Dynamics of functional meanings in discourse

THE DYNAMICS OF FUNCTIONAL MEANINGS IN DISCOURSE Volume I LOH BOON LIANG (B.SC. (HONS), NUS) A THESIS SUBMITTED FOR THE DEGREE OF MASTER OF ARTS (ENGLISH LANGUAGE) DEPARTMENT OF ENGLISH LANGUAGE AND LITERATURE NATIONAL UNIVERSITY OF SINGAPORE 2010 ACKNOWLEDGEMENTS The opportunity to undertake truly invigorating research does not come easily. I would like to express my heartfelt thanks to my supervisor Kay O’Halloran. I am deeply grateful to her for accepting me into her research group, despite my newness to the field of Systemic Functional Linguistics. I am equally indebted to Kevin Judd for believing I was suitable for the challenging research project. I would also like to thank all research staff and fellow postgraduate students at the Multimodal Analysis Lab, Interactive and Digital Media Institute, National University of Singapore, for all their personal support, fellowship and friendship through the years: Brad, Marissa, Nizah, Yiqiong, Liu Yu, Melany, Monica, Dezheng, Sabine, Thiha, Victor and Bertrand. Just four years ago I was not even aware of the existence of linguistics. I owe it to Tara Mohanan and Kim Chonghyuck for opening up this world for me in their linguistics undergraduate teaching, and writing recommendations for my application into this MA program. This work is part of the Socio-Cultural Modeling Project at the Multimodal Analysis Lab, supported by the US Air Force Office of Scientific Research (AFOSR) through the Asian Office of Aerospace Research and Development (AOARD) under Research Grant FA2386-09-1-4008 AOARD 094008 ii TABLE OF CONTENTS Volume I TITLE PAGE .................................................................................................................. i ACKNOWLEDGEMENTS ...........................................................................................ii TABLE OF CONTENTS ............................................................................................. iii SUMMARY .................................................................................................................. vi LIST OF TABLES .......................................................................................................vii LIST OF FIGURES ...................................................................................................... xi Chapter 1: INTRODUCTION ....................................................................................................... 1 Chapter 2: DATA AND METHODOLOGY ................................................................................... 4 2.1 The CNN Story ................................................................................................... 4 2.2 Systemics (The Software) ................................................................................ 10 Chapter 3: SYSTEMIC FUNCTIONAL ANALYSES ................................................................. 18 3.1 Logical Complexes ........................................................................................... 18 3.1.1 Concise Complexes ............................................................................ 19 3.1.2 Detailed Complexes ........................................................................... 24 3.2 Theme ............................................................................................................... 32 3.2.1 Wave Metaphor for the Textual Metafunction ................................... 32 3.2.2 Short Range Theme - Rheme Dynamics ............................................ 37 3.2.3 Thematic Progression (Daneš) ........................................................... 48 3.2.4 Marked Themes .................................................................................. 55 3.2.5 Interpersonal Themes ......................................................................... 56 3.3 Transitivity ........................................................................................................ 59 3.3.1 Relational Clauses .............................................................................. 59 3.3.2 Material Clauses ................................................................................. 67 3.3.3 Minority Type Clauses ....................................................................... 71 3.3.4 Some Individual Clauses .................................................................... 73 3.4 Ergativity........................................................................................................... 80 3.4.1 Medium .............................................................................................. 80 iii 3.4.2 Agency ............................................................................................... 88 3.5 Mood ................................................................................................................. 92 3.5.1 Mood and Speech Function ................................................................ 92 3.5.2 Modality and Mood Adjuncts ............................................................ 97 3.5.3 Modality in Transitivity and Grammatical Metaphor ...................... 100 3.5.4 Tense and Polarity ............................................................................ 101 3.5.5 Rankshifting ..................................................................................... 105 3.6 Grammatical Metaphors.................................................................................. 107 3.6.1 Experiential Shifts ............................................................................ 110 3.6.2 Logical, Interpersonal and Textual Shifts ........................................ 116 3.6.3 High Density Clauses ....................................................................... 119 3.6.4 Distribution of Grammatical Metaphors .......................................... 122 3.6.5 ‘Void’ of Grammatical Metaphors ................................................... 125 3.6.6 Parallel Patterns with Rankshifted and Interrupting clauses ............ 128 Chapter 4: RECURRENCE PLOTS ............................................................................................ 130 4.1 Clauses as Vectors .......................................................................................... 130 4.2 Types of Recurrence Plots .............................................................................. 134 4.2.1 XOR count: Difference in Meaning ................................................. 134 4.2.2 AND count: Similarity in Meaning .................................................. 135 4.2.3 OR count: Sum of Meanings ............................................................ 136 4.2.4 Reduced Vectors .............................................................................. 137 4.2.5 Tag Decomposition .......................................................................... 137 4.3 Plots of Individual Metafunctions ................................................................... 139 4.3.1 Mood ................................................................................................ 139 4.3.2 Theme ............................................................................................... 145 4.3.3 Transitivity ....................................................................................... 154 4.4 Three Level Investigation ............................................................................... 161 4.4.1 First Level of Delicacy ..................................................................... 162 4.4.2 Second Level of Delicacy ................................................................ 165 4.4.3 Third Level of Delicacy ................................................................... 168 4.5 Combined Metafunctions ................................................................................ 172 Chapter 5: SINGULAR VALUE DECOMPOSITION (SVD) ................................................... 178 iv 5.1 Experiment 1: Main Story ............................................................................... 180 5.1.1 Feature F0 ......................................................................................... 184 5.1.2 Feature F1 ......................................................................................... 189 5.1.3 Feature F2 ......................................................................................... 208 5.1.4 Feature F3 ......................................................................................... 211 5.1.5 Feature F4 ......................................................................................... 215 5.2 Experiment 2: Comments ............................................................................... 219 Chapter 6: CONCLUSION ...................................................................................................... 221 REFERENCES .......................................................................................................... 224 v SUMMARY This thesis explores the dynamics of meaning in discourse, based on the theoretical framework of Systemic Functional (SF) Grammar. Meaning is a multidimensional concept, spanning the textual, ideational and interpersonal metafunctions (Halliday 1978). In this thesis, meaning is described not just qualitatively in words, but also quantitatively as well, in discrete and non-discrete numbers, vectors and matrices, and also presented visually in the form of colour diagrams of the clauses, the grammar trees, and a variety of plots inspired from research in Dynamical Systems (Judd 2009a, b, c). The goal of this research is to arrive at a deeper understanding of language and its meaning in our world. The main body of this thesis is organized as follows; Chapter 1: Introduction considers the philosophical motivations for this research; Chapter 2: Data and Methodology introduces the text and the computer software (Systemics) for analysis; Chapter 3: Systemic Functional Analyses interprets and explains the linguistic significance of the completed analysis; Chapter 4: Recurrence Plots and Chapter 5: Singular Value Decomposition (SVD) advances the linguistic analysis by exploring the novel use of mathematical tools to extract and visualize interesting patterns of meaning in the discourse; and Chapter 6: Conclusion provides a summary of the main achievements of this research. vi LIST OF TABLES Table 3-1: Theme analysis of first clause .................................................................... 32 Table 3-2: Theme analysis of Set A - clauses with the word “crisis”.......................... 33 Table 3-3: Theme analysis of Set B - clauses with the word “credit” ......................... 34 Table 3-4: Thematic progression in successive clauses ............................................... 39 Table 3-5: Example of RH TH: R19 T20 ............................................................ 42 Table 3-6: Actual frequencies of short range thematic progression types ................... 42 Table 3-7: Predicted frequencies of short range thematic progression types .............. 45 Table 3-8: Thematic progression with Theme substitution ......................................... 47 Table 3-9: Short range Thematic Progression type 3 (TP3) with implicit [T]............. 49 Table 3-10: Short range Thematic Progression type 3 (TP3) with explicit [T] ........... 50 Table 3-11: Long range Thematic Progression type 3 (TP3) ...................................... 51 Table 3-12: Thematic Progression type 1 (TP1) .......................................................... 52 Table 3-13: Thematic Progression type 2 (TP2) .......................................................... 53 Table 3-14: The consequence of clause 50 & 51 ......................................................... 54 Table 3-15: Thematic Progression type 2 (TP2) with complex Theme ....................... 54 Table 3-16: Marked Themes functioning as location in time ...................................... 55 Table 3-17: Interpersonal Themes ............................................................................... 56 Table 3-18: Clause 56 (in opposition to clause 55) ..................................................... 56 Table 3-19: Interpersonal Modal-Metaphors as Theme............................................... 57 Table 3-20: Transitivity types in the text ..................................................................... 59 Table 3-21: All relational attributive clauses in the text (underlined: Attribute) ........ 60 Table 3-22: All negative Finites in the text ................................................................. 61 Table 3-23: All relational identifying clauses in the text (underlined: Value) ............ 62 Table 3-24: Tightness of encoding vs. decoding equivalence relations ...................... 65 Table 3-25: Relational clauses with rankshifted material processes underlined ......... 66 Table 3-26: All material clauses in the text ................................................................. 68 Table 3-27: Clauses with Goals in non-ranking material processes ............................ 69 Table 3-28: Construal of the impact of the crisis in clauses without Goal .................. 70 Table 3-29: Minority transitive process types I – mental (6 clauses) .......................... 71 Table 3-30: Minority transitive process types II – behavioural/verbal/existential (3 clauses) ................................................................................................. 72 vii Table 3-31: Metaphorical verbal clause 44 interpreted as relational meaning ............ 72 Table 3-32: A relational clause which ultimately construes material and mental processes ................................................................................................... 73 Table 3-33: A mental clause which portrays material events in grammatical metaphors ................................................................................................. 74 Table 3-34: Congruent usage of grammatical metaphors in clause 23 ........................ 74 Table 3-35: Word group expansion in clause 23 ......................................................... 75 Table 3-36: Categories of frequency occurring Mediums in the text .......................... 83 Table 3-37: Medium “we” in clauses at/near boundaries of logico complexes ........... 84 Table 3-38: Voice in the text (ranking clauses) ........................................................... 88 Table 3-39: Types of Effective Voice (ranking) clauses in the Text ........................... 88 Table 3-40: Material clauses in the text with non-ellipsed Agent ............................... 89 Table 3-41: Agency in the salad analogy ..................................................................... 