in Ypsilanti, Michigan. Language Variation and Change, 16: 153167. Findlay, J. M., & Coates, K. (2002). Parallel destinies: Canadian-American relations west of the Rockies. McGill-Queen's Press-MQUP. Flemming, E. (2003). The relationship between coronal place and vowel backness. Phonology, 20.03: 335-373. Foulkes, P. & Docherty, G. (2006). The social life of phonetics and phonology. Journal of Phonetics, 34: 409-438. Foulkes, P. (2010). Exploring social-indexical knowledge: a long past but a short history. Journal of Laboratory Phonology, 1: 5-39. Freeman, V. (2014). Bag, beg, bagel: Prevelar raising and merger in Pacific Northwest English. 32. University of Washington Working Papers in Linguistics. Garrett, A. & Johnson, K. (2013). Phonetic bias in sound change. In A. Yu (ed.), Origins of sound change: approaches to phonologization, 201-227. Goldinger, Stephen D. "The cognitive basis of spontaneous imitation: Evidence from the visual world." Proceedings of Meetings on Acoustics. 19.1. Acoustical Society of America, 2013. Goldstone, R. L., & Hendrickson, A. T. (2010). Categorical perception. Wiley Interdisciplinary Reviews:Cognitive Science, 1.1: 69-78. 92 Grosvald, M. & Corina, D. (2012) The production and perception of sub-phonemic vowel contrasts and the role of the listener in sound change. In M-J Solé, & D. Recasens (Eds.) The initiation of sounds change: Perception, production, and social factors. 77-101. Philidelphia: John Benjamins Publishing Company. Hagiwara, R. (1997) "Dialect variation and formant frequency: The American English vowels revisited." The Journal of the Acoustical Society of America 102: 655. Hagiwara, R. (2006) "Vowel production in Winnipeg." The Canadian Journal of Linguistics/La revue canadienne de linguistique 51.2: 127-141. Harrington, J. (2012) The coarticulatory basis of diachronic high back vowel fronting. In M-J Solé, & D. Recasens (Eds.) The initiation of sounds change: Perception, production, and social factors. 103-122. Philidelphia: John Benjamins Publishing Company. Harrington, J. (2007). Evidence for a relationship between synchronic variability and diachronic change in the Queen's annual Christmas broadcasts. In J. Cole & J.Hualde (Eds.), Laboratory Phonology 9: 125-143. Berlin: Mouton. Hay, J., Drager, K. and Thomas, B. (2013). Using nonsense words to investigate vowel merger. English Language and Linguistics 17.2: 241-269. Herrero, B. P. (2008). “Deep and raspy” or “high and squeaky”: a cross-linguistic study of voice perception and voice labeling. Experimental Linguistics:109. Irons, T. (2007). "On the status of low back vowels in Kentucky English: More evidence of merger." Language variation and change 19.2: 137. Johnstone, B. & Kiesling, S. (2008). Indexicality and experience: exploring the meanings of /aw/-monophthongization in Pittsburgh. Journal of Sociolinguistics, 12.1: 5-33. 93 Johnson, K. (2008). 15 Speaker Normalization in Speech Perception. The handbook of speech perception, 363. Jose, B. (2010). The Apparent-Time Construct and Stable Variation: Final /z/ Devoicing in Northwestern Indiana. Journal of Sociolinguistics, 14.1: 34-59. Klatt, D. H. (1979). Speech perception: A model of acoustic-phonetic analysis and lexical access. Journal of phonetics, 7.312: 1-26. Koerner, E. F. K. (1989). On ‘Unrewriting the History of Linguistics’. In Practicing linguistic historiography: selected essays. 50. John Benjamins Publishing. Koerner, E. F. K. (Ed.). (1995). Writings in General Linguistics: On Sound Alternation (1881) and Outline of Linguistic Science (1883). 11. John Benjamins Publishing. Labov, W. (1972). The social stratification of (r) in New York City Department Stores. In W. Labov (ed.), Sociolinguistic Patterns. Philadelphia, PA: University of Pennsylvania Press, 43-54. Labov, W. (1972). The social motivation of a sound change. In W. Labov (ed.), Sociolinguistic Patterns. Philadelphia, PA: University of Pennsylvania Press, 1-41. Labov, W., Ash, S. & Boberg, C. (2006). Canada. The Atlas of North American English. Berlin, Boston: Mouton de Gruyter. Labov, W. (1963). The social motivation of a sound change. Word-Journal of the International Linguistic Association, 19.3: 273-309. Labov, W. (2007). Transmission and Diffusion. Language, 83(2), 344-387. Ladefoged, P. (1980). What are linguistic sounds made of?. Language, 485-502. Ladefoged, P. (2006). A Course in Phonetics (5th ed.). Thomson Wadsworth. 94 Lawson, E., Scobbie, J., & Stuart-Smith, J. (2011). The social stratification of tongue shape for postvocalic /r/ in Scottish English. Journal of Sociolinguistics, 15.2: 256-268. Liberman, A. M., Harris, K. S., Hoffman, H. S., & Griffith, B. C. (1957). The discrimination of speech sounds within and across phoneme boundaries. Journal of experimental psychology, 54.5: 358. Liberman, A. M., Harris, K. S., Kinney, J. A., & Lane, H. (1961). The discrimination of relative onset-time of the components of certain speech and nonspeech patterns. Journal of experimental psychology, 61.5: 379. Liljencrants, J., & Lindblom, B. (1972). Numerical simulation of vowel quality systems: The role of perceptual contrast. Language, 839-862. Lobanov, B. M. (1971). Classification of Russian vowels spoken by different speakers. The Journal of the Acoustical Society of America, 49.2B: 606-608. Love, J. & Walker, A. (2013). Football versus football: effect of topic on /r/ realization in American and English sports fans. Language and Speech, 56.4: 443-460. McGurk, H., & MacDonald, J. (1976). Hearing lips and seeing voices. Nature, 264: 746-748. Maegaard, M., Jensen, T., Kristiansen, T., & Jorgensen, J. (2013). Diffusion of language change: accommodation to a moving target. Journal of Sociolinguistics, 17.1: 3-36. Majors, T. (2005). "Low back vowel merger in Missouri speech: Acoustic description and explanation." American speech 80.2: 165-179. Meyerhoff, M. (2011). Introducing sociolinguistics. Taylor & Francis. Miller, C. (2012). Reading in Time: Emily Dickinson in the Nineteenth Century. Univ of Massachusetts Press. 95 Montgomery, C. (2012). The effect of proximity in perceptual dialectology. Journal of Sociolinguistics, 16.5: 638-668. Nearey, T. M. (1989). Static, dynamic, and relational properties in vowel perception. The Journal of the Acoustical Society of America, 85.5: 2088-2113. Ohala, J. (2012) The listener as a source of sound change: An update. In M-J Solé, & D. Recasens (Eds.) The initiation of sounds change: Perception, production, and social factors. 21-36. Philadelphia: John Benjamins Publishing Company. Peterson, G. E., & Barney, H. L. (1952). Control methods used in a study of the vowels. The Journal of the Acoustical Society of America, 24.2: 175-184. Pierrehumbert, J. B. (2001). lenition and contrast. Frequency and the emergence of linguistic structure, 45: 137. Pierrehumbert, J. B., Bent, T., Munson, B., Bradlow, A. R., & Bailey, J. M. (2004). The influence of sexual orientation on vowel production (L). The Journal of the Acoustical Society of America, 116.4: 1905-1908. Pinker, S. (1999). The language mavens. The Workings of Language: From Prescriptions to Perspectives, 3-14. Pisoni, D. B. (1973). Auditory and phonetic memory codes in the discrimination of consonants and vowels. Perception & Psychophysics, 13.2: 253-260. Purnell, T. C. (2008). Prevelar raising and phonetic conditioning: Role of labial and anterior tongue gestures. American Speech, 83.4: 373-402. Raphael, L. J. (1972). Preceding Vowel Duration as a Cue to the Perception of the Voicing Characteristic of Word‐Final Consonants in American English. The Journal of the Acoustical Society of America, 51.4B: 1296-1303. 