89 Table 3-42: Hidden Agent in the salad analogy........................................................... 90 Table 3-43: Rankshifted processes with government/policymaker as potential Agent ........................................................................................................ 90 Table 3-44: Ellipsed Agent: government ..................................................................... 90 Table 3-45: Rankshifted Agent: government ............................................................... 91 Table 3-46: Mood of clauses in the text....................................................................... 92 Table 3-47: Speech Function in the text ...................................................................... 92 Table 3-48: Clauses with non-declarative mood in the text ........................................ 93 Table 3-49: Nexus of clause 35 and 36 ........................................................................ 95 Table 3-50: Clauses with Speech Function = Command ............................................. 96 Table 3-51: Clause with Speech Function = Offer ...................................................... 96 Table 3-52: Modality Orientation in the text ............................................................... 97 Table 3-53: Clauses with Modal Finites (examples) ................................................... 98 Table 3-54: Relational clauses that construe modality .............................................. 100 Table 3-55: Some modalised versions of statements in Table 3-54 .......................... 101 Table 3-56: Tense in the text ..................................................................................... 101 Table 3-57: Clauses with present, past, future tenses (examples) ............................. 102 Table 3-58: Clauses with future tense at the start of the text ..................................... 103 Table 3-59: Polarity in the text .................................................................................. 104 Table 3-60: Clauses with positive/negative polarity (examples) ............................... 104 Table 3-61: Clauses with double embedding in the text ............................................ 105 Table 3-62: Alternate version of clause 17, with no embedding ............................... 105 viii Table 3-63: Grammatical metaphor examples demonstrating all the grammatical functions ............................................................................ 107 Table 3-64: Types of grammatical metaphors in the text .......................................... 108 Table 3-65: Metaphorical and congruent use of the word “crisis” ............................ 109 Table 3-66: Statistics of Experiential grammatical metaphors in the text ................. 110 Table 3-67: Experiential grammatical metaphors shifting Process to Entity ............ 111 Table 3-68: Examples of clauses with experiential grammatical metaphors shifting Process to Entity........................................................................ 111 Table 3-69: Experiential grammatical metaphors shifting Entity to Modifier .......... 112 Table 3-70: Examples of clauses with experiential grammatical metaphors shifting Entity to Modifier...................................................................... 112 Table 3-71: Experiential grammatical metaphors shifting Process to Quality .......... 113 Table 3-72: Examples of clauses with experiential grammatical metaphors shifting Process to Quality ..................................................................... 113 Table 3-73: Experiential grammatical metaphors with shifts from Circumstance .... 114 Table 3-74: Clause showing shifts from Circumstance ............................................. 114 Table 3-75: Experiential grammatical metaphors with shifts from Quality .............. 114 Table 3-76: Clause showing shift from Quality......................................................... 114 Table 3-77: Metaphors representing the least frequent experiential shifts ................ 115 Table 3-78: Clause showing infrequent type experiential shifts ................................ 115 Table 3-79: Clauses with logical grammatical metaphors (underlined) .................... 116 Table 3-80: Clauses with interpersonal grammatical metaphors (underlined) .......... 117 Table 3-81: Clauses with textual grammatical metaphors (underlined) .................... 118 Table 3-82: Clauses with the most grammatical metaphors ...................................... 119 Table 3-83: Unpacking grammatical metaphors in clause 58 .................................... 120 Table 3-84: Experiential grammatical metaphors shifting Entity to Modifier, at the start of the discourse ......................................................................... 124 Table 3-85: Grammatical metaphor ‘void’ in red ...................................................... 126 Table 3-86: Grammatical metaphors in the second major logico complex leading to the ‘void’ ............................................................................... 127 Table 3-87: Occurrences of “salad” in the grammatical metaphor ‘void’ ................. 127 Table 3-88: Grammatical metaphors versus rankshifted and interrupting clauses .... 129 Table 4-1: All possible ergativity tags in a clause ..................................................... 131 Table 4-2: XOR Definition ........................................................................................ 135 Table 4-3: AND Definition ........................................................................................ 135 Table 4-4: OR Definition ........................................................................................... 136 ix Table 4-5: Decomposition of a tag representing intensive relational attributive process ...................................................................................................... 138 Table 4-6: Tags (35) used in the analysis of Figure 4-10 .......................................... 140 Table 4-7: Tags (21) used in the analysis of Figure 4-14 and Figure 4-22 ................ 145 Table 4-8: Tags (61) used in the analysis of Figure 4-24 and Figure 4-27 ................ 154 Table 4-9: < First Level > tags (9) analysed .............................................................. 162 Table 4-10: < Second Level > tags (24) analysed ..................................................... 165 Table 4-11: < Third Level > tags (32) analysed ........................................................ 168 Table 4-12: Combined metafunctions - tags (118) analysed ..................................... 172 Table 5-1: Overview of transitive processes in the main story.................................. 181 Table 5-2: Frequency distribution of all T1 process types in reduced matrix ............ 182 Table 5-3: Frequency distribution of all T1 tags in reduced matrix -77 tags ............. 182 Table 5-4: Clauses in the three tightest clusters of clause neighbours in T1 ............. 188 Table 5-5: Stretches of text that continuously exhibit similar F1 values ................... 194 Table 5-6: All Mann/Qual occurrences in the text .................................................... 196 Table 5-7: Clauses of the three deep red regions of F1 in Figure A-7....................... 199 Table 5-8: Clauses of the three deep blue regions of F1 in Figure A-7 ..................... 202 Table 5-9: Clauses of the faint blue cluster in Figure A-7 ......................................... 202 Table 5-10: Clauses of the two faint red clusters in Figure A-7 ................................ 204 Table 5-11: The outliers: very faint blue clauses in Figure A-7 ................................ 205 Table 5-12: Clauses with metaphorical material processes in the text (T1m/Proc/Mat) ..................................................................................... 212 Table 5-13: Frequencies of tags not related to mental processes but significant in F3 (in bold), with other associated tags not significant in F3 (not in bold) ............................................................................................ 213 Table 5-14: All Attribute/Circumstance in the text (bold)......................................... 213 Table 5-15: All Location/Time in the text (bold) ...................................................... 213 Table 5-16: All circumstances of transitivity, T1 or T1m – 25 tags ........................... 214 x LIST OF FIGURES Figure 1-1: Meaning in text as arising from a web of functional relations (examples of each relation in italics) .......................................................... 3 Figure 2-1: Website layout of CNN article .................................................................... 5 Figure 2-2: Main story with ten comments .................................................................... 9 Figure 2-3: Clausal analysis with the Systemics software (Clause Page) ................... 12 Figure 2-4: Grammatical metaphor analysis in Systemics .......................................... 14 Figure 2-5: Analysis of metaphors of transitivity in Systemics ................................... 15 Figure 2-6: System network for Theme analysis (TH), with selections for clause 48 highlighted (Grammar Page)..................................................... 16 Figure 3-1: Overview of logical complexes................................................................. 20 Figure 3-2: Logical complexes of the discourse (concise) .......................................... 24 Figure 3-3: Logical complexes in detail – part 1 ......................................................... 25 Figure 3-4: Logical complexes in detail – part 2 ......................................................... 27 Figure 3-5: Elaborations in the second major complex (Figure 3-4), highlighted....... 28 Figure 3-6: Logical complexes in detail – part 3 ......................................................... 29 Figure 3-7: Logical complexes in detail – part 4 ......................................................... 30 Figure 3-8: String 1 and String 2 – lexical strings of “crisis” and “credit”, respectively ............................................................................................... 35 Figure 3-9: The salad analogy (Theme underlined) ..................................................... 41 Figure 3-10: Thematic progression possibilities for k = 2 ........................................... 45 Figure 3-11: Three major patterns of Thematic Progression (Daneš 1974): ............... 48 Figure 3-12: Identification in clause 34 ....................................................................... 63 Figure 3-13: Logical structure leading to clause 23 – a “downward” progression...... 76 Figure 3-14: Lexical string of ups and downs – all words associated with rising/falling in the text........................................................................... 77 Figure 3-15: All Mediums in the text, and their clausal context ................................. 82 Figure 3-16: Reference chains 1, 2, 5, 6 corresponding to all participants in the discourse belonging to categories 1, 2, 5, 6 of Table 3-36 ..................... 86 Figure 3-17: Items in Reference chains 1, 2, 5, 6 highlighted in the text .................... 87 Figure 3-18: First major logico complex in the text (underlined in red: non-declarative mood) ............................................................................ 94 Figure 3-19: Second major logico complex in the text (underlined in red: non-declarative mood) ............................................................................ 94 xi Figure 3-20: Third major logico complex in the text (underlined in red: non-declarative mood) ............................................................................ 95 Figure 3-21: Mood Adjunct analysis of clause 55 ....................................................... 98 Figure 3-22: Mood Adjunct analysis of clause 50 ....................................................... 99 Figure 3-23: Piano roll for Tense ............................................................................... 102 Figure 3-24: Piano roll for Tense (enlarged) ............................................................. 103 Figure 3-25: Grammatical metaphor analysis of clause 58 ....................................... 120 Figure 3-26: Opposition between clause 55 and clause 58 ........................................ 121 Figure 3-27: Colour code for Figure 3-28.................................................................. 122 Figure 3-28: Grammatical metaphor density against logical structure ...................... 123 Figure 3-29: Grammatical metaphor ‘void’ in second major logico-complex .......... 125 Figure 4-1: A vector, in row and column form. ......................................................... 130 Figure 4-2: Mapping a clause into tags (clause 7) ..................................................... 130 Figure 4-3: A vector representing the ergativity analysis of clause 7........................ 131 Figure 4-4: Ergativity analysis of clause 8 ................................................................ 134 Figure 4-5: A vector representing the ergativity analysis of clause 8........................ 134 Figure 4-6: Difference in ergativity meaning between clause 7 and 8 ...................... 134 Figure 4-7: Similarity in ergativity meaning between clause 7 and 8 ....................... 136 Figure 4-8: Sum of ergativity meaning between clause 7 and 8 ................................ 136 Figure 4-9: Transitivity analysis of clause 7 .............................................................. 137 Figure 4-10: Differences in Mood between all clauses (M1 tags –XOR) ................. 141 Figure 4-11: Scale numbering in Figure 4-10 ............................................................ 142 Figure 4-12: Mood analysis of clause 55 ................................................................... 143 Figure 4-13: Logical complexing context around clause 55 ...................................... 144 Figure 4-14: Differences in Theme between all clauses (TH1 tags –XOR) ............... 146 Figure 4-15: Patterns in thematic differences (from Figure 4-14) ............................. 147 Figure 4-16: “Red Block 1”: clause 28-35 – consistent TH1 choices ........................ 147 Figure 4-17: “Red Block 2”: clause 58-63 – consistent TH1 choices ........................ 148 Figure 4-18: “Red Block 3”: clause 73-78 – consistent TH1 choices ........................ 149 Figure 4-19: “Blue Strip 1”: clause 24 – unique TH1 choices ................................... 150 Figure 4-20: “Blue Strip 2”: clause 42, 43 – unique TH1 choices ............................. 150 Figure 4-21: “Blue Strip 3”: clause 71 – unique TH1 choices ................................... 150 Figure 4-22: Similarities in Theme between all clauses (TH1 tags –AND) ............... 152 xii Figure 4-23: “Blue Square” clauses in Figure 4-22 ................................................... 153 Figure 4-24: Differences in Transitivity between all clauses (T1 tags –XOR) .......... 155 Figure 4-25: Transitivity analysis of clause 50 .......................................................... 155 Figure 4-26: Lexical string of words amplifying scale .............................................. 157 Figure 4-27: Sum of similarities and differences in Transitivity between all clauses (T1 tags –OR) ........................................................................... 158 Figure 4-28: Two Phases in Transitivity (T1 tags –OR) ............................................ 159 Figure 4-29: < First Level > Speech Function, Mood, Mood Metaphor tags –XOR plot............................................................................................. 163 Figure 4-30: Analysis of clauses with greatest changes in First Level ...................... 164 Figure 4-31: < Second Level > Int1 tags –XOR plot ................................................. 166 Figure 4-32: Analysis of clauses in “Red Square”..................................................... 166 Figure 4-33: Logico-complexing context of “Red Square” ....................................... 167 Figure 4-34: < Third Level > Int1 and TH1 tags –XOR plot ..................................... 169 Figure 4-35: Change in meaning at clause 42, 43 compared across three levels of delicacy .................................................................................. 170 Figure 4-36: Clause 42 & 43 TH Analysis ................................................................. 171 Figure 4-37: Combined Metafunctions XOR plot ..................................................... 173 Figure 4-38: Patterns in Combined Metafunctions XOR plot ................................... 174 Figure 4-39: Clauses in the “Light Band”.................................................................. 174 Figure 4-40: Clauses in the “Dark Band” .................................................................. 176 Figure 5-1: Visualization of SVD in two dimensions ................................................ 179 Figure 5-2: Tightest clusters in Figure A-3................................................................ 187 Figure 5-3: Meaning of F0 and F1 in SVD ................................................................ 190 Figure 5-4 a, b: Contribution of presence and absence of tags to F1......................... 192 Figure 5-5: Stretches of constant F1 values (in the clause list of Figure A-4) .......... 194 Figure 5-6: Deep red regions of F1 in Figure A-7 ..................................................... 198 Figure 5-7: Transitivity of some clauses in red region (2) in Figure A-7 .................. 200 Figure 5-8: Transitivity of a pair of clauses in red region (3) in Figure A-7 ............. 200 Figure 5-9: Deep blue regions of F1 in Figure A-7 ................................................... 201 Figure 5-10: Faint regions of F1 in Figure A-7 ......................................................... 203 Figure 5-11: Clauses with metaphors of transitivity in the “tail” of red cluster (2) in Figure 5-6 ........................................................................ 206 Figure 5-12: Doublets and triplets of blue relational identifying clauses in Figure A-9 ........................................................................................ 209 xiii Figure 5-13: Tints for Reln/Attr tags in tag list in Figure A-12................................. 212 Figure 5-14: Examples of mixed red & blue tinted clauses in Figure A-13 .............. 212 Figure 5-15: gmakens using Smick’s style ................................................................ 220 Figures in Volume II Figure A-1: Main story T1 F0 tag wheel .................................................................... 391 Figure A-2: Main story T1 F0 tag neighbours ........................................................... 392 Figure A-3: Main story T1 F0 clause neighbours ...................................................... 393 Figure A-4: Main story T1 F1 tag wheel .................................................................... 394 Figure A-5: Main story T1 F1 text ............................................................................. 395 Figure A-6: Main story T1 F1 tag neighbours ........................................................... 396 Figure A-7: Main story T1 F1 clause neighbours ...................................................... 397 Figure A-8: Main story T1 F2 tag wheel .................................................................... 398 Figure A-9: Main story T1 F2 text ............................................................................. 399 Figure A-10: Main story T1 F2 tag neighbours ......................................................... 400 Figure A-11: Main story T1 F2 clause neighbours .................................................... 401 Figure A-12: Main story T1 F3 tag wheel .................................................................. 402 Figure A-13: Main story T1 F3 text ........................................................................... 403 Figure A-14: Main story T1 F3 tag neighbourhood ................................................... 404 Figure A-15: Main story T1 F3 clause neighbourhood .............................................. 405 Figure A-16: Main story T1 F4 tag wheel .................................................................. 406 Figure A-17: Main story T1 F4 text ........................................................................... 407 Figure A-18: Main story T1 F4 tag neighbours ......................................................... 408 Figure A-19: Main story T1 F4 clause neighbours .................................................... 409 Figure A-20: Comments T1 F1 tag wheel .................................................................. 410 Figure A-21: Comments T1 F2 tag wheel .................................................................. 411 Figure A-22: Comments T1 F3 tag wheel .................................................................. 412 xiv _____________________________________________________________________ Chapter 1: INTRODUCTION _____________________________________________________________________ As physical beings in a physical world, we are surrounded by physical phenomena on many scales in time and space: the ripples on a water surface, the circulation of air, and the movements of the sun and moon. These physical events are not random but follow fixed laws. In recent centuries, mankind has made tremendous progress towards understanding the laws of the physical world, and can now predict and control the evolution of physical systems with unprecedented precision. Our physical existence, however, is merely a platform for our social existence. As social beings in a social world, we are surrounded by social phenomena on many scales, and there exist deeper forces and principles behind social events, which we seek to comprehend. But presently, we cannot predict social events with as much success as physical events; hence, we do not understand the social world as well as the physical world. The analogy is not perfect, however; we are not merely surrounded by social phenomena, but we actively engage in them, playing a part in the construction of their rules, and their evolution. Language is the instrument with which we define our social identity and conduct our daily social interactions. Linguistic phenomena, be it a spoken/written text or a multimodal website/video presentation, are a reflection of social-cultural events and trends. This thesis explores the linguistic dynamics of one textual discourse, affording us a glimpse of the grand scale of the universe of social-cultural dynamics. 1 The dynamics of the physical world are an illusion: the motions of the planets, the scatter of billiard balls – these changes can be seen as logical outcomes of the laws of physics, which do not change. Events in the social world too are an illusion: they simply form a platform on which we execute our personal agendas. But personal choices are heavily influenced by our social circumstances and opportunities. It is only as a whole that we can hope to describe these “social forces”. Michael Halliday’s Systemic Functional (SF) Linguistics is a theory of language and meanings based on the choices we make in producing each instance of language; the function of each utterance as a reflection of our momentary intention in the immediate social context. Meaning is described under the metafunctional principle proposed by Halliday (1978) which is described using system networks from Halliday & Matthiessen (2004). By collecting and analysing large linguistic datasets reflecting real events in society with Systemic Functional Theory, and combining this with computerised mathematical analysis from the framework of Dynamical Systems Theory (Judd 2009a, b, c), it is hoped that we will be able to better describe and model these events. This is the aim of the Socio Cultural Modelling Project led by Kay O’Halloran and Kevin Judd at the Multimodal Analysis Lab, Interactive Digital Media Institute (IDMI), and this thesis represents the beginnings of the project. The study of meanings in language1 is a complex problem that challenges theoreticians on all fronts. To begin with, we can think of meaning in text as arising from meaning in words. This is what dictionaries attempt to do. However, while this can be helpful in many situations, it fails to cover the meaning derived from the 1 The term “language” in this thesis refers to natural language (as opposed to “artificial language” such as those created for programming machines), unless stated otherwise. 2 patterns of words within sentences and patterns of sentences within a discourse. Meaning in text is not confined in words, nor is it confined in sentences; it is an ongoing dynamic process of orchestration of functional relations between words, sentences, discourse and social context, see Figure 1-1: Figure 1-1: Meaning in text as arising from a web of functional relations (examples of each relation in italics) There is often the philosophical question of whether linguistics is like a snake trying to swallow itself: language was created to describe the world, so how can it be used to describe meanings within itself (Firth 1948)? A prudent way to proceed might be to rely heavily upon a kind of language very different from natural language – mathematics – to describe natural language, thus invoking a new meaning potential. Meaning in natural language is fuzzy (Halliday 1995) and its grammar is flexible, while mathematics is precise and its rules are strict. Two chapters in this thesis, Chapter 4: Recurrence Plots and Chapter 5: Singular Value Decomposition (SVD), are devoted to the application of mathematical methods from linear algebra to analyse the units of systemic functional meaning. The next chapter, Chapter 2: Data and Methodology, will explain how data is collected and analysed with software, and how the analysis is stored in a format suitable for mathematical analysis. 3 _____________________________________________________________________ Chapter 2: DATA AND METHODOLOGY _____________________________________________________________________ 2.1 THE CNN STORY The subject for analysis is a news article2 on the Financial Crisis of 2008, from an international news website, CNN.com, dated 10 October 2008. A screenshot of the article, in its original form and layout on the website, is shown in Figure 2-1. This text is used to demonstrate how the results of systemic functional analysis can be interpreted dynamically in terms of shifts in functionality, the relations and changes between clauses, logogenetic patterns, and the meanings that are generated as a result. The topic of financial crisis was chosen as it was a global phenomenon that impacted all levels of society - affecting many countries, people, and institutions, and it was an event that extended significantly forwards and backwards in time. The scale and complexity of the event would mean that any representation of it would stretch the potential of meaning-making resources in language. This particular website was chosen because it allowed internet readers to write in their comments. There were a total of 117 comments. In addition to the main story, the ten latest comments are analysed, but the focus of the analysis is on the main story. 2 The complete link for the article is: http://edition.cnn.com/2008/POLITICS/10/09/smick.crisis/index.html 4 Figure 2-1: Website layout l of CNN article The figures which follow are excerptss (preserving the original formatting) from the website showing the full text of the main story, ory, followed by ten reader comments. 5 6 7 8 Figure 2-2: Main story with ten comments This article appeared on 10 October 2008, at the very height of the Financial Crisis. In the previous month, Lehman Brothers (4th largest investment bank in US) filed for bankruptcy, Merrill Lynch (3rd biggest investment bank in US) sold itself and US insurer AIG came to the brink of collapse, triggering shockwaves worldwide. These events in the US impacted on financial markets in Europe, Asia and all corners of the globe, and plunged economies into recession. In the week before 10 October, global markets lost U$6 trillion in panic selling, and governments injected U$6 trillion into the financial system in attempts to salvage the situation, with little effect. These events were disseminated by media worldwide, and people everywhere feared for their jobs and financial investments. This CNN story and commentary captures a minuscule snapshot of the spectacular scale of events and interactions that unfolded. 9 Appendix 1.1 (p. 227) displays the text of the CNN main story organised into 74 ranking clauses, numbered from “clause 7” to “clause 80”. The ten comments have also been organised giving a total of 103 clauses, in Appendix 1.2 (p. 230). 