96 Reed, C. E. (1961). The pronunciation of English in the Pacific Northwest. Language, 559:564. Riebold, J. M. (2012). Please merge ahead: The vowel space of Pacific Northwestern English. Abstract for a paper presented at the Northwest Linguistics Conference, University of Washington, Seattle, WA. Riebold, J. M. (2015). The Social Distribution of a Regional Change:/æg, ɛg, eg/in Washington State (Doctoral dissertation). Robins, R. H. (2013). A short history of linguistics. Routledge. Rosen, N., & Skriver, C. (2015). Vowel patterning of Mormons in Southern Alberta, Canada. Language & Communication. Sharma, D. & Sankaran, L. (2011). Cognitive and social forces in dialect shift: gradual change in London Asian speech. Language Variation and Change, 23: 399-428. Smith, E. E., & Medin, D. L. (2002). The exemplar view. Foundations of cognitive psychology: Core readings, 277-292. Solé, M.-J. (2012) Natural and unnatural patterns of sound change. In M-J Solé, & D. Recasens (Eds.) The initiation of sounds change: Perception, production, and social factors. 123146. Philadelphia: John Benjamins Publishing Company. Stanford, J. (2008). A sociotonetic analysis of Sui dialect contact. Language Variation and Change, 20.3: 409-450. Stanford, J. & Kenny, L. (2013). Revisiting transmission and diffusion: an agent-based model of vowel chain shifts across large communities. Language Variation and Change, 25: 119153. 97 Stevens, K. N., & Keyser, S. J. (2010). Quantal theory, enhancement and overlap. Journal of Phonetics, 38.1: 10-19. Szakay, A. (2012). Voice quality as a marker of ethnicity in New Zealand: from acoustics to perception. Journal of Sociolinguistics, 16.3: 382-397. Trawick-Smith, B. (2011, November 10). What’s with the Western US and Velars? [Web log post]. Retrieved from http://dialectblog.com/2011/11/10/the-western-us-and-velars/. Wassink, A. B., Squizzero, R., Scanlon, M., Schirra, R., & Conn, J. (2009) Effects of Style and Gender on Fronting and Raising of/æ/,/e:/and/ε/before/g/in Seattle English. Wassink, A. B., & Riebold, J. M. (2013). Individual variation and linguistic innovation in the American Pacific Northwest. In Chicago Linguistic Society (CLS 49) Workshop on Sound Change Actuation: 18-20. Wassink, A. B. (2014). Vowel Raising in Washington State: What's the BAG deal? Presented at the first meeting of the Cascadia Workshop in Sociolinguistics (CWSL). March 1-2, 2014 (U of Victoria). Wassink, A. B. (2015) Vowel Raising in Washington State: the cat's out of the B(ɛ̝g). Presentation given as part of the panel, "Western Speech". Annual meeting of the American Dialect Society, Portland OR Yu, A. (2013). Individual differences in socio-cognitive processing and the actuation of sound change. In A. Yu (ed.), Origins of sound change: approaches to Phonologization. 201227. Zeller, C. (1997) "The investigation of a sound change in progress:/æ/to/e/in midwestern American English." Journal of English linguistics 25.2: 142-155. 98 Zichermann, G., & Cunningham, C. (2011). Gamification by design: Implementing game mechanics in web and mobile apps. " O'Reilly Media, Inc." 99 APPENDIX A: PILOT STUDY STIMULI ORTHOGRAPHIC PICTORIAL exhumation bag segment bagel protagonist bagpipe lab beggar mandrake Craigie Hall agony dagger flake dragon pragmatic egg magnolia flag fable Pregnant exit Jaguar impregnate leg dishrag magazine rutabaga maggot exile magnet pagan magpie haggard nag Copenhagen negative 100 break Niagara (falls) fragment Maggie (simpson) phlegmatic peg existential plague elect Raggedy Anne (ragdoll) flexible reggae vague Megan Fox wagtail stag nutmeg Clegg (Nick) blab tag hack magma negligent Vegas exaltation Viagra jagged wag agriculture wagon / winnebago handbag zigzag megabyte stalactite segregation stalagmite jab mega man legume drag magnitude Legos 101 Baghdad Haggis zigzag baby stagger crab suspect cab braggart cable Blake rabies diagonal bacon apex bake / bakery slacken cake bootleg cupcake Winnebago Scooby Doo tag hay(stack) undertaker earthquake Drake pancake lag rake reneg (rattle)snake regular steak vagabond toothaceh muskeg cactus trachea (snap, crackle, pop) eclair pterodactyl 102 subtract raccoon fragrance axe gag raquet segue checkers nag cheque/paycheque Lagos textbook Tobago wax agnostic tax hag yak quaker quack stagette Legolas coagulant (daniel) Craig integrity Baggins (bilbo/frodo) Shaganappi Ewan McGregor bake airplane eczema Pythagoras windbag Reagan magnate carl sagan gulag san diego icebreaker schwarzaneggar degradation Winnipeg 103 thorax Shaggy vagary Quagmire (family guy) aggravate Hagrid excerpt Jacob (twilight dab Beck(ham) sag bread exhortation gun legacy monkey tobacco flower jagged hamburger stagnant unicorn vagrancy throne magnesium king flack bamboo acrimony forest/rainforest Deb vegetables/salad fake machete The Keg chili pepper flex mail carrier 104 APPENDIX B: SUBJECT QUESTIONAIRRE Subject Code: ____________________ QUESTIONNAIRE Questions on this form are optional. Any question you do not wish to answer can be left blank. 1. Date of Birth? ________________________________ 2. Gender ________________________________ 3. Sexual Orientation ________________________________ 4. What is your highest level of education? _______________________________ 5. What is your parents’ highest level of education? ________________________________ 6. Where is your home town? (City, Province, Country) ________________________________ 7. Where have you lived? (City, # of years) __Ex: Athens, Alabama, USA, 10 years_ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ 8. What is your first language(s)? ________________________________ 9. Other Languages Spoken (list language, rate fluency from 1-3 where 1 = rudimentary and 3 = fluent, and list age learned) _Ex: French, 3, 6 years______________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ 105 12. Do you consider yourself: (circle) Rural / Urban Liberal / Conservative Family-focused / Friend-focused 12. Introverted / Extroverted Tech-savvy / Tech-apathetic Religious / Nonreligious Leader / Follower Thinker / Feeler Handedness Left / Right 106 APPENDIX C: CALIBRATION VOWEL WORD LIST Word Heed Hid Hayed Hoist Had Hud Hawed Hide Hoed Head Hood Who’d Herd Hod How Hard Here Hare Whore Vowel i ɪ eɪ ɔɪ æ ʌ ɑ aɪ o / oʊ ɛ ʊ U ɜ˞ ɑ aʊ ɑɹ ɪɹ ɛɹ ɔɹ 107 APPENDIX D: EXPERIMENTAL STIMULI WORD LISTS Condition Word/Nonword Number Shared Pictorial Word 20 6 Pictorial Nonword 20 6 Total 40 Orthographic Word 20 6 Orthographic Nonword 20 6 40 Auditory - Clear Word 15 6 Auditory – Clear Nonword 15 6 Auditory – Merged Word 15 4 Auditory – Merged Nonword 15 4 Auditory – Mismatched Word 15 4 Auditory – Mismatched Nonword 15 4 90 Total: Shared Wordlist – Word: 1. 2. 3. 4. 5. 6. 170 12 (see below) Shared Wordlist – NonWord: Egg Stegosaurus Dragon Flag Shrek Black 1. 2. 3. 4. 5. 6. 108 Bregg (brain + egg) Snegg (snake + egg) Dag (dog + flag) Chag (chair + flag) Fek (fox + Shrek) Vack (Vest + Yak) 170 Pictorial Words: Pictorial Nonwords: 1. Magpie 2. Baggins 3. Hagrid 4. Magnet 5. Jaguar 6. Sandiego 7. Leg 8. Pregnant 9. Legolas 10. Regan 11. Freckle 12. Neck 13. Yak 14. Cactus Orthographic Words: 1. vag 2. yag 3. skag 4. kag 5. pag 6. negg 7. jegg 8. hegg 9. regg 10. tegg 11. yeck 12. seck 13. dack 14. gak Orthographic NonWords: 1. Stalagmite 2. Zigzag 3. Plague 4. Magma 5. Dagger 6. Negative 7. Winnipeg 8. Beg 9. Segregate 10. Reggae 11. Wreck 12. Peck 13. Stalactite 14. Attack 1. plag 2. trag 3. vag 4. twag 5. pag 6. deg 7. feg 8. jeg 9. zeg 10. yeg 11. sleck 12. teck 13. chack 14. skack 109 Auditory Words – Clear: 1. 2. 3. 4. 5. 6. 7. 8. 9. Auditory NonWords – Clear: Maggot Haggis Plague Mega Winnipeg Pregnant Beg Peck Yak Auditory Words – Merged: 1. 2. 3. 4. 5. 6. 7. 8. 