2.2 SYSTEMICS (THE SOFTWARE) Analysis of the text in the CNN article is performed with advanced versions of Systemics, a software for systemic functional analysis developed by Kevin Judd and Kay O’Halloran (O'Halloran 2003). The interface of Systemics is organized into “Pages” which fulfill different aspects of the analysis. • Text Page: allows decomposition of the text into ranking clauses and addition of symbols to further organize the text in preparation for analysis. o || clause boundary o [[rankshifted clause]] o [rankshifted group] o o o {ellipsed words} • Clause Page: zooms into each ranking clause, allows the user to annotate the textual, interpersonal and ideational functionality of words, word groups, and embedded clauses, as tags in a “Clause Table”, and further interpretation of the analysis to be encoded in a “Analysis Table” 10 • Interclausal Page: allows clause complexes to be drawn as links between clauses, revealing logical meaning • Discourse Page: allows reference chains and lexical strings to be drawn between the clauses, revealing cohesion • Search Page: allows user to selectively retrieve any combination of types of tags that have been input in the Clause Page, and display their distribution and statistics • Grammar Page: contains a (expandable) comprehensive database of SF Grammar, principally compiled from Halliday (1994), Halliday (1998b) and Martin (1992). This represents Halliday’s description of the meaning potential of language (English written text) encoded in the form of system networks. Appendix 7: The Grammar, p. 413, shows all system networks used in this thesis, extracted from the Grammar Page. Comprehensive descriptions of Halliday’s Systemic Functional systems are found elsewhere, e.g. Halliday & Matthiessen (2004), thus are not reproduced here. For a complete description of all the Pages of Systemics, please refer to O'Halloran (2003). The most important Page in concern for this thesis is the Clause Page (shown in Figure 2-3 next page) because it shows how tags are input, organized and named, and these tag names will be used for advanced mathematical analysis later in Chapter 4 and Chapter 5. 11 Figure 2-3: Clausal analysis with the Systemics software (Clause Page) Analysis done at the Clause Page is entered into two tables: a Clause Table, which represents direct labeling of clause constituents, and an Analysis Table, which is based on the interpretation of the clause table analysis and the discourse context. For the Clause Table, columns match words/parts of the clause, but for the Analysis 12 table, columns are arbitrarily arranged. For both tables, the tags are organized metafunctionally into Keys (rows): Clause Table S: TH: M: T: E: MET: Structure Theme Mood Transitivity Ergativity Grammatical metaphor (see Figure 2-4) Analysis Table S: Tex: Int: Exp: Structure Textual Interpersonal Experiential The tags3 are colour-coded in the Clause Page according to their Key. The Key numberings (e.g. “2” in “M2”) generally (but not always) describe the level of embedding. For example, in Figure 2-3, “M2”, “T2”, “E2” and “Int2”, “Exp2” correspond to the analysis of the embedded clause “of forcing the banks to start lending” (for this thesis, Themes in embedded clauses are not analysed). The exception to this Key numbering convention is “TH#”. The Key “TH2” is not used to encode embedded clauses, but to encode the ranking clause analysis in the event that the ranking clause is part of a marked clause complex construction. In these rare cases, “TH1” will encode the Theme/Rheme at the level of clause complex (see the analysis of clauses 79-80 in Appendix 2: SF Analysis of Clauses, p. 329 for an 3 Tag names used in Systemics shall appear in a different font (Calibri) in this thesis. In tag names the wildcard character * denotes any letter, number, or slash; and # denotes any number. 13 example). Grammatical metaphors (Halliday 1998b; Martin 1992) are analysed in Systemics according to the conventions shown in Figure 2-4 below: Figure 2-4: Grammatical metaphor analysis in Systemics The Key “MET1” labels the entire functional element (mood/transitivity element) in which the grammatical metaphor resides. The grammatical metaphor itself is labeled at the “MET1m” Key. For example, “bank” in “a private/global bank clearing facility” (clause 76) is a grammatical metaphor with a mainly experiential function (hence “Exp”) and congruently represents an Entity (“Ent”) but has been shifted to operate as a Modifier (“Mod”) of the Entity “clearing facility”. For an introduction of the grammatical functions (Modifier, Entity, etc) describing the grammatical metaphors, see section 3.6: Grammatical Metaphors, p. 107. In some clauses, there are “T#m” tags together with “T#”. These are cases of clauses with ideational metaphors of transitivity (Halliday 1994: 344) where there is a discrepancy between realized and interpreted transitivity process type. The clause is double coded: “T#m” encodes the metaphorically realized meaning, and “T#” encodes the congruent interpreted meaning. For an example, see Figure 2-5 next page. The “m” suffix is also used in the “E#” and “Exp#” Keys. 14 Figure 2-5: Analysis of metaphors of transitivity in Systemics Having introduced all the Keys in the tables, we now explain the theoretical significance of the tags. The tags in the Clause Table and Analysis Table represent, for the instance of the particular clause/word group/word, the choices made by the writer according to system networks stored in the Grammar Page. In Systemic Functional Theory, the choices represent meaning created in the text, and the system network represents the meaning potential. For example, consider the “TH1” analysis of clause 48 in Figure 2-3 (p. 12), which has two tags: Topic/Theme Rheme The slash “ / ” in the tag name “Topic/Theme” is denoted in Figure 2-3 (p. 12) as new line spacing between “Topic” and “Theme”, and represents hierarchical ordering of the tags, i.e. sub-branching of nodes in a system network. The two tags “Topic/Theme” and “Rheme” represent end nodes from the “TH” system network, shown in Figure 2-6 next page, which is an excerpt from the Grammar Page. In Figure 2-6, the main branches of “TH” (in blue) are: “Text” (Textual), “Int” (Interpersonal), “Topic”, and “Rheme”. All these branches, except “Rheme”, eventually end with a “Theme” end node. Note that tag hierarchy usually but not 15 always follows a part-whole relationship, e.g. in “Topic/Theme”, “Theme” is not a pure subset of “Topic” since it’s possible that a Theme is not a Topic. In fact, Topic is a component of Theme. Figure 2-6: System network for Theme analysis (TH), with selections for clause 48 highlighted (Grammar Page) The names “Topic/Theme” and “Rheme” displayed in the Clause Page (see Figure 2-3, p. 12) do not represent the full tag names stored in the Systemics database. Their complete names follow the full Grammar Page hierarchy (compiled in Appendix 7, p. 413) and hence, are: 16 Cl/TH1/Topic/Theme Cl/TH1/Rheme “Cl” denotes the tag’s source is from the Clause Table (“An” for Analysis Table). The tags contain complete descriptions of the system hierarchies. We see that even the simplest tag names are highly hierarchical. The full tag names will be used for mathematical calculations in Chapter 4 and Chapter 5. The Clause and Analysis Tables in the Clause Page can be printed in Systemics. The complete set of printouts representing the comprehensive SF analysis of the main story and the comments are compiled in Appendix 2: SF Analysis of Clauses, p. 234. The SF analysis of the main story is also presented in the form of matrices in Appendix 3: SF Analysis Matrices, p. 368. Printouts from other Pages of Systemics (Interclausal, Discourse) are also used in this thesis. The main story analysis is comprehensively explained and interpreted in the next chapter. 17 _____________________________________________________________________ Chapter 3: SYSTEMIC FUNCTIONAL ANALYSES _____________________________________________________________________ In this chapter, the SF analysis of the main story is discussed. The chapter is composed of six sections: 3.1 Logical Complexes, 3.2 Theme, 3.3 Transitivity, 3.4 Ergativity, 3.5 Mood, and 3.6 Grammatical Metaphors. The analyses discussed in sections 3.2 - 3.6 are based on parts of the analysis compiled in Appendix 2 and Appendix 3. 3.1 LOGICAL COMPLEXES Clause complexing and cohesion between clauses in a text can be described in terms of the type of logico-semantic relation (Expansion/Projection). The basic systems described by Halliday (1994: 216-221) are reproduced in Appendix 7.2 p. 418. To extend Halliday’s terminology, a group of clauses linked by logico-semantic relations (whether structural or non-structural) shall be called a logico-semantic complex, or logical complex/logico-complex4. It is useful to begin the SF analysis of the CNN article by examining the logico-complexes in the article, for this generates a large-scale map of meaning in the entire discourse in the logical metafunction, within which we can contextualize further analyses in the textual, interpersonal and experiential metafunctions, which create more localized meanings (at the level of the clause). The logical complexing encodings were performed in Systemics, in the Interclausal Page, with the following conventions: 4 Not to be confused with the term “clause complex”, which only refers to structurally linked clauses. 18 Logical Link Label Legend • • • • • = + × “ ‘ Elaboration Expansion Extension Enhancement Locution Idea Projection • Taxis (structural links) • Cohesive (non-structural links) Hypotactic links have arrows which point towards the dependent clause. Text Colour Key Conjunctive Adjunct (Theme) • Structural Conjunction (Theme) • Interrupting Clause or Interrupting Clause Complex • Rankshifted Clause Complex • The clause complexing between an interrupting clause and its ranking clause, and between clauses within a rankshifted/ interrupting clause complex, are also analysed, and such internal clauses are numbered in oval rather than rectangular boxes. 3.1.1 Concise Complexes The article consists of: • 74 Ranking clauses • 25 Rankshifted/Embedded clauses 19 • 8 Interrupting/Included clauses The overall logical organization of the ranking clauses is shown in Figure 3-1: Figure 3-1: Overview of logical complexes Over the following pages, Figure 3-2 displays the logical analysis of the ranking, interrupting and embedded clauses in concise format: for Expansion, only the basic divisions into elaborating, extending and enhancing relations are made. Note that clause 2 and clause 3 are not ranking clauses, but represent a rankshifted complex, forming the title of the article. 20 21 22 23 Figure 3-2: Logical complexes of the discourse (concise) The concise complex analysis provides the basic large-scale logical map of the discourse. To provide a finer and more grounded sense of the logical development of the text, a more delicate analysis of all the complexes is presented in sequence in the next section 3.1.2. 3.1.2 Detailed Complexes In the figures of logico-complexes that follow: Figure 3-3 to Figure 3-7, the level of detail of analysis is beyond Figure 3-2, which only subdivided the logicosemantic relation into elaborating, extending, enhancing types. For Figure 3-3 to Figure 3-7, the complete grammar for analysing Expansion is used, as explained in Halliday (1994: 225-241, 328-329), and compiled in Appendix 7.2 p. 419. 24 Figure 3-3: Logical complexes in detail – part 15 The article begins with a small logico-complex of clause 7-10 (see Figure 3-3 above) where the author humorously points out that amidst the chaos and confusion of the crisis, there is something which is clear: that our leaders don’t how to lead us through this troubled times: “…driving policy looking through the rearview mirror” (clause 9, 10). By ridiculing the policymakers, the author appears authoritative and to have complete knowledge of the situation. After clause 10 in Figure 3-3 is the first major logico complex of the article, 5 In Figure 3-3 and similar figures that follow, the link label “positive” is ambiguous: 1) */enhancement/condition/positive (for clause 13, 18 & >) 2) */extension/additive/positive (for all other instances) 25 spanning clause 11-23, describing the failure of the bailout package and the origins of the crisis. Clause 11-14 describe the promises of the bailout package; here again, the author injects a fair amount of derision by exaggerating the leaders’ expectations of the bailout: “…some magic pill, which, if gulped down…” (clause 12, 13). An extending replacive relation, marked by “instead”, follows, with clause 15-16 describing the failure of the bailout: “Instead, stock markets collapsed and credit markets remained frozen”. The remainder of the first major complex, clauses 17-23, forms an enhancing causal relation with clause 15-16, providing the historical background leading to the financial crisis. The cause of the crisis, according to David Smick, seems to center upon securitization by the banks, or as he puts it, the creation of “paper instruments”. This entire logico complex proceeds logically, describing events in causal and temporal sequence, and ends dramatically with the statement that we are experiencing a “downward spiral”. The second major logico complex, shown in Figure 3-4 next page, elaborates on the cause of the crisis, by explaining what exactly the paper instruments were, how they were created, and their effects on the financial system. The complex begins with a rhetorical question “So what are these paper instruments…?” (clause 24), the answer to which constitutes the rest of the complex. The first relation in the complex is elaborating clarification, with clause 25: “I like to use a salad analogy” which reconstrues the paper instruments as salad. The author seems to have a habit of using lexical metaphors to deride that what he aims to debunk (bailout package = “magic pill”; asset-backed securities = “paper instrument” = “salad”). Then he explains what was the standard form of loans in the past (“syndicated loans”) in clause 26, and uses an adversative “but” (clause 27) to begin the account of the creation of paper 26 instruments. Figure 3-4: Logical complexes in detail – part 2 Clause 28-32 (in Figure 3-4 above) form the largest clause complex in the article: “Instead of making simple loans, and holding them until maturity, a bank collected all its loans together, then diced and sliced them up into a big, beautiful tossed salad.” Incidentally, this is probably the most crucial point in the discourse, as it describes the making of the “salad”, the key to understanding the crisis. This is followed by clause 33, explaining the purpose of the “salad” - “to sell for huge fees…around the world” and its dire consequences, which is the subject of clause 3441: “…you’re dead” (clause 36), “…distrust heightened” (clause 40). A noteworthy semantic pattern in the second logico complex (Figure 3-4) is the many uses of elaborations atop the hierarchy of logical relations; these are highlighted in Figure 3-5 on the next page. This is not surprising given that the 27 purpose of the entire complex is to explain the complex financial term “securitization”. Figure 3-5: Elaborations in the second major complex (Figure 3-4), highlighted In the third major complex, shown in Figure 3-6 next page, the author highlights the extent of the crisis, and the consequences. Like in the second major complex, he begins with a question, in clause 42 & 43: “So what does this salad boycott mean for the future, and why have financial markets collapsed so brutally?” and answers it, i.e. clarifying himself in the rest of the complex. Clauses 44-58 and clauses 59-67 form two sub-complexes in an extending additive relation. The former sub-complex portents collapsing credit markets, and the latter provides the orientations of the government and the people towards the crisis. The main argument of the sub-complex spanning clause 44-58 is the logical cluster of clause 44-47, which predicts a “serious credit crunch in 2009”. The reasons are given in two logical clusters: clause 48-52 portray the immense scale of monetary losses in the crisis, and clause 53-58 express concerns about who the banks are willing to lend to. 28 Figure 3-6: Logical complexes in detail – part 3 After having described the effects of the crisis in financial terms thus far (“markets”, “interest rates”, “loans”, “credit”, etc) the discourse starts to turn towards the impact of the crisis on people in clause 58 (“jobs creation”). While people have earlier figured in the article as “bankers”, “investors”, in clause 67 the more down-toearth “American workers” appears. Clause 60-65 cautions that the crisis may be too large for government guarantees to work, this implies the vulnerability of the people, which clause 65-66 explicitly forewarns. 