9. yag skag trag pag deg zeg yeg teck skak Auditory NonWords – Merged: 1. B_g 2. P_g 3. N_g 4. L_g 5. _g 6. Dr_g 7. St_g 8. Pr_g 9. R_g 10. M_g 11. H_k 12. F_k 13. P_k 14. R_k 15. Bl_k 1. D_g 2. J_g 3. Qu_g 4. Gl_g 5. V_g 6. Y_g 7. Str_g 8. Th_g 9. Sp_g 10. Ch_g 11. F_k 12. G_k 13. V_k 14. Th_k 15. Sw_k 110 Auditory Words – Mismatched: Word Target Word Vowel Replacement New Word Vowel 1. Magpie æ ɛ Mɛgpie 2. Jaguar æ ɛ Jɛguar 3. Dragon æ ɛ Drɛgon 4. Shaggy æ ɛ Shɛggy 5. Zigzag æ ɛ Zigzɛg 6. Pregnant ɛ æ Prægnant 7. Stegosaurus ɛ æ Stægosaurus 8. Mega ɛ æ Mæga 9. Winnipeg ɛ æ Winnipæg 10. Keg ɛ æ Kæg 11. Cactus æ ɛ Cɛctus 12. Black æ ɛ Blɛck 13. Freckle ɛ æ Fræckle 14. Crack æ ɛ Crɛck 15. Yak æ ɛ Yɛk 111 Auditory NonWords – Mismatched: Word Target Word Vowel Replacement New Word Vowel 1. vag æ ɛ vɛg 2. dag æ ɛ dɛg 3. trag æ ɛ trɛg 4. yag æ ɛ yɛg 5. skag æ ɛ skɛg 6. cheg ɛ æ chæg 7. yeg ɛ æ yæg 8. queg ɛ æ quæg 9. peg ɛ æ pæg 10. theg ɛ æ thæg 11. thak æ ɛ thɛk 12. vak æ ɛ vɛk 13. skak æ ɛ skɛk 14. gek ɛ æ gæk 15. strek ɛ æ stræk 112 APPENDIX E: PLOTS OF ANOVA RESULTS Main Effects of: Vowel Target F2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 ɛ 0.4 Calibration ɛ 0.6 0.8 F1 æ 1 1.2 1.4 1.6 Calibration æ 1.8 Main Effects of: Context 0.7 0.6 0.5 F2 0.4 0.3 0.2 0.1 0 0 0.2 Before [g] 0.4 Calibration ɛ 0.6 0.8 1 Before [k] 1.2 1.4 Calibration æ 1.6 1.8 113 F1 0.8 Main Effects of: Word F2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Calibration ɛ 0.2 Words 0.4 0.6 Nonwords F1 0.8 1 1.2 1.4 1.6 Calibration æ 1.8 Main Effects of: Merge F2 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 Mergers 0.6 Splitters 0.8 F1 0.7 1 1.2 1.4 Calibration æ 1.6 1.8 114 Main Effects of: Order 0.8 0.7 0.6 F2 0.4 0.5 0.3 0.2 0.1 0 0 0.2 OPA 0.4 PAO Calibration e AOP APO POA 0.6 OAP F1 0.8 1 1.2 1.4 1.6 Calibration ae 1.8 Main Effects of: Order (with Target) F2 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 AOP - ɛ OPA - ɛ PAO - ɛ APO - ɛ OAP - ɛ POA - ɛ 0.2 0.4 OPA - æ Calibration e 0.6 PAO - æ AOP - æ APO - æ 0.8 POA - æ F1 0.9 1 OAP - æ 1.2 1.4 1.6 Calibration ae 115 1.8 Two-Way Interactions between Target and Context F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.2 ɛg 0 0.2 0.4 Calibration ɛ æg 0.8 F1 0.6 ɛk 1 1.2 1.4 æk Calibration æ 1.6 1.8 Two-Way Interactions between Target and Merge 0.8 0.7 0.6 0.5 F2 0.4 0.3 0.2 0.1 0 0 Splitters' [ɛ] Mergers' [ɛ] 0.2 0.4 Calibration [ɛ] 0.6 F1 0.8 Mergers' [æ] Splitters' [æ] 1 1.2 1.4 Calibration [æ] 1.6 1.8 116 Two-Way Interactions between Target and Word 0.8 0.7 0.6 F2 0.4 0.5 0.3 0.2 0.1 0 0 0.2 Nonwords with [ɛ] 0.4 Calibration [ɛ] 0.8 Words with [ɛ] 1 Words with [æ] F1 0.6 Nonwords with [æ] 1.2 1.4 1.6 Calibration [æ] 1.8 Two-Way Interactions between Context and Merge F2 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 [_g] Merger 0.2 [_g] Splitter 0.4 0.6 0.8 F1 0.9 1 [_k] Splitter [_k] Merger 1.2 1.4 Calibration æ 1.6 1.8 117 Two-way Interactions between Context and Word 0.8 0.7 0.6 F2 0.4 0.5 0.3 0.2 0.1 0 0 Words that end in g Nonwords that end in g Calibration e 0.2 0.4 0.6 F1 0.8 1 Nonwords that end in 1.2 k Words that end in k 1.4 Calibration ae 1.6 1.8 TwoWay interaction between Stimulus Type and Word F2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Calibration [ɛ] 0.2 Auditory Nonwords 0.4 Orthographic Words Auditory Words Pictorial Words Orthographic Pictorial Nonwords Nonwords 0.6 F1 0.8 1 1.2 1.4 Calibration ae 1.6 1.8 118 Three-way Interaction between StimType, Target, and Context F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.5 Orthographic [ɛg] Auditory [ɛg] 0 Auditory [æg] Pictorial [ɛg] Calibration ɛ 0.5 F1 Auditory [ɛk] Pictorial [æg] Orthographic [ɛk] Orthographic [æg] Pictorial [ɛk] 1 Orthographic [æk] Auditory [æk] 1.5 Pictorial [æk] Calibration æ 2 Three-way Interaction between StimType, Target, and Word 0.7 0.6 0.5 Auditory W [ɛ] Auditory NW [ɛe] Pictorial W [ɛ] Orthographic W [ɛ] Pictorianl NW [ɛ] Orthographic NW [ɛ] Calibration ɛ F2 0.4 0.3 0.1 0 0 0.2 0.4 Auditory NonWord [æ] Pictorial Word [æ] Pictorial Nonword [æ] 0.2 0.6 Orthographic Word [æ] Auditory Word [æ] Orthographic Nonword [æ] 0.8 F1 0.8 1 1.2 1.4 1.6 Calibration ae1.8 119 Three-way Interactions between Target, Context, and Merge 1 0.9 0.8 0.7 F2 0.5 0.6 0.4 0.3 0.2 0.1 0 -0.5 ɛgM ɛgS 0 aegM 0.5 Calibration ɛ ɛkM F1 ægS ɛkS 1 aekS 1.5 aekM Calibration ae 2 Three-way interactions between Stimtype, Context, and Word F2 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Auditory Words [_g] Orthographic Words [_g] Pictorial Words [_g] Auditory NonWords [_g] 0.2 0.4 0.6 Calibration ɛ Pictorial Nonwords [_g] F1 0.8 Auditory Nonwords [_k] Pictorial Nonwords [_k] Auditory Words [_k] 1 Orthographic Words [_k] 1.2 Pictorial Words [_k] 1.4 Orthographic Nonwords [_k] 1.6 1.8 Calibration ae 120 Three-way Interactions between Target, Context, and Word 0.8 0.7 0.6 0.5 0.3 0.2 0.1 0 -0.1 -0.5 Word [ɛg] 0 Nonwords [ɛg] Nonwords [ɛk] Calibration ɛ Words [æg] 0.5 Nonwords [æg] F1 0.9 F2 0.4 1 Words [ɛk] Words [æk] 1.5 Nonwords [æk] Calibration æ 2 121 APPENDIX F: POST HOC RESULTS Post hoc testing of the Analysis of Variance of the normalized score at 50% of the Calibration [ɛ], of the non-[œ] stimuli, as influenced by StimType:Target:Context:Word:Merge Main Effect: StimType diff lwr upr p (adjusted) O-A 0.08337593 0.01526796 0.1514839 0.0115068 P-A 0.09575212 0.0274172 0.16408704 0.002962 P-O 0.01237619 -0.05680438 0.08155677 0.9075402 Main Effect: Target diff e-ae -0.1591878 lwr -0.2056059 Main Effect: Context diff lwr k-g 0.448036 0.4011005 Main Effect: Word diff W-NW 0.1375615 upr -0.1127697 p (adjusted) 0 upr 0.4949715 p (adjusted) 0 lwr 0.09078881 upr 0.1843342 p (adjusted) 0 Two-Way Interaction: Target:Context diff lwr ae:k-ae:g 0.76050616 0.67731523 e:k-e:g 0.15469208 0.07288201 e:k-ae:k -0.58683142 -0.68367482 upr 0.8436971 0.23650214 -0.489 p (adjusted) 0 0.0000074 0 Two-Way Interaction: Target:Word diff lwr e:NW-ae:NW -0.24925665 -0.33062064 ae:W-ae:NW 0.06413977 -0.02212835 e:W-e:NW 0.23681741 0.15165063 e:W-ae:W -0.07657901 -0.16644264 upr -0.16789265 0.1504079 0.32198419 0.01328461 p(adjusted) 0 0.2233932 0 0.1260187 122 Two-Way Interaction: Context:Word diff lwr k:NW-g:NW 0.40814003 0.32423072 g:W-g:NW 0.09648346 0.02773386 k:W-k:NW 0.16959594 0.06810373 k:W-g:W 0.48125251 0.39188582 upr 0.4920493 0.1652331 0.2710882 0.5706192 p(adjusted) 0 0.0017914 0.0001068 0 Two-Way Interaction: Target:Merge diff lwr e:M-ae:M -0.07808959 -0.16407133 ae:S-ae:M 0.1080304 0.02233037 e:S-e:M -0.05355878 -0.13954052 e:S-ae:S -0.23967877 -0.3253788 upr 0.007892156 0.193730431 0.032422963 -0.153978743 p(adjusted) 0.0905102 0.0066338 0.3779186 0 Three-way Interaction: StimType:Target:Context diff lwr upr O:ae:g-A:ae:g 0.185571926 0.055573565 0.315570286 A:e:g-A:ae:g 0.128183534 0.012992976 0.243374092 A:ae:k-A:ae:g 0.902681515 0.741941531 1.063421498 P:ae:k-P:ae:g 0.790179769 0.608468421 0.971891117 O:ae:k-O:ae:g 0.628157679 0.445274201 0.811041156 O:e:k-O:e:g 0.338715333 0.159062125 0.518368540 P:e:k-P:e:g 0.197838625 0.016001220 0.379676030 A:e:k-A:ae:k -0.715916866 -0.893929371 -0.537904362 O:e:k-O:ae:k -0.425917280 -0.639486150 -0.212348410 P:e:k-P:ae:k -0.559897272 -0.773466142 -0.346328402 O:e:k-A:e:k 0.201047676 0.014189497 0.387905856 P:e:g-A:ae:g 0.139438691 0.009268316 0.269609066 O:ae:k-A:ae:g 0.813729605 0.641999739 0.985459471 P:ae:k-A:ae:g 0.897174588 0.725444722 1.068904454 A:e:k-A:ae:g 0.186764648 0.049668098 0.323861198 O:e:k-A:ae:g 0.387812324 0.217328103 0.558296546 P:e:k-A:ae:g 0.337277316 0.166793094 0.507761537 A:ae:k-O:ae:g 0.717109589 0.544504370 0.889714807 P:ae:k-O:ae:g 0.711602662 0.528719185 0.894486139 O:e:k-O:ae:g 0.202240399 0.020526092 0.383954705 A:ae:k-P:ae:g 