29 The end of the third major complex is not as clearly discernible as the end of the second major complex, because the third major complex is closely related to the next small complex (clause 68-72), shown in Figure 3-7 below. Clauses 66 to 69 are similarly dwelling on the global nature of the crisis, and could be analysed as being in some extending relation to each other. Figure 3-7: Logical complexes in detail – part 4 However, the small complex of clause 68-72 is also arguably part of the last complex of clause 73-80 (see Figure 3-7 above), since it begins the discussion of solutions to the crisis. Hence the clauses 68-72 are best analysed as separate from either of the complexes before and after it, as a complex by itself. This is actually quite meaningful, because, within this complex, Smick is doing the same thing: pushing away the opinions of others and presenting his own as more correct. In clause 68-69 he dismisses the policy leaders’ perspective, and in clause 70-72, he points out “It’s Y” by starting off with “It’s not X”. This is why clause 68-69 and clause 70-72 are linked by a “extension/additive/positve” relation. 30 The article ends with the logico complex of clause 73-80 (see Figure 3-7 in previous page), wherein he describes the measures that financial institutions and governments should take to recover from the crisis. There are three enhancing “means” relations (see Figure 3-7), it seems as if the author here is addressing the second part of the title of his article: clause 3: “how to stop it” (it = the crisis). Looking back at Figure 3-2 on p.24, where all enhancing relations were marked “×”, we find the last complex is almost entirely made of enhancing relations (as opposed to elaborations/extensions). This seems to be a pattern in the discourse – ending with enhancing relations – the ending sections of the first, second and third major complexes are also almost purely linked by enhancement relations (see Figure 3-2). The author builds up logical force, usually of the causal type, towards the end of major parts of the discourse. 31 3.2 THEME Theme analysis (Halliday 1994: 37-39; Fries 1995: 1-4) describes the textual organization at the clause level. The Theme is the point of departure of the message conveyed by the clause, the way the clause is introduced; it gives the reader a sense of what the message is about. A clause can be divided into two portions, Theme and Rheme, and the division is determined by the author’s choice of Topic (Topical Theme), which in English is the first transitivity element in the clause, i.e. the participant, process or circumstance. Before the Topic there may be further textual and interpersonal Themes, and the remainder of the clause after the Topic is Rheme. 3.2.1 Wave Metaphor for the Textual Metafunction The flow of the discourse gives rise to some interesting thematic dynamics. The continuous alternation between Theme and Rheme has been likened to a wave by Matthiessen (1992), consisting of peaks of prominence (Theme) and troughs of nonprominence (Rheme). We begin by examining the very start of the series of waves. The first clause in the text has “credit crisis” in its Theme (see Table 3-1 below), orientating the entire article towards it. This is a direct and head-on approach to the subject of the credit crisis (The direct gaze of David Smick in the image accompanying the text in Figure 2-1, p. 5, reinforces this effect of directness). The fact that the Theme is marked brings about further textual prominence. Clause No. Clause, with Theme underlined 7 At this point [in the credit crisis], at least one thing is certain Table 3-1: Theme analysis of first clause 32 The thematic progression (Theme ↔ Rheme progressions) in the discourse that follows from clause 7 can be analysed by retrieving all clauses in the text with occurrences of the word “crisis”. At the beginning of the text, the word “crisis” starts off appearing in the Theme, and thereafter it appears in the Rheme. This signals the change in phase of a wave that builds up sharply then subsequently relaxes out. At the end of the discourse, however, “crisis” reappears in the Theme. Clause Location of Clauses, with Theme underlined, and nouns containing No. word “crisis” in bold 7 Theme At this point [in the credit crisis], at least one thing is certain 17 Theme6 This is because the credit crisis reflects something [[ {which is} more fundamental than a serious problem [of mortgage defaults] ]]. 21 Theme Therefore, the housing crisis was a mere trigger [for a collapse [of trust [in paper]]] Rheme After all, the U.S. banks alone so far 50 have lost upwards of $2 trillion from their collective asset base. 60 Rheme The great uncertainty is [[whether government has the power [[to rescue the financial system]] in times of crisis.]] 65 Rheme Given such massive exposure, government guarantees [in a time of crisis] become meaningless. 69 Theme when the solution [to the credit crisis] will be global. Table 3-2: Theme analysis of Set A - clauses with the word “crisis” 6 “This is because” is considered as a structural conjunction here on the basis that two statements of the form “X. This is because Y.” can always be written as “X because Y”. In other words, we treat the process “is” as non-congruent. However, the other parts of this thesis will treat it as congruent. 33 The same pattern appears if we look at all clauses with the word “credit”: Clause Location of Clauses, with Theme underlined, and nouns containing No. word “credit” in bold 7 Theme At this point [in the credit crisis], at least one thing is certain 16 Theme and credit markets remained frozen. 17 Theme6 This is because the credit crisis reflects something [[{which is} more fundamental than a serious problem [of mortgage defaults] ]]. 45 Rheme the world will face a serious credit crunch in 2009 52 Rheme Translation : [[The U.S. financial system will have a whopping $15 trillion to $20 trillion less credit available next year [[than {what} was around a year and a half before]].]] 59 Theme Apart from the economic pain [[resulting from shrinking credit markets]], we are about to see an earthquake [in the relationship [between government and financial markets]]. 69 Theme when the solution [to the credit crisis] will be global. Table 3-3: Theme analysis of Set B - clauses with the word “credit” There is a return to the Theme at the end, represented by clause 69 in both Set A and Set B. By turning to the system of cohesion - Halliday & Matthiessen (2004), a plausible explanation for the patterns in Table 3-2 and Table 3-3 may be found by combining the lexical string of “crisis” (String 1) and the lexical string of “credit” (String 2), in Figure 3-8 next page: 34 Figure 3-8: String 1 and String 2 – lexical strings of “crisis” and “credit”, respectively Numbers in dark blue on the left denote clause numbers in which the words appear. The system network for lexical string analysis is in Appendix 7.3 p. 419. String 1 and String 2 represent two potentially divergent streams of cohesive meaning that spring forth from the first clause in the discourse, clause 7. They represent the pathways that are taken to move forward and expand upon the semantic space opened up by the word “credit crisis” in the Theme of clause 7. The most salient feature of the above analysis is that the two strings merge at the front (clause 7, 17) and the end (clause 69). This pattern parallels the Theme-Rheme dynamics seen previously in the tables of Set A and Set B, because both represent a kind of initial build-up, subsequent release, and return to the initial state. More remarkable is how precise the point of divergence of String 1 and String 2 (at clause 17 - 21) matches the 35 point in the discourse where the ‘thematic pulses’, the words “crisis” and “credit”, migrate from Theme to Rheme. The wave metaphor for describing the textual metafunction provides further insight to the analysis. The flow of textual meaning in the discourse begins with “credit crisis” in the Theme for the first clause, and this represents the crest of a pulse of prominence, the entry point into the semantic space. As the discourse progresses from clause 7 - 21, “credit” and “crisis” gradually transit into the Rheme; we reach the trough of the wave. This is the view from thematic analysis of the clause, but this is a limited view of textual dynamics because while it sees the shape of the wave, it does not see its composition. Nevertheless the notion of composition is important: the clauses that compose the thematic waves are themselves composed of words which also compose lexical strings, streams of unfolding meaning in the form of taxonomic relations, that travel from clause to clause like the thematic waves do. This allows us to conceive of the thematic waves as being composed of lexical strings, not just structurally, but also semantically, since not only do words exist in the Theme and Rheme, but also, each word exists within a semantic space associated with it, and the lexical string represents the author’s choice of movement through that semantic space, along the discourse. The thematic progression also represents a kind of movement through semantic space. Lexical strings can be considered as streams of meaning that influence the flow of thematic waves. With this wave metaphor, the lexical strings in Figure 3-8 can then be construed as streams responsible for the upwelling and subsiding of thematic waves. The concentration of the two streams as one at first, from clause 7 to 17, concentrates 36 the semantic space, resulting in high prominence, as the theme analysis reveals. The subsequent divergence of the stream in clause 17 to 21 dissipates this prominence, resulting in a thematic shift to the Rheme. This is representative of the nature of the text – the author begins with the theme of “credit crisis”, then moves in to elaborate on the origin and development of the crisis, and the participants in the crisis takes over as the Topical Theme, e.g. “the US banks” in clause 50 - see Table 3-2 (p.33) and “the world” in clause 45 - see Table 3-3 (p. 34). As the end of the discourse draws closer, see Figure 3-8 (p.35) near clause 69, the streams converge again, and this reconcentrates the semantic space, regenerating the peak of prominence as represented by the Theme in clause 69. 3.2.2 Short Range Theme - Rheme Dynamics Previously we examined the thematic progressions of selected participants through the discourse, which often occur over long distances, but this overlooks the short range dynamics between clauses that immediately follow each other. This section focuses on the short range dynamics between successive clauses, ignoring the longer distance thematic movements. Table 3-4 over the next two pages will show thematic dynamics from clause number i to i + j, for all i, for j=1. (In contrast, j could be larger than 1 in Figure 3-8, even up to 29.) At such close range, the lexical associations become less important than the reference chains, so only reference chains are considered in the analysis here. Note that the items in a reference chain need not be participants of transitivity, but non-participants of transitivity are possible too, e.g. ellipsed verbs. Here “participants” just refers to participants in reference. Also, in this analysis, references that take entire ranking clauses as their referent are not included. 