0.795686695 0.624323895 0.967049495 O:ae:k-P:ae:g 0.706734786 0.525023438 0.888446133 O:e:k-P:ae:g 0.280817505 0.100282919 0.461352091 123 p(adjusted) 0.0002000 0.0146819 0.0000000 0.0000000 0.0000000 0.0000001 0.0195210 0.0000000 0.0000000 0.0000000 0.0223528 0.0235609 0.0000000 0.0000000 0.0005370 0.0000000 0.0000000 0.0000000 0.0000000 0.0146541 0.0000000 0.0000000 0.0000251 P:e:k-P:ae:g A:ae:k-A:e:g O:ae:k-A:e:g P:ae:k-A:e:g O:e:k-A:e:g P:e:k-A:e:g A:ae:k-O:e:g O:ae:k-O:e:g P:ae:k-O:e:g P:e:k-O:e:g A:ae:k-P:e:g O:ae:k-P:e:g P:ae:k-P:e:g O:e:k-P:e:g O:e:k-A:ae:k P:e:k-A:ae:k A:e:k-O:ae:k P:e:k-O:ae:k A:e:k-P:ae:k O:e:k-P:ae:k diff lwr upr 0.230282497 0.049747911 0.410817083 0.774497980 0.612751721 0.936244240 0.685546071 0.512873962 0.858218179 0.768991054 0.596318945 0.941663163 0.259628790 0.088195480 0.431062101 0.209093782 0.037660471 0.380527092 0.853584523 0.683150525 1.024018520 0.764632613 0.583796908 0.945468318 0.848077596 0.667241891 1.028913301 0.288180324 0.108527116 0.467833531 0.763242824 0.590508015 0.935977633 0.674290914 0.491285124 0.857296704 0.757735897 0.574730108 0.940741687 0.248373634 0.066536229 0.430211039 -0.514869190 -0.719705362 -0.310033018 -0.565404199 -0.770240371 -0.360568027 -0.626964957 -0.814960320 -0.438969593 -0.476452289 -0.690021159 -0.262883419 -0.710409940 -0.898405303 -0.522414577 -0.509362263 -0.722931133 -0.295793393 p(adjusted) 0.0018509 0.0000000 0.0000000 0.0000000 0.0000497 0.0039148 0.0000000 0.0000000 0.0000000 0.0000109 0.0000000 0.0000000 0.0000000 0.0005090 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 Three-way Interaction: StimType:Target:Word diff lwr upr O:ae:NW-A:ae:NW 0.2074225577 0.049595841 0.365249274 O:e:NW-O:ae:NW -0.3737406325 -0.563804995 -0.183676271 P:e:NW-P:ae:NW -0.3050081136 -0.490003179 -0.120013048 O:e:W-O:e:NW 0.2277594586 0.041832422 0.413686495 P:e:W-P:e:NW 0.3503885835 0.160013811 0.540763356 O:e:NW-A:ae:NW -0.1663180748 -0.330057379 -0.002578770 O:ae:W-A:ae:NW 0.1696502532 0.005186820 0.334113686 P:e:W-A:ae:NW 0.1952911444 0.031191625 0.359390664 A:e:NW-O:ae:NW -0.2717179284 -0.429544645 -0.113891212 P:e:NW-O:ae:NW -0.3625199968 -0.547515062 -0.177524931 A:e:W-O:ae:NW -0.1976537529 -0.387718115 -0.007589391 A:e:NW-P:ae:NW -0.2142060451 -0.372032761 -0.056379329 O:e:NW-P:ae:NW -0.3162287493 -0.506293111 -0.126164387 A:ae:W-A:e:NW 0.2243102116 0.033740022 0.414880401 O:ae:W-A:e:NW 0.2339456239 0.069482191 0.398409057 124 p(adjusted) 0.0010777 0.0000000 0.0000049 0.0036518 0.0000001 0.0426057 0.0361567 0.0057302 0.0000013 0.0000000 0.0330867 0.0005787 0.0000038 0.0067523 0.0002161 P:ae:W-A:e:NW P:e:W-A:e:NW A:ae:W-O:e:NW O:ae:W-O:e:NW P:ae:W-O:e:NW P:e:W-O:e:NW A:ae:W-P:e:NW O:ae:W-P:e:NW P:ae:W-P:e:NW O:e:W-P:e:NW diff lwr upr 0.2237461547 0.062762192 0.384730118 0.2595865150 0.095486996 0.423686034 0.3263329158 0.108314140 0.544351691 0.3359683280 0.140357975 0.531578681 0.3257688589 0.133074734 0.518462983 0.3616092192 0.166304735 0.556913703 0.3151122801 0.101498378 0.528726182 0.3247476923 0.134059144 0.515436241 0.3145482232 0.126852357 0.502244089 0.2165388229 0.035797086 0.397280560 Three-way Interaction: StimType:Context:Word diff lwr upr A:k:NW-A:g:NW 0.374187855 2.459325e-01 0.50244325 O:k:NW-O:g:NW 0.584491000 3.823758e-01 0.78660619 P:k:NW-P:g:NW 0.387546284 1.982481e-01 0.57684451 O:k:NW-A:k:NW 0.229860320 2.570968e-02 0.43401096 P:k:W-P:k:NW 0.311225911 8.619120e-02 0.53626062 A:k:W-A:g:W 0.469940442 2.574537e-01 0.68242717 O:k:W-O:g:W 0.394700570 2.118778e-01 0.57752329 P:k:W-P:g:W 0.592903704 3.990040e-01 0.78680345 O:k:NW-A:g:NW 0.604048176 4.170381e-01 0.79105827 P:k:NW-A:g:NW 0.411672092 2.385940e-01 0.58475022 A:k:W-A:g:NW 0.526143923 3.391338e-01 0.71315401 O:k:W-A:g:NW 0.510774606 3.473138e-01 0.67423542 P:k:W-A:g:NW 0.722898003 5.484580e-01 0.89733801 A:k:NW-O:g:NW 0.354630680 2.052077e-01 0.50405365 P:k:NW-O:g:NW 0.392114916 2.028167e-01 0.58141314 A:k:W-O:g:NW 0.506586747 3.044716e-01 0.70870194 O:k:W-O:g:NW 0.491217430 3.106704e-01 0.67176442 P:k:W-O:g:NW 0.703340827 5.127966e-01 0.89388503 A:k:NW-P:g:NW 0.350062047 2.006391e-01 0.49948502 O:k:NW-P:g:NW 0.579922368 3.778072e-01 0.78203756 A:k:W-P:g:NW 0.502018115 2.999029e-01 0.70413331 O:k:W-P:g:NW 0.486648798 3.061018e-01 0.66719578 P:k:W-P:g:NW 0.698772195 5.082280e-01 0.88931640 A:g:W-A:k:NW -0.317984375 -4.811631e-01 -0.15480569 O:g:W-A:k:NW -0.258113819 -4.102787e-01 -0.10594891 125 p(adjusted) 0.0003557 0.0000160 0.0000664 0.0000014 0.0000023 0.0000001 0.0000942 0.0000018 0.0000030 0.0051767 p(adjusted) 0.0000000 0.0000000 0.0000000 0.0125820 0.0003944 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000021 diff lwr upr P:g:W-A:k:NW -0.244193556 -3.978726e-01 -0.09051453 P:k:W-A:k:NW 0.348710148 1.560082e-01 0.54141206 A:g:W-O:k:NW -0.547844695 -7.603314e-01 -0.33535797 O:g:W-O:k:NW -0.487974140 -6.921248e-01 -0.28382350 P:g:W-O:k:NW -0.474053877 -6.793356e-01 -0.26877220 A:g:W-P:k:NW -0.355468611 -5.558030e-01 -0.15513420 O:g:W-P:k:NW -0.295598056 -4.870680e-01 -0.10412808 P:g:W-P:k:NW -0.281677793 -4.743533e-01 -0.08900233 O:k:W-A:g:W 0.454571125 2.624845e-01 0.64665775 P:k:W-A:g:W 0.666694522 4.651824e-01 0.86820669 A:k:W-O:g:W 0.410069887 2.059192e-01 0.61422053 P:k:W-O:g:W 0.606823967 4.141221e-01 0.79952588 A:k:W-P:g:W 0.396149624 1.908679e-01 0.60143131 O:k:W-P:g:W 0.380780307 1.966955e-01 0.56486516 p(adjusted) 0.0000141 0.0000002 0.0000000 0.0000000 0.0000000 0.0000005 0.0000305 0.0001164 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 Three-way Interaction: Target:Context:Word diff lwr upr ae:k:NW-ae:g:NW 0.813473615 0.69408726 0.93285997 e:g:W-e:g:NW 0.132284944 0.03106842 0.23350147 e:k:NW-ae:k:NW -0.732678343 -0.87200774 -0.59334895 e:k:W-e:k:NW 0.308341633 0.16528850 0.45139477 ae:k:W-ae:g:W 0.752940758 0.60509587 0.90078564 e:k:W-e:g:W 0.266240844 0.13330356 0.39917813 e:k:W-ae:k:W -0.414596331 -0.57719467 -0.25199799 e:g:W-ae:g:NW 0.122896061 0.02368165 0.22211048 ae:k:W-ae:g:NW 0.803733236 0.66731697 0.94014950 e:k:W-ae:g:NW 0.389136905 0.26542506 0.51284875 ae:k:NW-e:g:NW 0.822862498 0.70180720 0.94391780 ae:k:W-e:g:NW 0.813122119 0.67524288 0.95100136 e:k:W-e:g:NW 0.398525788 0.27320258 0.52384899 ae:g:W-ae:k:NW -0.762681137 -0.89497555 -0.63038672 e:g:W-ae:k:NW -0.690577554 -0.81949924 -0.56165586 e:k:W-ae:k:NW -0.424336710 -0.57293659 -0.27573683 ae:k:W-e:k:NW 0.722937964 0.56876606 0.87710987 e:k:W-ae:g:W 0.338344427 0.20213381 0.47455504 ae:k:W-e:g:W 0.680837175 0.53600244 0.82567191 p(adjusted) 0.0000000 0.0019330 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0043508 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.000000 0.0000000 0.0000000 126 Three-way Interaction: Target:Context:Merge diff lwr e:g:M-ae:g:M 0.133066721 0.02475167 ae:k:M-ae:g:M 0.827769288 0.68915194 ae:g:S-ae:g:M 0.147509323 0.0404527 e:k:M-ae:k:M -0.581147576 -0.7423237 e:k:S-e:k:M 0.002207789 -0.15601072 ae:k:S-ae:g:S 0.693751124 0.55590512 e:k:S-e:g:S 0.195606755 0.05973215 e:k:S-ae:k:S -0.592430947 -0.75308748 upr 0.24138177 0.96638663 0.25456595 -0.41997146 0.1604263 0.83159712 0.33148136 -0.43177441 p(adjusted) 0.0048958 0 0.0007937 0 1 0 0.0003506 0 Post hoc testing of the Analysis of Variance of the normalized score at 50% of the Calibration [æ], of the non-[œ] stimuli, as influenced by StimType:Target:Context:Word:Merge Main Effect: Target diff lwr upr p(adjusted) e-ae 0.4762357 0.4137821 0.5386892 0 Main Effect: Context diff k-g -0.7919444 lwr -0.854333 upr -0.7295558 p(adjusted) 0 Main Effect: Word diff W-NW 0.1536749 lwr 0.08802312 upr 0.2193266 p(adjusted) 4.70E-06 Main Effect: Merge diff S-M -0.4023558 lwr -0.4657429 upr -0.3389688 p(adjusted) 0 Two-Way Interaction: Target:Context diff lwr e:g-ae:g 0.5607509 0.4759516 ae:k-ae:g -0.6968537 -0.8052571 e:k-ae:g -0.3489462 -0.4548806 ae:k-e:g -1.2576046 -1.3666626 e:k-e:g -0.9096971 -1.0163012 e:k-ae:k 0.3479075 0.2217139 upr 0.6455502 -0.5884503 -0.2430119 -1.1485466 -0.8030931 0.474101 p(adjusted) 0 0 0 0 0 0 127 Two-Way Interaction: StimType:Word diff lwr A:W-O:NW 0.22166738 0.043372 O:W-O:NW 0.25651325 0.09091575 P:W-O:NW 0.23292061 0.06405554 upr p(adjusted) 0.39996277 0.0053465 0.42211076 0.0001519 0.40178567 0.0012095 Two-Way Interaction: Target:Merge diff lwr e:M-ae:M 0.32621374 0.21522474 ae:S-ae:M -0.55017197 -0.66079732 e:S-ae:M 0.07398338 -0.03690087 ae:S-e:M -0.8763857 -0.98711605 e:S-e:M -0.25223036 -0.36321936 e:S-ae:S 0.62415534 0.51353 upr 0.4372027 -0.4395466 0.1848676 -0.7656554 -0.1412414 0.7347807 p(adjusted) 0 0 0.3157889 0 0 0 Two-Way Interaction: Context:Merge diff lwr k:M-g:M -0.9369165 -1.0470297 g:S-g:M -0.4923986 -0.5791384 k:S-g:M -1.1404365 -1.2503211 g:S-k:M 0.4445179 0.3345214 k:S-k:M -0.20352 -0.3325551 k:S-g:S -0.6480379 -0.7578056 upr -0.82680333 -0.40565876 -1.03055196 0.55451448 -0.07448489 -0.53827028 p(adjusted) 0 0 0 0 0.0003026 0 Three-Way Interaction: Stimtype:Target:Context diff lwr upr A:e:g-A:ae:g 0.63274487 0.448526785 0.816962956 O:e:g-A:ae:g 0.53754167 0.357083327 0.718000007 P:e:g-A:ae:g 0.55785722 0.373639136 0.742075308 A:ae:k-A:ae:g -0.752831 -0.987317131 -0.518344865 O:ae:k-A:ae:g -0.56737003 -0.803429303 -0.331310754 P:ae:k-A:ae:g -0.72620036 -0.962259639 -0.49014109 A:e:k-A:ae:g -0.27839965 -0.500613763 -0.056185532 O:e:k-A:ae:g -0.44255484 -0.677040973 -0.208068707 P:e:k-A:ae:g -0.29242819 -0.526914326 -0.057942061 A:e:g-O:ae:g 0.69162131 0.501116014 0.88212661 O:e:g-O:ae:g 0.59641811 0.409546018 0.783290198 P:e:g-O:ae:g 0.61673366 0.426228365 0.807238961 A:ae:k-O:ae:g -0.69395456 -0.933411676 -0.454497438 128 p(adjusted) 0 0 0 0 0 0 0.0025263 0.0000001 0.0027288 0 0 0 0 O:ae:k-O:ae:g P:ae:k-O:ae:g O:e:k-O:ae:g A:e:g-P:ae:g O:e:g-P:ae:g P:e:g-P:ae:g A:ae:k-P:ae:g O:ae:k-P:ae:g P:ae:k-P:ae:g A:e:k-P:ae:g O:e:k-P:ae:g P:e:k-P:ae:g A:ae:k-A:e:g O:ae:k-A:e:g P:ae:k-A:e:g A:e:k-A:e:g O:e:k-A:e:g P:e:k-A:e:g A:ae:k-O:e:g O:ae:k-O:e:g P:ae:k-O:e:g A:e:k-O:e:g O:e:k-O:e:g P:e:k-O:e:g A:ae:k-P:e:g O:ae:k-P:e:g P:ae:k-P:e:g A:e:k-P:e:g O:e:k-P:e:g P:e:k-P:e:g A:e:k-A:ae:k O:e:k-A:ae:k P:e:k-A:ae:k A:e:k-O:ae:k A:e:k-P:ae:k O:e:k-P:ae:k P:e:k-P:ae:k diff -0.50849359 -0.66732392 -0.3836784 0.53256953 0.43736633 0.45768188 -0.85300634 -0.66754537 -0.8263757 -0.37857499 -0.54273018 -0.39260353 -1.38557587 -1.2001149 -1.35894524 -0.91114452 -1.07529971 -0.92517306 -1.29037266 -1.1049117 -1.26374203 -0.81594131 -0.98009651 -0.82996986 -1.31068822 -1.12522725 -1.28405759 -0.83625687 -1.00041206 -0.85028542 0.47443135 0.31027616 0.4604028 0.28897038 0.44780072 0.28364552 0.43377217 lwr -0.749491402 -0.908321738 -0.623135518 0.344022013 0.252490488 0.269134365 -1.090908859 -0.906998588 -1.065828923 -0.604391232 -0.780632701 -0.630506055 -1.625195204 -1.441273893 -1.600104229 -1.138768753 -1.314919045 -1.164792399 -1.527113736 -1.343211023 -1.502041359 -1.040533615 -1.216837578 -1.066710932 -1.550307555 -1.366386244 -1.52521658 -1.063881104 -1.240031397 -1.08990475 0.204504999 0.030160345 0.180286991 0.017676315 0.176506651 0.002211512 0.152338158 upr -0.267495773 -0.426326109 -0.14422128 0.721117052 0.622242169 0.646229403 -0.615103813 -0.428092146 -0.586922482 -0.15275874 -0.304827655 -0.154701009 -1.145956534 -0.958955906 -1.117786242 -0.683520284 -0.835680376 -0.68555373 -1.053631593 -0.866612368 -1.025442704 -0.591349014 -0.743355435 -0.593228789 -1.071068885 -0.884068257 -1.042898593 -0.608632635 -0.760792727 -0.610666081 0.744357702 0.590391972 0.740518618 0.560264447 0.719094782 0.565079537 0.715206184 129 p(adjusted) 0 0 0.0000113 0 0 0 0 0 0 0.000003 0 0.0000048 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.0000007 0.0155924 0.0000055 0.0251952 0.0000048 0.0461858 0.000032 Three-Way Interaction: Stimtype:Target:Word diff lwr A:e:NW-A:ae:NW 0.544280113 0.32031333 O:e:NW-A:ae:NW 0.428901487 0.19137891 P:e:NW-A:ae:NW 0.407562932 0.17757201 A:e:W-A:ae:NW 0.535302726 0.29778015 O:e:W-A:ae:NW 0.505054717 0.28141465 P:e:W-A:ae:NW 0.636877451 0.39889493 A:e:NW-O:ae:NW 0.710418952 0.46415479 O:e:NW-O:ae:NW 0.595040326 0.33638632 P:e:NW-O:ae:NW 0.573701772 0.32194645 O:ae:W-O:ae:NW 0.374010823 0.11450738 A:e:W-O:ae:NW 0.701441565 0.44278756 O:e:W-O:ae:NW 0.671193557 0.42522649 P:e:W-O:ae:NW 0.80301629 0.54393986 A:e:NW-P:ae:NW 0.465682581 0.21941841 O:e:NW-P:ae:NW 0.350303954 0.09164995 P:e:NW-P:ae:NW 0.3289654 0.07721008 A:e:W-P:ae:NW 0.456705193 0.19805119 O:e:W-P:ae:NW 0.426457185 0.18049012 P:e:W-P:ae:NW 0.558279919 0.29920348 A:ae:W-A:e:NW -0.50375982 -0.78971953 O:ae:W-A:e:NW -0.336408129 -0.59058786 P:ae:W-A:e:NW -0.468612141 -0.7186325 A:ae:W-O:e:NW -0.388381194 -0.68507769 P:ae:W-O:e:NW -0.353233515 -0.6154663 A:ae:W-P:e:NW -0.36704264 -0.65774464 P:ae:W-P:e:NW -0.331894961 -0.58732574 A:e:W-A:ae:W 0.494782433 0.19808594 O:e:W-A:ae:W 0.464534424 0.17883053 P:e:W-A:ae:W 0.596357158 0.29929233 A:e:W-O:ae:W 0.327430742 0.06122933 O:e:W-O:ae:W 0.297182733 0.04329084 P:e:W-O:ae:W 0.429005467 0.16239359 A:e:W-P:ae:W 0.459634754 0.19740197 O:e:W-P:ae:W 0.429386746 0.17965902 P:e:W-P:ae:W 0.56120948 0.29856002 130 upr 0.7682469 0.66642407 0.63755385 0.77282531 0.72869479 0.87485997 0.95668312 0.85369433 0.82545709 0.63351427 0.96009557 0.91716063 1.06209273 0.71194675 0.60895796 0.58072072 0.7153592 0.67242425 0.81735635 -0.21780011 -0.0822284 -0.21859178 -0.0916847 -0.09100073 -0.07634063 -0.07646419 0.79147892 0.75023831 0.89342199 0.59363215 0.55107463 0.69561735 0.72186754 0.67911447 0.82385894 p(adjusted) 0 0.0000003 0.0000005 0 0 0 0 0 0 0.0001621 0 0 0 0 0.0006057 0.0012025 0.0000006 0.000001 0 0.0000006 0.0009469 0.0000001 0.0011636 0.0006741 0.0022139 0.0013332 0.0000035 0.0000075 0 0.0034242 0.007334 0.00001 0.0000007 0.0000014 0 Three-Way Interaction: Target:Context:Merge diff lwr e:g:M-ae:g:M 0.32970864 0.19783431 ae:k:M-ae:g:M -0.97780128 -1.14656887 e:k:M-ae:g:M -0.58686229 -0.75142909 ae:g:S-ae:g:M -0.71746302 -0.84780521 ae:k:S-ae:g:M -1.13526636 -1.30329798 e:k:S-ae:g:M -0.83137458 -0.99594138 ae:k:M-e:g:M -1.30750991 -1.47726274 e:k:M-e:g:M -0.91657093 -1.08214796 ae:g:S-e:g:M -1.04717166 -1.17878704 e:g:S-e:g:M -0.25754246 -0.3904903 ae:k:S-e:g:M -1.46497499 -1.63399614 e:k:S-e:g:M -1.16108322 -1.32666025 e:k:M-ae:k:M 0.39093899 0.19470593 ae:g:S-ae:k:M 0.26033825 0.09177292 e:g:S-ae:k:M 1.04996745 0.88035971 e:g:S-e:k:M 0.65902847 0.49360018 ae:k:S-e:k:M -0.54840406 -0.74400452 e:k:S-e:k:M -0.24451229 -0.43714444 e:g:S-ae:g:S 0.7896292 0.65820099 ae:k:S-ae:g:S -0.41780333 -0.5856318 ae:k:S-e:g:S -1.20743253 -1.37630797 e:k:S-e:g:S -0.90354076 -1.06896905 e:k:S-ae:k:S 0.30389177 0.10829132 131 upr 0.46158296 -0.80903369 -0.4222955 -0.58712083 -0.96723474 -0.66680779 -1.13775709 -0.7509939 -0.91555627 -0.12459462 -1.29595384 -0.99550619 0.58717204 0.42890358 1.2195752 0.82445676 -0.35280361 -0.05188015 0.92105741 -0.24997486 -1.03855709 -0.73811247 0.49949223 p(adjusted) 0 0 0 0 0 0 0 0 0 0.0000001 0 0 0 0.0000806 0 0 0 0.0030457 0 0 0 0 0.0000705 Post hoc testing of the Analysis of Variance of the normalized score at 50% of the Calibration [ɛ], of the auditory stimuli, including [œ], as influenced by: Target:Context:Word:Merge:Label Label Notation A [æ] E [ɛ] se Switched ɛ sa Switched æ _ Neutral Fa Neutral (word forms with æ) Fe Neutral (word forms with ɛ) Fb Both (word forms with both æ and ɛ) Main Effect: Target diff ae-_ -0.12822532 e-_ -0.20495711 e-ae -0.07673178 lwr -0.2001949 -0.2748744 -0.1478432 upr -0.056255782 -0.135039766 -0.005620351 p(adjusted) 