37 Clause to Clause 7 8 --> --> 8 9 9 10 11 --> --> --> 10 11 12 12 13 --> --> 13 14 14 15 16 17 18 19 --> --> --> --> --> --> 15 16 17 18 19 20 20 21 22 23 24 25 26 --> --> --> --> --> --> --> 21 22 23 24 25 26 27 27 28 --> --> 28 29 29 30 31 32 33 --> --> --> --> --> 30 31 32 33 34 34 --> 35 35 36 37 --> --> --> 36 37 38 38 --> 39 39 40 41 42 43 --> --> --> --> --> 40 41 42 43 44 44 45 46 47 --> --> --> --> 45 46 47 48 TH --> TH TH --> RH RH --> TH RH --> RH policymakers --> Those bailout package --> This magic pill --> which which --> taxpayer-funded bailout financial paper instruments --> these paper instruments collapse --> {} bankers --> the bankers simple loans --> them the bank -> {} the bank -> {} its loans --> {} its loans --> them financial salad --> piece of lettuce piece of lettuce (with toxic waste) -> piece of salad overall salad --> salad global investors --> No one financial markets -> The markets 38 Clause to Clause 48 49 50 --> --> --> 49 50 51 51 52 53 54 55 56 --> --> --> --> --> --> 52 53 54 55 56 57 57 58 59 --> --> --> 58 59 60 60 --> 61 61 62 63 64 65 66 --> --> --> --> --> --> 62 63 64 65 66 67 67 --> 68 68 69 70 --> --> --> 69 70 71 71 --> 72 72 73 74 75 76 --> --> --> --> --> 73 74 75 76 77 77 --> 78 78 --> 79 79 --> 80 TOTAL instances TH --> TH TH --> RH RH --> TH RH --> RH the U.S. Banks --> Most banks is-->{} the banks --> they lend --> do it Warren Buffet --> who government --> government whether government has the power to rescue the financial system in times of crisis -->it we--> American workers American workers-> Our policy leaders in Washington the world --> the world's money the world's money--> more than $6 trillion are sitting --> {}, the sidelines --> idle global money markets bank clearing facility -->bankers The bankers --> The central banks 8 2 9 13 Table 3-4: Thematic progression in successive clauses 39 The sparse distribution of these progressions means that most of the thematic progressions in the text are long-range. The occurrence of progressions on multiple scales in the text makes the text highly complex, a reflection of the complexity of the credit crisis itself. Among the short range progressions, the RH RH progression is the most frequent; there are 13 of them. There is a section of the text with a particularly high concentration of RH RH progressions: clause 26-35. This is where the author uses the salad analogy, and reflects the intensely local nature of the thematic progression at that point in the text. This is a textual effect that focuses the reader. An event as complex as the financial crisis involves many participants, introduced in all manner of ways into the text, creating a jungle of cohesive links, but during the salad analogy, this jungle is set aside to focus on what is immediately in front of the reader: the banks and the loans (the salad). The “salad analogy” section ranging from clause 28 to 41 is a special part of the text that deserves closer inspection, see Figure 3-9, next page. The salad analogy can be split into two sections. The distinction between the two sections is not well defined, but there is a gradual transition. The first section, clause 28-36, deals with the how and why of making the salad, and the clauses have Themes involving single entities (“a bank” in clause 30, “you” in clause 36), and describe a hypothetical, fantasy scenario. In the second portion of the salad analogy, clause 37-41, the description moves towards reality, highlights actual events that have happened (investors stopped ordering salad and distrust heightened) and the Themes of the clauses now feature universal/global entities (“global investors” in clause 38, “No one” in clause 39, “global interest rates” in clause 41). 40 Figure 3-9: The salad analogy (Theme underlined) Moving back to Table 3-4 statistics on p. 39, we find the progression RH TH is also fairly frequent. There are nine of these, distributed fairly evenly in the text. This is not surprising given that many clauses have large Rhemes, which contain many participants, and the author can choose to thematise anything in the previous clause’s Rheme. An example of this is given in Table 3-5: Clause Clause with Theme underlined, reference items in bold No. 19 In recent years, our banks, borrowing to maximize the leverage of their assets at unheard-of levels, produced mountains of financial paper instruments (called asset-backed securities) with little means of measuring their value 41 20 Incredibly, these paper instruments were insured by more dubious paper instruments. Table 3-5: Example of RH TH: R19 T20 Next in terms of frequency is TH TH (eight occurrences). This is an important thematic progression, because the prominence of repeating a Theme. TH TH occurs throughout the text, and is fairly spread out. But in the places where they occur, they generate intensity to the reader because of their prominence. They are therefore used only on the most important participants in the text: the policymakers, the banks, and the world’s money (see Table 3-4, p. 38-39). Finally, we turn to TH RH (two occurrences). This is surprisingly rare in the text. TH RH is used fairly commonly in general, to provide further descriptions of participants introduced in the Theme, but this is seldom a style the author adopts in this article. Predictive Model To better understand the frequencies of the various progression types, we can try to build a simple predictive model based on a few assumptions. TH TH 8 TH RH 2 RH TH 9 RH RH 13 Total 32 Table 3-6: Actual frequencies of short range thematic progression types The statistics of RH RH as the most frequent short range progression in 42 Table 3-6 is really not unexpected. Considering the Rheme is almost always a larger element than the Theme and therefore likely to contain more participants, and assuming all participants have equal probability to progress to all other participants further down the discourse, RH RH progressions are more probable, and should be more frequent than TH TH progressions. To develop and test this argument quantitatively, let the average number of participants in the Theme be NT and that for the Rheme be NR. Suppose the Rheme contains on average k times as many participants as the Theme: NR = k NT The potential for a short range progression to occur is taken to be proportional to P, defined as the total number of possible progressions, calculated by multiplying the average number of participants in the referred element in the preceding clause and the average number of participants in the referring element in the succeeding clause. Note that this assumes only one progression occurs between successive clauses. With this, potential P for different possible short range thematic progressions are, (for TH (for TH (for RH (for RH TH) RH) TH) RH) PTT = NT NT = NT2 PTR = NT NR = kNT2 PRT = NR NT = kNT2 PRR = NR NR = k2NT2 Based on these assumptions, the frequencies of progressions TH TH, TH RH, RH TH and RH RH should be in the ratio 1:k:k:k2. For example, if k = 2, the ratio of progression possibilities is 1:2:2:4, as illustrated by Figure 3-10: 43 44 Figure 3-10: Thematic progression possibilities for k = 2 For the text in the CNN article, calculations reveal that k = 1.65 (see Appendix 4: Theme/Rheme Calculation p. 385), so the ratio is approximately 1 : 1.7 : 1.7 : 2.7. Normalising this distribution to the total count of 32 gives the following predicted frequency distribution in Table 3-7 (rounded to one decimal place): TH TH 4.5 TH RH 7.5 RH TH 7.5 RH RH 12.4 Total 31.9 Table 3-7: Predicted frequencies of short range thematic progression types A comparison of Table 3-7 with Table 3-6 (p. 42) shows the model makes a few accurate predictions; it arrives at the count of RH RH as 12.4, almost exactly the correct figure of 13, and the count of RH TH as 7.5, close to the actual figure of 9. This accuracy is remarkable because of the range of assumptions of the model and the approximate nature of the quantitative measures used. This means that on a statistical average, progressions from Rheme are approximately random - there is no preference, at each instance of clause to clause thematic movement, to progress from RH to TH or from RH to RH. 45 More significant are the deviations from prediction. There is an overabundance of TH TH (actual: 8, predicted: 4.5), while there is an underoccurrence of TH RH (actual: 2, predicted: 7.5). This means that the thematic progressions that start from the Theme element prefer progressing to the Theme rather than the Rheme. Some cases of thematic progression that do not fall under the four categories discussed above are the clauses with a completely anaphoric element as Theme, (coded differently in Systemics as “*/Topic/WH/Theme” or “*/Topic/TH/Theme”, see Appendix 7.1.1 System of TH p. 414) that refer back to a substantial portion of text that may extend beyond a single ranking clause. Some examples of these cases are listed in Table 3-8: Clause Clause with “referring” Clause Referred text No. Theme in bold No. 17 This is because [[the credit 15 Instead, stock markets collapsed 16 and credit markets remained crisis reflects something [[ {which is} more fundamental than a serious frozen. problem [of mortgage defaults] ]]]]. 64 This means [[government 62 In the United Kingdom, , the collected assets [of even if it wanted || to {bail the major banks] are four times out the system}]]. the nation's gross domestic product (GDP). 63 A similar situation exists in many Euro zone countries. 46 Clause Clause with “referring” Clause No. Theme in bold No. 79 And while that is 78 happening, Referred text The central banks, , need to begin to refashion the world's financial architecture. Table 3-8: Thematic progression with Theme substitution Textual progressions such as those in Table 3-8 as indicative of points in the text where the method of development, a concept described by Martin (1992) and Fries (1981), is more local. This allows the packaging of a large passage of preceding ideas into a participant in a single clause, achieving a kind of “textual focussing effect”, generating argumentative force as well as cohesion. Using the wave metaphor for the thematic flow, not only does a substituted Theme represent a single peak of prominence, but represents a previous wave as well. Going back to Table 3-4 (p. 39), let’s consider the progressions that involve changes in thematic status, i.e. movements from Theme to Rheme (two occurrences) or Rheme to Theme (nine occurrences). That there is such an imbalance should not be surprising, for the movement RH TH is the hallmark of the most elementary thematic progression out of the three main types described by Daneš (1974) and adapted by Fries (1981, 1995) as shown in Figure 3-11: 47 3.2.3 Thematic Progression (Daneš) Figure 3-11: Three major patterns of Thematic Progression (Daneš 1974): (TP1) Simple linear progression (TP2) Constant Theme (TP3) Derived Themes Legend: Progression within a clause (by experiential relation) Progression between clauses (by direct/indirect reference) The movement TH RH described in Table 3-4 (p. 39) however, does not form part of any of the main types of thematic progression in Figure 3-11, and is indicative of a much rarer kind of method of development of the text. The movement TH TH (eight occurrences), is about as common as RH TH, and marks the second pattern of thematic progression in Figure 3-11: TP2. But most common in the text is RH RH (13 occurrences), which also does not fall under any of Daneš’ (1974) main types. 48 3.2.3.1 Thematic progression TP3 Applying the framework in Figure 3-11 to our text, we find that the overwhelming majority of thematic progression in the text, on all scales, is TP3. Table 3-9 presents a first example of TP3: Clause Thematic No. Progression 76 [T] T76 Clause with Theme in bold [T] We need a private/public global bank clearing Participants facility. involved in the crisis 77 [T] T77 The bankers don't trust each other 78 [T] T78 The central banks, , need to begin to refashion the world's financial architecture. 80 [T] T80 the major governments [of the world], , should begin major fiscal efforts [[to stimulate their weakening economies]]. Table 3-9: Short range Thematic Progression type 3 (TP3) with implicit [T] Clause 76, 77, 78, 80, which conclude the article, have Themes derived from the context of the financial crisis, the participants and entities that are involved in the crisis. [T] does not explicitly appear in the preceding text (though each participant, “we”, “banks”, “bankers” and “governments” have been individually mentioned in some earlier part, their earlier appearances are not necessary for the textual development in clause 76 to 80). Table 3-10 presents further examples of TP3, but with [T] that have explicit appearance in the text: 49 Clause Thematic No. Progression 33 [T] 34 [T] T33 T34 Clause with Theme in bold [T] The idea was [[to sell “securitization” individual servings [of diversified financial (Rheme of salad] around the world]]. clause 27); the The only problem : [[[under an occasional making of piece of lettuce] was a speck of toxic waste “salad” (clause [in the form of a defaulting subprime 28-32) mortgage]]]. 73 75 [T] [T] T73 T75 The challenge will be [[to reform our “the solution to financial system quickly]] the credit The first step should be efforts [[to make crisis” (Theme the market [for future asset -backed paper] of clause 69) more transparent and credible]]. Table 3-10: Short range Thematic Progression type 3 (TP3) with explicit [T] Clause 33, 34 follow immediately from the author’s salad analogy; the Themes essentially refer to the salad – “the idea” (behind the salad) and “the only problem” (with the salad). The effect of using TP3 here is that the text no longer focuses on the salad itself, but brings the fact that it’s a problem to the forefront of attention. The curt appearance of “problem” as a Theme by itself seems to be akin to throwing in a wedge into the discussion, putting a stop to all the elaborate material description about the salad making. Clause 73, 75 expand upon “the solution to the credit crisis”, with the Themes “the challenge” (in resolving the crisis) and “the first step” (in resolving the crisis). However, the function of having TP3 here seems to be the reverse of its use in clause 33, 34 – instead of stopping a train of textual development of salad making, in clause 73, 75 it kick-starts the discussion about the solution to the crisis. 50 Clause 33, 34, 73, 75 are four key examples of inverted Token-Value constructions, a striking pattern in the text that the author habitually employs. These constructions will be examined again, in greater detail, in section 3.3: Transitivity, p. 62. So far we’ve examined TP3 on a short range, but TP3 also appears frequently on larger scales in the text. Table 3-11 presents some examples: Clause Clause with Topical Theme in bold No. 15 Instead, stock markets collapsed 18 Global investors, , have declared a buyers' strike [against the sophisticated paper assets [of securitization] [[that financial institutions use {the sophisticated paper assets [of securitization]} || to measure || and offload risk]]]. 41 when global interest rates began to rise. 66 Yet because of the interconnected web [of global financial relationships], we are all vulnerable to the threat. Table 3-11: Long range Thematic Progression type 3 (TP3) The widespread use of TP3 can be attributed to the fact that the topic of discussion, the financial crisis, involves a large number of inter-related entities, participants and concepts, described in a financial jargon of which knowledge is assumed. Hence it is easy to employ a multitude of specific themes that derive from the overarching theme of a financial crisis, without prior mention. When clause 15 first introduces “stock markets” into the text, there is no prior mention of stocks or markets. The knowledge is assumed from the context of discussion or from our shared culture. Clause 18 brings in “global investors”, which 51 can be argued to derive from the implicit overarching theme of “credit”, and similarly for clause 41, which introduces “global interest rates”. A “web of global financial relationships” enters into the text in clause 66. Arguably, the examples of TP3 given in Table 3-11 can be considered alternatively as TP1 or TP2 which involve shorter range cohesive links to nearby clauses. For example, “global interest rates” in clause 41 can be analysed as progression by bridging reference from “salad” (essentially meaning “loans”, on which the interest is imposed upon) in clause 38. However, such a localized view of progression may fail to capture the global thematic development and the hierarchical nature of the cohesion of the discourse. Though TP3 is the dominant mode of thematic progression, there are notable points in the discourse that clearly progress by TP1 and TP2. 3.2.3.2 Thematic Progression TP1 Clause 11-13 is the earliest example of TP1 in the text. It is used to facilitate the portrayal of the “bailout”. Clause Thematic No. Clause (bold: Theme; underlined: the part of Rheme that Progression progresses) Begin with the U.S. Treasury's $700 billion bailout package. 11 12 R11 T12 This was presented as some magic pill 13 R12 T13 which, if gulped down, would quickly restore financial stability. Table 3-12: Thematic Progression type 1 (TP1) 52 The process of “bailout” is deliberately introduced by Smick in the form of a complex grammatical metaphor, the “US Treasury’s $700 billion bailout package”, so that it can be convincingly re-expressed as a physical entity in the text, a “magic pill”, that is “gulped down”. TP1, being a linear progression, allows “bailout” to be stretched across three clauses, extending its initial introduction, and allows “bailout” to gradually progress, via transitivity and cohesion, from a process in the real financial world into a metaphorical pill that the author ultimately intends as a parody. 3.2.3.3 Thematic Progression TP2 Clause Thematic No. Progression : most policymakers lack a clue of what is really at stake. 8 9 Clause with Theme in bold T8 T9 Those with some knowledge are driving policy After all, the U.S. banks alone so far during the crisis have 50 lost upwards of $2 trillion from their collective asset base. 51 T50 T51 Most banks are leveraged by more than 10 to 1. Table 3-13: Thematic Progression type 2 (TP2) Clause 8-9 is the first example of TP2 in the text, and is employed here to facilitate the introduction of “policymakers” into the text. In clause 50-51, TP2 allows the same participants, US banks, to be developed in parallel in two clauses that essentially present two separate facts, so that they can be easily combined later into a single consequence in clause 52: 53 Clause Clause No. 52 Translation: The U.S. financial system will have a whopping $15 trillion to $20 trillion less credit available next year than was around a year and a half before. Table 3-14: The consequence of clause 50 & 51 While the examples of TP2 in Table 3-13 involve clauses with simple Themes, Table 3-15 presents a more complex TP2, involving a complex nominal Theme: Clause Thematic No. Clause with Theme in bold, constant element underlined Progression Before the last decade, bankers simply lent in the form of 26 syndicated loans. 27 T26 T27 But with the huge expansion [of the global economy] in the 1990s, , the bankers came up with an idea [[called securitization]]. 27’ T26 T27’ But in the last decade, the bankers came up with an idea [[called securitization]]. Table 3-15: Thematic Progression type 2 (TP2) with complex Theme The fact that it’s a TP2 between clause 26 and 27 is more difficult to recognise, since there is nothing constant about T26 and T27 in their entirety, because T26 is a point in time, while T27 is a nominalised process of “expansion”. The constancy is more apparent if we consider instead the pair: clause 26 and clause 27’, a reduced version of clause 27. The basic progression, in clause 26 clause 27’, forms the backbone onto which the author hinges the extra fact “the expansion of the global economy produced an ocean of new capital”, to give the more complex progression clause 26 clause 27. The repetition of themes in TP2 provides textual prominence, but there is also the sense of repetition in the experiential metafunction with 54 “expansion” and “ocean” in clause 27, so the resulting prominence is multiplied. 3.2.4 Marked Themes There are 12 clauses with marked Themes in the text, including both circumstances of transitivity as well as dependent beta clauses. A large number of them realize temporal meaning, see Table 3-16. These include the very first and last clauses in the text. These clauses are concentrated in the introduction of the text (clause 7, 19, 26, 27) where the author introduces key events that led to the financial crisis: banks borrowing on a massive scale and insuring their assets. Clause Theme Rheme No. 7 At this point [in the at least one thing is certain. credit crisis], 19 In recent years, our banks, , produced mountains [of financial paper instruments ([[called asset -backed securities]])] with little means [[of measuring their value]]. 26 Before the last bankers simply lent in the form of syndicated loans. decade, 27 But with the huge , the expansion [of the bankers came up with an idea [[called securitization]]. global economy] in the 1990s, 79, 80 And while that is the major governments [of the world], , should begin major fiscal efforts [[to stimulate their weakening economies]]. Table 3-16: Marked Themes functioning as location in time 55 The presence of these marked temporal Themes anchors key points in the article well, grounding the discourse at key points in time and making the text progress as if everything was down to earth, and factual. In fact all five clauses in Table 3-16 can even be considered belonging to the same thematic progression, of type TP3, where the derived theme [T] is a historical timeline of events. 3.2.5 Interpersonal Themes There are only four clauses in the text with interpersonal Themes. Two are shown in Table 3-17, the rest in Table 3-19. Clause Interpersonal No. Theme 20 Incredibly, 55 True, Topical Theme Rheme these paper were insured by more dubious paper instruments instruments. the banks will still lend Table 3-17: Interpersonal Themes The interpersonal Themes in both clauses in Table 3-17 are conveying the author’s incredulity; in clause 20 it is towards past events, and in clause 55 it is towards future events. Although the Mood Adjunct “True” in clause 55 functions as an adjunct of positive polarity within the clause, but seen in the context of the discourse, e.g. with clause 56 (Table 3-18), it actually serves to convey the author’s sense of distrust towards the willingness of banks to lend in the wake of the crisis. Clause Clause No. 56 -- but the fear is they will do it only to people such as Warren Buffett, Table 3-18: Clause 56 (in opposition to clause 55) 56 The effect of the interpersonal Themes in clause 20 and 55 is that the author displays his negative attitudes towards wrongful practices leading to and resulting from the financial crisis, and in doing so, he positions himself as someone who has a deep understanding of the financial system. The other two clauses in the text with interpersonal themes are clauses 70 and 71: Clause Interpersonal Topical Theme Rheme No. Theme 70 It is not that the world lacks money; 71 it is that the world's is sitting on the sidelines money Table 3-19: Interpersonal Modal-Metaphors as Theme Modality construes the area of meaning that lies between ‘yes’ and ‘no’, the fine gradation between two opposite poles. Now, the statements in clause 70 & 71 lie precisely at the ends of the pole; clause 70 reads as “not true” and clause 71 as “true”, hence they are very closely related to the construal of modality. Therefore the elements “it is not that” and “it is that” are analysed here as interpersonal metaphors of modality (they should actually be “metaphors of polarity”). The reason for treating both instances of “is” as metaphorical is that they cannot be congruent processes since “it” is a nonrepresentative “it”, so no attribution or identification takes place. Also, I have not considered clause 70 & 71 as predicated Theme constructions because such an analysis is problematic as there is no transitive element that is predicated. Admittedly, the functionality of “it is” in clause 70 & 71 bears some similarity to predicated Theme constructions because it seems to be foregrounding 57 something, i.e. the polarity of the statements that follow (or their Finites). For a discussion of these classification issues, see Collins (1991: 34). Collins classifies such clauses as clefts (predicated themes) but admits that “convincing evidence (for doing so) is difficult to find.” Ultimately, the elements “it is not that” and “it is that” function to create a contrast of polarity, and this contrast functions to highlight the truth of clause 71. 58 3.3 TRANSITIVITY Transitivity analysis (Halliday 1994, chapter 5) of the CNN main story reveals that the process types are primarily relational or material, as shown in Table 3-20. In this section we will explore the diversity of process types, their distribution and their functional properties. Relational 36 Material 29 Mental 6 Behavioural 1 Verbal 1 Existential 1 Total 74 Table 3-20: Transitivity types in the text 3.3.1 Relational Clauses The most common type of transitive process is the relational process. There are 36 relational processes in the text, divided into 21 attributive and 15 identifying types. Table 3-21 below lists all the attributive relational clauses, with some useful further sub-divisions: Relational Attributive (21) Intensive (10) 7 16 21 36 37 61 62 65 66 67 At this point in the credit crisis, at least one thing is certain and credit markets remained frozen. Therefore, the housing crisis was a mere trigger for a collapse of trust in paper, and you're dead. The overall salad looked delicious, It seems doubtful. In the United Kingdom, for example, the collected assets of the major banks are four times the nation's gross domestic product (GDP). Given such massive exposure, government guarantees in a time of crisis become meaningless. Yet because of the interconnected web of global financial relationships, we are all vulnerable to the threat. The collapse of, say, a major European bank would hardly leave American workers immune. 59 Possessive (7) 8 47 48 50 57 70 76 : most policymakers lack a clue of what is really at stake. to remove the toxic salad from bank balance sheets. Policymakers have no means of forcing the banks to start lending After all, the U.S. banks alone so far during the crisis have lost upwards of $2 trillion from their collective asset base. who don't need loans. It is not that the world lacks money; We need a private/public global bank clearing facility. Circumstantial (4) 45 69 71 72 the world will face a serious credit crunch in 2009 when the solution to the credit crisis will be global. it is that the world's money is sitting on the sidelines -- more than $6 trillion in idle global money markets alone. Table 3-21: All relational attributive clauses in the text (underlined: Attribute) Most of the attributive relational processes are intensive, and the Carriers include important entities involved in the crisis, including “you” (clause 36) and “we” (clause 66). The Attributes are mostly something negative, i.e. “frozen”, “collapse”, “dead”, “doubtful”, “meaningless”, “vulnerable”, generating an overall tone of gloom and doom in the text. It is interesting to note that all the circumstantial attributive clauses have “the world” or “global” in it, reflecting the scale of the circumstances of the financial crisis. The possessive attributive clauses show a very interesting trend – six out of seven are negative in polarity in some form of realisation: either the negativity is encoded in the Finite: “don’t need”; encoded lexically: “lack”, “remove”, “lost”; encoded in the complement: “no means”, or in a modal-metaphorical clause: “it is not that”. This predominance of negative possessive polarity seems to reinforce the notion in clause 50 that the crisis has generated losses of staggering proportions, and to convey a sense to the reader that “whatever we have now is in danger of being lost”. This correlation between the negative polarity and the possessive type clause 60 is all the more significant, considering that for the large number of clauses in the text, (74 ranking + 25 rankshifted + 8 interrupting = 107 total clauses, and excluding 27 non-finite clauses: 80 total finite clauses) negative Finites only ever appears four times (see Table 3-22), and two of these occur in the possessive clauses mentioned in Table 3-21 in the previous page. Clause Location of Finite Clause (negative element in bold) No. 57 ranking clause who don't need loans. 64 embedded clause This means [[government cannot bail out the system even if it wanted to.]] 70 ranking clause It is not that the world lacks money; (in Modal Metaphor) 77 ranking clause The bankers don't trust each other. Table 3-22: All negative Finites in the text Next we examine the second common type of relational process in the text: the identifying relational processes, in Table 3-23: Relational Identifying (15) Intensive -encoding (7) 33 34 56 58 60 73 75 The idea was to sell (for huge fees) individual servings of diversified financial salad around the world. The only problem: under an occasional piece of lettuce was a speck of toxic waste in the form of a defaulting subprime mortgage. -- but the fear is they will do it only to people such as Warren Buffett, What is uncertain is the amount of lending to borrowers engaged in entrepreneurial risk, the center of business reinvention and job creation. The great uncertainty is whether government has the power to rescue the financial system in times of crisis. The challenge will be to reform our financial system quickly The first step should be efforts to make the market for future asset-backed paper more transparent and credible. Intensive -decoding (5) 12 24 42 52 64 This was presented as some magic pill So what are these paper instruments, these asset-backed or mortgage-backed securities? So what does this salad boycott mean for the future Translation: The U.S. financial system will have a whopping $15 trillion to $20 trillion less credit available next year than was around a year and a half before. This means government cannot bail out the system even if it wanted to. 61 Circumstantial -encoding(3) 17 20 22 This is because the credit crisis reflects something more fundamental than a serious problem of mortgage defaults. Incredibly, these paper instruments were insured by more dubious paper instruments. {which was} followed by a de-leveraging of the entire global financial system. Table 3-23: All relational identifying clauses in the text (underlined: Value) Halliday (1967, 1968) and Halliday & Matthiessen (2004: 230) define encoding and decoding types of identifying clauses, based on whether the Token is an Identifier (encoding) or an Identified element (decoding). Looking at Table 3-23, the encoding clauses