0.0000913 0.0000000 0.0307692 Main Effect: Context diff k-g 0.2422242 lwr 0.1907012 upr 0.2937473 p(adjusted) Main Effect: Word diff W-NW 0.02012197 lwr upr -0.04560166 0.0858456 Main Effect: Label diff e-a -0.11363199 fa-a 0.06341507 fb-a -0.01929093 fe-a 0.02075313 oe-a 0.20259579 sa-a -0.04245673 se-a -0.04017165 fa-e 0.17704706 fb-e 0.09434105 fe-e 0.13438511 lwr -0.22326455 -0.15294431 -0.16628789 -0.23641199 0.08772616 -0.19583393 -0.17017775 -0.03985025 -0.05344652 -0.12323274 upr -0.003999417 0.279774454 0.127706021 0.277918240 0.317465415 0.110920474 0.089834452 0.393944369 0.242128623 0.392002963 132 0 p(adjusted) 0.5482414 p(adjusted) 0.0359183 0.9870240 0.9999267 0.9999974 0.0000028 0.9907708 0.9823304 0.2058434 0.5251444 0.7605303 oe-e sa-e se-e fb-fa fe-fa oe-fa sa-fa se-fa fe-fb oe-fb sa-fb se-fb oe-fe 0.31622777 0.07117526 0.07346034 -0.08270600 -0.04266194 0.13918071 -0.10587180 -0.10358672 0.04004406 0.22188672 -0.02316580 -0.02088072 0.18184266 diff sa-fe -0.06320985 se-fe -0.06092477 sa-oe -0.24505251 se-oe -0.24276743 se-sa 0.00228508 0.20034812 -0.08295984 -0.05743904 -0.32068342 -0.36071703 -0.08041021 -0.34784227 -0.33146109 -0.23555667 0.07017351 -0.20577585 -0.18435381 -0.07804710 lwr -0.34226580 -0.32785031 -0.40295552 -0.37808328 -0.16694823 0.432107420 0.225310352 0.204359716 0.155271407 0.275393144 0.358771642 0.136098674 0.124287654 0.315644787 0.373599926 0.159444259 0.142592377 0.441732421 upr 0.215846095 0.206000764 -0.087149509 -0.107451585 0.171518391 Two-Way Interaction: Target:Context diff lwr upr ae:g-_:g -0.37849900 -0.47141182 -0.28558617 e:g-_:g -0.25401402 -0.34781315 -0.16021490 _:k-_:g -0.07674478 -0.19200335 0.03851380 ae:k-_:g 0.52559806 0.39472292 0.65647320 e:k-_:g -0.18815686 -0.29922688 -0.07708684 e:g-ae:g 0.12448497 0.03197655 0.21699339 _:k-ae:g 0.30175422 0.18754358 0.41596486 ae:k-ae:g 0.90409706 0.77414386 1.03405026 e:k-ae:g 0.19034214 0.08035995 0.30032432 _:k-e:g 0.17726925 0.06233642 0.29220207 ae:k-e:g 0.77961209 0.64902374 0.91020044 e:k-e:g 0.06585717 -0.04487478 0.17658911 ae:k-_:k 0.60234284 0.45557915 0.74910653 e:k-_:k -0.11141208 -0.24082548 0.01800131 e:k-ae:k -0.71375492 -0.85725262 -0.57025722 Two-way Interaction: Context:Label 133 0.0000000 0.8568124 0.6852222 0.9656986 0.9999150 0.5346786 0.8882971 0.8666374 0.9998543 0.0002591 0.9999417 0.9999390 0.4000006 p(adjusted) 0.9973283 0.9971964 0.0000733 0.0000017 1.0000000 p(adjusted) 0.0000000 0.0000000 0.4025443 0.0000000 0.0000217 0.0017819 0.0000000 0.0000000 0.0000129 0.0001672 0.0000000 0.5339674 0.0000000 0.1379672 0.0000000 k:a-g:a g:e-g:a k:e-g:a g:fa-g:a k:fa-g:a g:fb-g:a k:fb-g:a g:fe-g:a k:fe-g:a g:oe-g:a k:oe-g:a g:sa-g:a k:sa-g:a g:se-g:a k:se-g:a g:e-k:a k:e-k:a g:fa-k:a k:fa-k:a g:fb-k:a k:fb-k:a g:fe-k:a k:fe-k:a g:oe-k:a k:oe-k:a g:sa-k:a k:sa-k:a g:se-k:a k:se-k:a k:e-g:e g:fa-g:e k:fa-g:e g:fb-g:e k:fb-g:e g:fe-g:e k:fe-g:e g:oe-g:e diff lwr upr 0.855385614 0.685726721 1.02504451 0.144831536 0.012151093 0.27751198 0.103722739 -0.056377510 0.26382299 0.285372119 0.020935571 0.54980867 0.350464311 -0.007602672 0.70853129 0.215291399 0.032729201 0.39785360 0.240453836 0.017404480 0.46350319 0.264816956 0.003641891 0.52599202 NA NA NA 0.505862187 0.366353114 0.64537126 0.338403485 0.168744592 0.50806238 0.030580943 -0.138505045 0.19966693 diff lwr upr 1.268005507 0.900213807 1.63579721 0.111996043 -0.047668405 0.27166049 0.363434089 0.167953552 0.55891463 -0.710554078 -0.883650894 -0.53745726 -0.751662875 -0.946574388 -0.55675136 -0.570013495 -0.856867283 -0.28315971 -0.504921303 -0.879848459 -0.12999415 -0.640094215 -0.853840242 -0.42634819 -0.614931778 -0.864149400 -0.36571416 -0.590568658 -0.874418658 -0.30671866 NA NA NA -0.349523427 -0.527908374 -0.17113848 -0.516982129 -0.719818388 -0.31414587 -0.824804671 -1.027161977 -0.62244737 0.412619893 0.028394513 0.79684527 -0.743389571 -0.937943275 -0.54883587 -0.491951525 -0.716831601 -0.26707145 -0.041108796 -0.204847789 0.12263020 0.140540584 -0.126114729 0.40719590 0.205632775 -0.154075901 0.56534145 0.070459864 -0.115301607 0.25622133 0.095622301 -0.130053092 0.32129769 0.119985421 -0.143435882 0.38340672 NA NA NA 0.361030651 0.217360372 0.50470093 134 p(adjusted) 0.0000000 0.0172482 0.6805157 0.0200833 0.0627209 0.0055047 0.0203637 0.0429024 NA 0.0000000 0.0000000 0.9999996 p(adjusted) 0.0000000 0.5423601 0.0000000 0.0000000 0.0000000 0.0000000 0.0004601 0.0000000 0.0000000 0.0000000 NA 0.0000000 0.0000000 0.0000000 0.0214081 0.0000000 0.0000000 0.9999705 0.9133891 0.8458736 0.9958620 0.9871081 0.9744457 NA 0.0000000 k:oe-g:e g:sa-g:e k:sa-g:e g:se-g:e k:se-g:e g:fa-k:e k:fa-k:e g:fb-k:e k:fb-k:e g:fe-k:e k:fe-k:e g:oe-k:e k:oe-k:e g:sa-k:e k:sa-k:e g:se-k:e k:se-k:e k:fa-g:fa g:fb-g:fa k:fb-g:fa g:fe-g:fa k:fe-g:fa g:oe-g:fa k:oe-g:fa g:sa-g:fa k:sa-g:fa g:se-g:fa k:se-g:fa g:fb-k:fa k:fb-k:fa g:fe-k:fa k:fe-k:fa g:oe-k:fa k:oe-k:fa g:sa-k:fa k:sa-k:fa g:se-k:fa k:se-k:fa 0.193571950 0.020475134 0.36666877 -0.114250593 -0.286785920 0.05828473 1.123173971 0.753783795 1.49256415 -0.032835492 -0.196148394 0.13047741 0.218602553 0.020130872 0.41707423 0.181649380 -0.099656584 0.46295534 0.246741572 -0.123958201 0.61744134 0.111568660 -0.094672253 0.31780957 0.136731097 -0.106080297 0.37954249 0.161094217 -0.117148074 0.43933651 NA NA NA 0.402139448 0.232819811 0.57145908 0.234680746 0.039769233 0.42959226 diff lwr upr -0.073141796 -0.267554834 0.12127124 1.164282768 0.784181344 1.54438419 0.008273304 -0.178003624 0.19455023 0.259711350 0.041952313 0.47747039 0.065092192 -0.361185792 0.49137018 -0.070080720 -0.364749877 0.22458844 -0.044918283 -0.366246730 0.27641016 -0.020555163 -0.369429597 0.32831927 NA NA NA 0.220490068 -0.049627939 0.49060807 0.053031366 -0.233822422 0.33988515 -0.254791177 -0.541306494 0.03172414 0.982633387 0.548154746 1.41711203 -0.173376076 -0.454434240 0.10768209 0.078061970 -0.224780557 0.38090450 -0.135172911 -0.516112786 0.24576696 -0.110010475 -0.511927440 0.29190649 -0.085647354 -0.509909828 0.33861512 NA NA NA 0.155397876 -0.206885181 0.51768093 -0.012060826 -0.386987981 0.36286633 -0.319883368 -0.694551626 0.05478489 0.917541196 0.420492943 1.41458945 -0.238468267 -0.608980032 0.13204350 0.012969778 -0.374327101 0.40026666 135 0.0123314 0.6444423 0.0000000 0.9999985 0.0152602 0.6857761 0.6356857 0.8949202 0.8606684 0.8325497 NA 0.0000000 0.0039069 p(adjusted) 0.9962093 0.0000000 1.0000000 0.0045788 1.0000000 0.9999855 1.0000000 1.0000000 NA 0.2692144 0.9999995 0.1490717 0.0000000 0.7541594 0.9999585 0.9980060 0.9999107 0.9999984 NA 0.9854737 1.0000000 0.2014160 0.0000000 0.6909905 1.0000000 k:fb-g:fb g:fe-g:fb k:fe-g:fb g:oe-g:fb k:oe-g:fb g:sa-g:fb k:sa-g:fb g:se-g:fb k:se-g:fb g:fe-k:fb k:fe-k:fb g:oe-k:fb k:oe-k:fb g:sa-k:fb k:sa-k:fb g:se-k:fb k:se-k:fb k:fe-g:fe g:oe-g:fe k:oe-g:fe g:sa-g:fe k:sa-g:fe g:se-g:fe k:se-g:fe g:oe-k:fe k:oe-k:fe g:sa-k:fe k:sa-k:fe g:se-k:fe k:se-k:fe k:oe-g:oe g:sa-g:oe k:sa-g:oe g:se-g:oe k:se-g:oe g:sa-k:oe k:sa-k:oe g:se-k:oe 0.025162437 -0.233012387 0.28333726 0.049525557 -0.242220290 0.34127140 NA NA NA 0.290570787 0.099872057 0.48126952 0.123112086 -0.090633942 0.33685811 -0.184710457 -0.398002031 0.02858112 1.052714107 0.662619300 1.44280891 -0.103295356 -0.309198150 0.10260744 0.148142690 -0.086624993 0.38291037 0.024363120 -0.294286685 0.34301293 NA NA NA 0.265408351 0.035651817 0.49516488 0.097949649 -0.151267973 0.34716727 diff lwr upr -0.209872893 -0.458700855 0.03895507 1.027551671 0.616947217 1.43815612 -0.128457793 -0.370982058 0.11406647 0.122980253 -0.144485490 0.39044600 NA NA NA 0.241045230 -0.025880723 0.50797118 0.073586529 -0.210263472 0.35743653 -0.234236014 -0.517743957 0.04927193 1.003188550 0.570687201 1.43568990 -0.152820913 -0.430812674 0.12517085 0.098617133 -0.201381740 0.39861601 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA -0.167458702 -0.345843649 0.01092625 -0.475281244 -0.653121399 -0.29744109 0.762143320 0.390245775 1.13404086 -0.393866143 -0.562773768 -0.22495852 -0.142428098 -0.345528294 0.06067210 -0.307822542 -0.510179848 -0.10546524 0.929602022 0.545376642 1.31382740 -0.226407442 -0.420961147 -0.03185374 136 1.0000000 0.9999999 NA 0.0000226 0.8380904 0.1821072 0.0000000 0.9411984 0.7214163 1.0000000 NA 0.0076211 0.9939749 p(adjusted) 0.2188087 0.0000000 0.9100891 0.9721662 NA 0.1317688 0.9999553 0.2500120 0.0000000 0.8820744 0.9991684 NA NA NA NA NA NA 0.0940820 0.0000000 0.0000000 0.0000000 0.5428217 0.0000237 0.0000000 0.0067966 k:se-k:oe k:sa-g:sa g:se-g:sa k:se-g:sa g:se-k:sa k:se-k:sa k:se-g:se 0.025030604 -0.199849472 0.24991068 1.237424564 0.853451813 1.62139732 0.081415101 -0.112639210 0.27546941 0.332853146 0.108404979 0.55730131 -1.156009463 -1.535927531 -0.77609140 -0.904571418 -1.300876405 -0.50826643 0.251438046 0.033999216 0.46887688 1.0000000 0.0000000 0.9883029 0.0000468 0.0000000 0.0000000 0.0075012 Post hoc testing of the four-way interaction Order:Target:Context:Merge in the Analysis of Variance of the normalized score at 50% of the Calibration [ɛ], of the auditory stimuli, including [œ]. diff lwr upr p AOP:g:M:ae-AOP:g:M:_ 0.490466924 0.790176954 0.190756894 0 AOP:g:M:e-AOP:g:M:ae 0.301672673 0.001962643 0.601382703 0.0452333 AOP:k:M:ae-AOP:g:M:ae 1.154155291 0.739913367 1.568397214 0 AOP:k:M:ae-AOP:k:M:_ 0.757237735 0.287130591 1.22734488 0.0000001 APO:g:M:ae-AOP:g:M:ae 0.44491765 0.030675726 0.859159573 0.0149793 APO:g:S:ae-APO:g:M:ae 0.579653237 1.01813999 0.141166484 0.0001009 APO:g:S:e-APO:g:M:e 0.465135026 0.924148452 0.0061216 0.0407156 APO:g:S:ae-APO:g:S:_ 0.680367905 1.0428385 0.31789731 0 APO:g:S:e-APO:g:S:_ 0.371426256 0.7406348 0.002217712 0.045612 APO:k:S:ae-APO:g:S:ae 0.881121512 0.374800622 1.387442402 0 APO:k:S:e-APO:k:S:ae 0.604321387 1.170405351 0.038237422 0.0167543 OAP:g:M:_-APO:g:M:_ 0.646396977 1.172003808 0.120790146 0.0007551 OAP:g:S:ae-APO:g:S:ae 0.336843436 0.006762599 0.666924272 0.0363893 OAP:k:M:_-APO:k:M:_ 0.810559368 1.544288045 0.076830691 0.0087067 OAP:g:S:_-OAP:g:M:_ 0.439348576 0.013908104 0.864789049 0.029938 OAP:g:S:ae-OAP:g:S:_ 0.458838748 0.760953462 0.156724034 0.0000008 OAP:g:S:e-OAP:g:S:_ 0.417528029 0.719642743 0.115413315 0.0000249 OAP:k:M:ae-OAP:g:M:ae 0.808673419 0.044984399 1.572362439 0.0194083 OAP:k:M:ae-OAP:k:M:_ 0.85979061 0.012553712 1.707027509 0.0397844 OAP:k:M:e-OAP:k:M:ae 0.880283446 1.712953581 0.04761331 0.0199806 OAP:k:S:_-OAP:k:M:_ 0.688092294 0.089005338 1.287179251 0.0038481 OAP:k:S:e-OAP:k:S:ae 0.748180475 1.213109289 0.28325166 0.0000001 OPA:g:M:e-APO:g:M:e 0.569690759 1.09529759 0.044083928 0.012623 OPA:k:M:ae-OPA:g:M:ae 1.358487066 0.599005731 2.117968401 0 OPA:k:S:ae-OPA:g:S:ae 0.882003138 0.16149588 1.602510395 0.0008499 OPA:k:M:ae-OPA:k:M:_ 0.85535398 0.008117082 1.702590879 0.0431638 137 PAO:g:M:_-APO:g:M:_ PAO:g:M:ae-APO:g:M:ae PAO:g:M:e-APO:g:M:e PAO:g:S:ae-APO:g:S:ae PAO:k:M:_-APO:k:M:_ PAO:g:S:ae-OAP:g:S:ae PAO:g:S:ae-OPA:g:S:ae PAO:g:M:ae-PAO:g:M:_ PAO:g:S:ae-PAO:g:M:ae PAO:g:S:e-PAO:g:M:e PAO:k:M:ae-PAO:g:M:ae PAO:k:M:ae-PAO:k:M:_ PAO:k:S:_-PAO:k:M:_ PAO:k:S:e-PAO:k:S:ae POA:g:M:_-APO:g:M:_ POA:g:S:_-APO:g:S:_ POA:g:S:_-OAP:g:S:_ POA:k:S:_-OAP:k:S:_ POA:g:S:_-PAO:g:S:_ POA:g:S:ae-PAO:g:S:ae POA:g:S:e-PAO:g:S:e POA:k:M:ae-POA:k:M:_ POA:k:M:e-POA:k:M:ae POA:k:S:ae-POA:k:S:_ diff 0.551113839 0.498708121 0.57401031 0.941579533 0.76119761 0.604736098 0.788181213 0.370671631 0.860634418 0.514798576 0.789728779 0.658476272 0.644054623 0.824476003 0.53525049 0.389873117 0.505187396 0.610233927 0.620368305 0.847885144 0.660228714 1.157508358 1.066075355 0.714142782 lwr 1.010127264 0.938973733 1.031140291 0.50309278 1.402174976 0.188751039 0.275570382 0.737650668 0.420368806 0.065483359 0.267508454 0.067939416 0.003077257 1.617728386 1.060857321 0.759081661 0.844937863 1.091273835 1.08787665 1.288150755 1.109543931 0.348978735 1.859327738 0.136264135 upr 0.092100413 0.05844251 0.116880329 1.380066287 0.120220244 1.020721157 1.300792045 0.003692593 1.300900029 0.964113793 1.311949104 1.249013129 1.285031989 0.031223619 0.009643659 0.020664573 0.165436929 0.129194019 0.15285996 0.407619532 0.210913496 1.966037981 0.272822972 1.292021429 p(adjusted) 0.0013774 0.005194 0.0004365 0 0.0017934 0.0000043 0.0000005 0.0428429 0 0.0040617 0.000001 0.0072129 0.0464649 0.0268604 0.0376394 0.0203923 0.0000019 0.0003305 0.00009 0 0.0000029 0.0000075 0.000061 0.0006659 Post hoc testing of the Analysis of Variance of the normalized score at 50% of the Calibration [æ], of the auditory stimuli, including [œ], as influenced by StimType:Target:Context:Word:Label Main effect of: Target diff lwr ae-_ -0.8039317 -0.8982424 e-_ -0.3861260 -0.4777474 e-ae 0.4178058 0.3246196 upr -0.7096211 -0.2945046 0.5109919 138 p(adjusted) 0 0 0 Main effect of: Context diff lwr k-g -0.4772643 -0.5500365 upr -0.4044921 p(adjusted) 0 Main effect of: Merge S-M -0.3180181 -0.3873292 -0.248707 0 Main effect of: Label diff lwr upr p(adjusted) e-a 0.5283996020 0.385775882 0.671023323 0.0000000 fa-a 0.7284952328 0.447027959 1.009962506 0.0000000 fb-a 0.7280385693 0.536806595 0.919270544 0.0000000 fe-a 0.8315440676 0.496991609 1.166096526 0.0000000 oe-a 0.9969859431 0.847549201 1.146422685 0.0000000 sa-a 0.3341268313 0.134594635 0.533659028 0.0000115 se-a 0.4531838763 0.284055724 0.622312029 0.0000000 fa-e 0.2000956307 -0.082071448 0.482262709 0.3817436 fb-e 0.1996389672 0.007378462 0.391899473 0.0352905 fe-e 0.3031444656 -0.031996969 0.638285901 0.1100773 oe-e 0.4685863411 0.317835640 0.619337042 0.0000000 sa-e -0.1942727707 -0.394790928 0.006245387 0.0654855 se-e -0.0752157257 -0.245505966 0.095074515 0.8832303 fb-fa -0.0004566635 -0.310047365 0.309134038 1.0000000 fe-fa 0.1030488349 -0.310716898 0.516814568 0.9951601 oe-fa 0.2684907104 -0.017180560 0.554161980 0.0832901 sa-fa -0.3943684015 -0.709153776 -0.079583027 0.0037120 se-fa -0.2753113564 -0.571758766 0.021136053 0.0911116 fe-fb 0.1035054983 -0.255030307 0.462041304 0.9881377 oe-fb 0.2689473739 0.071579909 0.466314839 0.0009712 sa-fb -0.3939117380 -0.631473675 -0.156349801 0.0000149 se-fb -0.2748546929 -0.487520883 -0.062188503 0.0023213 oe-fe 0.1654418755 -0.172655146 0.503538897 0.8155683 sa-fe -0.4974172363 -0.860448025 -0.134386448 0.0008823 se-fe -0.3783601913 -0.725610228 -0.031110155 0.0215842 sa-oe -0.6628591118 -0.868279037 -0.457439186 0.0000000 se-oe -0.5438020668 -0.719837802 -0.367766331 0.0000000 se-sa 0.1190570450 -0.101102756 0.339216846 0.7250363 139 Two-Way Interactions: Target:Context diff lwr ae:g-_:g -0.65263988 -0.7752504 e:g-_:g -0.18187395 -0.3056540 _:k-_:g -0.14055326 -0.2926518 ae:k-_:g -1.45044380 -1.6231505 e:k-_:g -0.88803248 -1.0346037 e:g-ae:g 0.47076592 0.3486891 _:k-ae:g 0.51208661 0.3613709 ae:k-ae:g -0.79780392 -0.9692940 e:k-ae:g -0.23539260 -0.3805283 _:k-e:g 0.04132069 -0.1103480 ae:k-e:g -1.26856985 -1.4408980 e:k-e:g -0.70615852 -0.8522836 ae:k-_:k -1.30989054 -1.5035642 e:k-_:k -0.74747921 -0.9182569 e:k-ae:k 0.56241132 0.3730476 upr -0.53002940 -0.05809389 0.01154531 -1.27773714 -0.74146124 0.59284273 0.66280231 -0.62631388 -0.09025690 0.19298940 -1.09624165 -0.56003342 -1.11621689 -0.57670152 0.75177508 p(adjusted) 0.0000000 0.0004179 0.0891918 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000587 0.9713480 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 Two-Way Interactions: Context:Word diff lwr k:NW-g:NW -0.37544898 -0.4786742 g:W-g:NW -0.01663223 -0.1591346 k:W-g:NW -0.96104297 -1.1536794 g:W-k:NW 0.35881675 0.2028088 k:W-k:NW -0.58559400 -0.7884248 k:W-g:W -0.94441075 -1.1697812 upr -0.2722237 0.1258701 -0.7684066 0.5148247 -0.3827631 -0.7190403 p(adjusted) 0.0000000 0.9906036 0.0000000 0.0000000 0.0000000 0.0000000 Two-Way Interactions: Target:Merge diff lwr ae:M-_:M -0.64919574 -0.805668801 e:M-_:M -0.31504210 -0.467032290 _:S-_:M -0.17038388 -0.324353898 ae:S-_:M -1.12742726 -1.283735590 e:S-_:M -0.62637835 -0.778233876 e:M-ae:M 0.33415364 0.179335827 _:S-ae:M 0.47881186 0.322049925 ae:S-ae:M -0.47823153 -0.637290728 e:S-ae:M 0.02281739 -0.131868221 _:S-e:M 0.14465822 -0.007629346 upr -0.49272267 -0.16305191 -0.01641386 -0.97111894 -0.47452282 0.48897145 0.63557379 -0.31917233 0.17750301 0.29694578 p(adjusted) 0.0000000 0.0000001 0.0201332 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.9983196 0.0737450 140 diff ae:S-e:M e:S-e:M ae:S-_:S e:S-_:S e:S-ae:S lwr -0.81238517 -0.31133625 -0.95704338 -0.45599447 0.50104892 upr -0.967036478 -0.461485632 -1.113640882 -0.608147631 0.346529950 Two-Way Interactions: Context:Merge diff lwr k:M-g:M -0.6093846 -0.74003391 g:S-g:M -0.4016090 -0.50533492 k:S-g:M -0.7474072 -0.87823815 g:S-k:M 0.2077756 0.07716471 k:S-k:M -0.1380226 -0.29104834 k:S-g:S -0.3457982 -0.47659085 p(adjusted) -0.65773386 0.0000000 -0.16118686 0.0000001 -0.80044589 0.0000000 -0.30384130 0.0000000 0.65556789 0.0000000 upr -0.47873531 -0.29788303 -0.61657624 0.33838656 0.01500316 -0.21500559 Two-Way Interactions: Context:Label diff lwr upr k:a-g:a -0.75720025 -0.980542802 -0.53385770 g:e-g:a 0.62833022 0.453666882 0.80299356 k:e-g:a -0.27374058 -0.484499930 -0.06298123 g:fa-g:a 0.45819223 0.110082376 0.80630209 k:fa-g:a 0.61794016 0.146573223 1.08930710 g:fb-g:a 0.56264525 0.322316520 0.80297399 k:fb-g:a 0.42821290 0.134586015 0.72183978 g:fe-g:a 0.61549508 0.271678709 0.95931146 k:fe-g:a NA NA NA g:oe-g:a 0.85512086 0.671468168 1.03877355 k:oe-g:a 0.64528640 0.421943846 0.86862895 g:sa-g:a 0.22935504 0.006766674 0.45194341 k:sa-g:a -0.57576819 -1.059936951 -0.09159942 g:se-g:a 0.36796155 0.157775902 0.57814721 k:se-g:a 0.01000527 -0.247329441 0.26733998 g:e-k:a 1.38553047 1.157662164 1.61339878 k:e-k:a 0.48345967 0.226874038 0.74004530 g:fa-k:a 1.21539249 0.837772098 1.59301287 k:fa-k:a 1.37514041 0.881578386 1.86870244 g:fb-k:a 1.31984551 1.038465719 1.60122529 k:fb-k:a 1.18541315 0.857337807 1.51348849 141 p(adjusted) 0.0000000 0.0000000 0.0000000 0.0002633 0.0939978 0.0000000 p(adjusted) 0.0000000 0.0000000 0.0009722 0.0007434 0.0008067 0.0000000 0.0000719 0.0000001 NA 0.0000000 0.0000000 0.0356600 0.0048037 0.0000003 1.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 g:fe-k:a k:fe-k:a g:oe-k:a k:oe-k:a g:sa-k:a k:sa-k:a g:se-k:a k:se-k:a k:e-g:e g:fa-g:e k:fa-g:e g:fb-g:e k:fb-g:e g:fe-g:e k:fe-g:e g:oe-g:e k:oe-g:e g:sa-g:e k:sa-g:e g:se-g:e k:se-g:e g:fa-k:e k:fa-k:e g:fb-k:e k:fb-k:e g:fe-k:e k:fe-k:e g:oe-k:e k:oe-k:e g:sa-k:e k:sa-k:e g:se-k:e k:se-k:e k:fa-g:fa g:fb-g:fa k:fb-g:fa g:fe-g:fa k:fe-g:fa diff lwr upr 1.37269533 0.999029196 1.74636147 NA NA NA 1.61232111 1.377491397 1.84715082 1.40248665 1.135468711 1.66950458 0.98655529 0.720167859 1.25294272 0.18143206 -0.324370343 0.68723447 1.12516181 0.869047198 1.38127641 0.76720552 0.471168636 1.06324240 -0.90207080 -1.117620267 -0.68652133 -0.17013798 -0.521168671 0.18089270 -0.01039006 -0.483918157 0.46313805 -0.06568496 -0.310225289 0.17885536 -0.20011732 -0.497201175 0.09696654 -0.01283514 -0.359608504 0.33393823 NA NA NA 0.22679064 0.037660049 0.41592123 0.01695618 -0.210912128 0.24482448 -0.39897518 -0.626104328 -0.17184603 -1.20409841 -1.690371438 -0.71782537 -0.26036866 -0.475357218 -0.04538011 -0.61832495 -0.879597265 -0.35705264 0.73193281 0.361615699 1.10224993 0.89168074 0.403683729 1.37967776 0.83638583 0.564885939 1.10788573 0.70195348 0.382311430 1.02159553 0.88923566 0.522951631 1.25551969 NA NA NA 1.12886144 0.905965492 1.35175739 0.91902698 0.662441343 1.17561261 0.50309562 0.247166190 0.75902505 -0.30202761 -0.802401152 0.19834594 0.64170213 0.396483250 0.88692102 0.28374585 -0.002916750 0.57040845 0.15974793 -0.401413416 0.72090927 0.10445302 -0.283455684 0.49236172 -0.02997933 -0.452982892 0.39302422 0.15730285 -0.301962827 0.61656852 NA NA NA 142 p(adjusted) 0.0000000 NA 0.0000000 0.0000000 0.0000000 0.9977503 0.0000000 0.0000000 0.0000000 0.9558320 1.0000000 0.9999297 0.6150781 1.0000000 NA 0.0041850 1.0000000 0.0000002 0.0000000 0.0035382 0.0000000 0.0000000 0.0000001 0.0000000 0.0000000 0.0000000 NA 0.0000000 0.0000000 0.0000000 0.7832579 0.0000000 0.0557917 0.9998539 0.9999274 1.0000000 0.9986633 NA g:oe-g:fa k:oe-g:fa g:sa-g:fa k:sa-g:fa g:se-g:fa k:se-g:fa g:fb-k:fa k:fb-k:fa g:fe-k:fa k:fe-k:fa g:oe-k:fa k:oe-k:fa g:sa-k:fa k:sa-k:fa g:se-k:fa k:se-k:fa k:fb-g:fb g:fe-g:fb k:fe-g:fb g:oe-g:fb k:oe-g:fb g:sa-g:fb k:sa-g:fb g:se-g:fb k:se-g:fb g:fe-k:fb k:fe-k:fb g:oe-k:fb k:oe-k:fb g:sa-k:fb k:sa-k:fb g:se-k:fb k:se-k:fb k:fe-g:fe g:oe-g:fe k:oe-g:fe g:sa-g:fe k:sa-g:fe diff lwr upr 0.39692862 0.041339575 0.75251767 0.18709416 -0.190526225 0.56471455 -0.22883719 -0.606012011 0.14833762 -1.03396042 -1.605917285 -0.46200356 -0.09023068 -0.460221586 0.27976023 -0.44818697 -0.846855267 -0.04951867 -0.05529491 -0.556772208 0.44618239 -0.18972726 -0.718819248 0.33936472 -0.00244508 -0.560953165 0.55606300 NA NA NA 0.23718070 -0.239736374 0.71409777 0.02734623 -0.466215795 0.52090826 -0.38858512 -0.881806332 0.10463609 -1.19370835 -1.848033172 -0.53938353 -0.24997861 -0.737728124 0.23777091 -0.60793490 -1.117780689 -0.09808910 -0.13443235 -0.474299147 0.20543444 0.05284983 -0.331210568 0.43691023 NA NA NA 0.29247560 0.041435767 0.54351544 0.08264114 -0.198738644 0.36402093 -0.33329021 -0.614071748 -0.05250868 -1.13841344 -1.651942487 -0.62488440 -0.19468370 -0.465738487 0.07637109 -0.55263999 -0.861693124 -0.24358685 0.18728218 -0.232195152 0.60675952 NA NA NA 0.42690796 0.124451605 0.72936431 0.21707350 -0.111001846 0.54514884 -0.19885786 -0.526420245 0.12870453 -1.00398109 -1.544509465 -0.46345271 -0.06025134 -0.379515412 0.25901272 -0.41820763 -0.770305187 -0.06611008 NA NA NA 0.23962578 -0.111761188 0.59101274 0.02979131 -0.343874824 0.40345745 -0.38614004 -0.759355889 -0.01292420 -1.19126327 -1.760617184 -0.62190936 143 p(adjusted) 0.0126549 0.9469042 0.7765863 0.0000001 0.9999798 0.0114301 1.0000000 0.9977580 1.0000000 NA 0.9452357 1.0000000 0.3288787 0.0000001 0.9302221 0.0045957 0.9935614 1.0000000 NA 0.0066762 0.9997849 0.0049504 0.0000000 0.4985248 0.0000001 0.9787492 NA 0.0001598 0.6457917 0.7757861 0.0000000 0.9999994 0.0048994 NA 0.5934063 1.0000000 0.0339802 0.0000000 g:se-g:fe k:se-g:fe g:oe-k:fe k:oe-k:fe g:sa-k:fe k:sa-k:fe g:se-k:fe k:se-k:fe k:oe-g:oe g:sa-g:oe k:sa-g:oe g:se-g:oe k:se-g:oe g:sa-k:oe k:sa-k:oe g:se-k:oe k:se-k:oe k:sa-g:sa g:se-g:sa k:se-g:sa g:se-k:sa k:se-k:sa k:se-g:se diff lwr upr -0.24753353 -0.613487755 0.11842070 -0.60548981 -1.000414668 -0.21056496 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA -0.20983446 -0.444664176 0.02499525 -0.62576582 -0.859878356 -0.39165328 -1.43088905 -1.920462831 -0.94131526 -0.48715930 -0.709512870 -0.26480574 -0.84511559 -1.112480981 -0.57775020 -0.41593136 -0.682318789 -0.14954392 -1.22105458 -1.726856991 -0.71525218 -0.27732484 -0.533439449 -0.02121023 -0.63528113 -0.931318011 -0.33924425 -0.80512323 -1.310593070 -0.29965339 0.13860651 -0.116850680 0.39406371 -0.21934977 -0.514818082 0.07611854 0.94372974 0.443597570 1.44386191 0.58577346 0.064069197 1.10747771 -0.35795629 -0.644197356 -0.07171522 144 p(adjusted) 0.6078168 0.0000191 NA NA NA NA NA NA 0.1435127 0.0000000 0.0000000 0.0000000 0.0000000 0.0000114 0.0000000 0.0192103 0.0000000 0.0000063 0.8926184 0.4368322 0.0000000 0.0116379 0.0020049