seem to be the ones making claims, while the decoding clauses are ones which explain things. There is a preponderance of encoding (ten of them) over decoding clause types (five of them). The separation of encoding/decoding clauses in Table 3-23 reveals striking patterns: (1) all the ten encoding clauses have the same thematic structure: with Value as Theme, in an inverted Token-Value configuration, and (2) all seven intensive encoding clauses have the same mood structure: the Token takes the form of a rankshifted clause. The reason for these striking patterns is an open issue, but it is possible that, for (1), the reason is to enable the text to develop by a more abstract thematic progression, namely TP3, which generates cohesion on a larger scale, and for (2), embedding propositions makes them less directly negotiable. The seven intensive encoding clauses will be discussed again in Chapter 5: Singular Value Decomposition (SVD), section 5.1.3: Feature F2, p. 209. An interesting trend in all the identifying clauses is that the author never identifies any participant or entity in the crisis – the identified element is always either a textual pronoun (e.g. “this”), or some abstract derived notion (e.g. “the idea”, “the 62 challenge”), with only few exceptions: (1) when the author is asking a question (clause 24, 42) – in which case he really isn’t identifying anything, but asking for things to be identified; and (2) some processes of the circumstantial types – the embedded processes of clause 17 & 20. This is a skilful tactic by Smick because the things which he avoids identifying directly are complex – “paper instruments”, “crisis”. Perhaps this also reflects his creative writing style, providing information in metaphors and detours rather than directly. Consider for example clause 34, which has encoding processes at both “T1” and “T2” levels. Figure 3-12: Identification in clause 34 Idd: Identified Idr: Identifier The choice of an encoding rather than a decoding configuration in clause 34 means that instead of identifying the “toxic waste” or “defaulting sub-prime mortgage”, the intangible entities: its location and “problem” gets identified. The author therefore avoids identifying the “real” participants. 63 Another significant difference between the encoding and decoding types is in the ergative analysis (see section 3.4: Ergativity p. 80 for an introduction to ergative elements). According to Halliday & Matthiessen (2004: 292), encoding clauses are in effective Voice, with Token as Agent and Value as Medium, and decoding clauses are in middle Voice, with Token as Medium and Value as Range. This means that encoding clauses, being in effective Voice, can be expressed in passive Voice (while for most decoding clauses, the passive form is strained and rare), allowing intangible Values to be foregrounded as Theme. For clause 34, (see Figure 3-12 on previous page) the author exploits this option twice, by passivising both “T1” and “T2” levels. As a result, the first three participants a reader encounters upon reading clause 34 are “problem”, “piece of lettuce”, and “speck of toxic waste”, all of which are abstract or metaphorical, and the reader only gets to the most recognisable entity “defaulting subprime mortgage” at the end of the clause. Smick appears to be building a pyramid of abstraction upon the familiar entity. In fact, not only clause 34, but all of the encoding clauses (except for nonfinite clause 22) are in passive Voice, as a result of their inverted Token-Value configuration. This seems to mask the agentivity in them, making the direction of identification and coding less direct and less obvious (compared to clauses in active Voice). But the difference in ergativity of encoding and decoding clauses is a very subtle distinction, as even Halliday & Matthiessen (2004: 294) cautions: “The ergative analysis of relational clauses is complex. In the attributive, the 64 Attribute is clearly analogous to a Range; but in the identifying the criteria tend to conflict. For the purposes of simplicity, we will interpret the Token as Medium and Value as Range in all types, although this does ignore some aspects of the patterning of such clauses in the text.” However, for the analysis of my text, this subtle distinction in ergativity seems to be important, because it reveals important aspects about the encoding clauses. Harvey (2001) has investigated the semantic patterns of encoding and decoding clauses, and states that: “The effective ergative structure of encoding clauses reflects the ‘tight-fit’ between the Token and Value (i.e. X=Y; Y=X) and the relation can be interpreted as categorical and exhaustive.” “The middle ergative structure of decoding clauses suggests a ‘loose-fit’ – the relation is more open-ended than the relation constructed in encoding clauses. In other words, decoding clauses, in general, construct a less equivalent relation than do encoding clauses.” To illustrate the difference in ‘tightness’, an example of each type from my text is discussed below: Clause No. Clause (italicized: my interpretations based on the clause alone) 73 ENCODING The challenge will be to reform our financial system quickly (and [the challenge will be] nothing else) 64 DECODING This means government cannot bail out the system even if it wanted to. (and possibly [this means] something else) Table 3-24: Tightness of encoding vs. decoding equivalence relations 65 Harvey (2001) describes the relations of equivalence in relational identifying clauses in technical and technocratic discourse as constructing “an apparently homogeneous world-view, the validity of which is not open to questioning or scrutiny.” This is the reason for Smick’s abundant use of encoding clauses, which construct the tightest equivalence – the construction of a grammatical and semantic fortress of impenetrability around his views. Moving away from our focus on encoding clauses, another property of the relational clauses in general is that many of them (13 out of 36) actually describe material processes, in rankshifted clauses or shifted grammatical metaphors: Relational Identifying (8) Intensive -encoding (6) 33 56 58 60 73 75 The idea was to sell (for huge fees) individual servings of diversified financial salad around the world. -- but the fear is they will do it only to people such as Warren Buffett, What is uncertain is the amount of lending to borrowers engaged in entrepreneurial risk, the center of business reinvention and job creation. The great uncertainty is whether government has the power to rescue the financial system in times of crisis. The challenge will be to reform our financial system quickly The first step should be efforts to make the market for future asset-backed paper more transparent and credible. Intensive -decoding (1) 64 This means government cannot bail out the system even if it wanted to. Circumstantial -encoding(1) 22 {which was} followed by a de-leveraging of the entire global financial system. Relational Attributive (5) Intensive (3) 16 21 67 and credit markets remained frozen. Therefore, the housing crisis was a mere trigger for a collapse of trust in paper, The collapse of, say, a major European bank would hardly leave American workers immune. Possessive (1) 48 Policymakers have no means of forcing the banks to start lending Circumstantial (2) 69 when the solution to the credit crisis will be global. Table 3-25: Relational clauses with rankshifted material processes underlined 66 The author could easily have written the material clauses as ranking clauses, but has rankshifted them as a form of linguistic distancing from real-world events, creating a complex, abstract text, making it appear more professional and authoritative. 3.3.2 Material Clauses This section explores the diversity of material clauses in the text. All the material clauses are listed in Table 3-26 below, with some further sub-divisions: Material Processes (29) with ... No Goal or Range (9) 15 26 40 41 43 53 54 55 79 Instead, stock markets collapsed Before the last decade, bankers simply lent in the form of syndicated loans. This distrust heightened when global interest rates began to rise. and why have financial markets collapsed so brutally? The cost of money is rising and the availability {is} shrinking. True, the banks will still lend And while that is happening, Goal (10) – underlined 9 29 30 31 32 35 49 51 74 78 Those with some knowledge are driving policy and holding them until maturity, a bank collected all its loans together, then {the bank} diced {its loans} and {the bank} sliced them up into a big, beautiful tossed salad. Eat that piece of salad, short of nationalizing the entire financial system. Most banks are leveraged by more than 10 to 1. to draw that global capital back into more productive uses. The central banks, working with the private institutions in providing enhanced data, need to begin to refashion the world's financial architecture. Range (10) - underlined 11 13 14 18 Begin with the U.S. Treasury's $700 billion bailout package. which, if gulped down, would quickly restore financial stability. The "shock and awe" of the sheer size of the taxpayer-funded bailout would somehow restore confidence. Global investors, now on the sidelines, have declared a buyers' strike against the sophisticated paper assets of securitization that financial institutions use to measure and offload risk. 67 19 25 28 38 46 80 In recent years, our banks, borrowing to maximize the leverage of their assets at unheard-of levels, produced mountains of financial paper instruments (called asset-backed securities) with little means of measuring their value I like to use a salad analogy. Instead of making simple loans but suddenly global investors were no longer ordering salad. regardless of how much money government spends the major governments of the world, including the Chinese, should begin major fiscal efforts to stimulate their weakening economies. Table 3-26: All material clauses in the text There is a large number (nine) of material clauses without a Goal or Range, and most of these (six of them: clause 15, 40, 41, 43, 53, 54) are used to construct the idea of something rising/falling. This trend of representing rising/falling is discussed in detail later in section 3.3.4, in Figure 3-14, p. 77. The other three clauses (clause 26, 55, 79) do not stand alone but are used very closely in conjunction with another clause immediately preceding or following it. There is also a large number (ten) of material clauses with Goal. This is significant because it shows the construction of a reality in which things are impacted upon, e.g. clause 29-32: loans, clause 49, 78: the financial system, clause 51: banks. This reflects the fact that the financial crisis is a phenomenon which involved real events with impact. But considering the great impact the crisis had on all levels of society, the question arises as to why there aren’t even more material clauses with Goals. One reason is that many material processes with Goals are rankshifted or realized as grammatical metaphors, see Table 3-27 next page (these examples were also in Table 3-25, p. 66). 68 Clause Ranking Clause (underlined: non-ranking material process; bold: No. Goal of underlined process) transitive type 22 33 48 60 64 73 Relational {which was} followed by a de-leveraging [of the entire Identifying global financial system]. Relational The idea was [[to sell individual servings Identifying [of diversified financial salad] around the world]]. Relational Policymakers have no means [[of forcing the banks to start Attributive lending]] Relational The great uncertainty is [[whether government has the power Identifying [[to rescue the financial system]] in times of crisis]]. Relational This means [[government cannot bail out the system || even Identifying if it wanted || to {bail out the system}]]. Relational The challenge will be [[to reform our financial system Identifying quickly]] Table 3-27: Clauses with Goals in non-ranking material processes Yet, only one of the six clauses in Table 3-27 (clause 22) construes an event which actually occurred; the remaining five only construe potential/possible events. The same can be said for the ten clauses with Goals listed in Table 3-26 (p. 68): only five of them construe events which actually occurred (clause 9, 30, 31, 32, 51). Therefore the question of how the text achieves the construal of the actual impact of the financial crisis remains. The answer is that impact is construed with a wide variety of transitivity configurations not involving a Goal; key examples of such clauses are listed in Table 3-28 next page. The clauses in Table 3-28 contain impacted elements not in the form of a Goal, but can be reworded (see rightmost column) with the impacted element as Goal in a material process. The financial crisis was not a physical event like an earthquake, war, or epidemic, but it had great material, mental and relational impact. 69 Clause Transitive type Clause (underlined: process, Clause No. bold: impacted participant) reworded with bold element as Goal 18 Material (with Range) 15 Material (no Goal/Range) 41 Material (no Goal/Range) 53 Material Global investors ... have investors stopped declared a buyer’s strike against buying the paper the sophisticated paper assets assets of of securitization ... securitization Instead, stock markets panic brought down collapsed the stock markets when global interest rates the crisis drove up began to rise. global interest rates The cost of money is rising banks are increasing (no Goal/Range) 16 21 36 the cost of money Relational and credit markets remained distrust has frozen Attributive frozen. the credit markets Relational Therefore, the housing crisis the housing crisis Attributive was a mere trigger for a collapse destroyed investor’s of trust in paper, trust and you’re dead. toxic salad killed you Relational Attributive 50 Relational After all, the U.S banks alone so the crisis has taken Attributive far during the crisis have lost away $2 trillion from upwards of $2 trillion from the banks’ asset base their collective asset base. 23 Mental As a result, we are experiencing this brought down the Affect the painful downward value of every asset reappraisal of the value of in the world virtually every asset in the world. Table 3-28: Construal of the impact of the crisis in clauses without Goal 70 3.3.3 Minority Type Clauses After considering the majority of transitive process types, the relational and material, we now turn to the minority types. Interestingly, many of the nine minority transitive type clauses (six mental, one behavioural, one verbal, one existential) involve metaphors which realize meanings of the majority transitive types in some form or other. There are grammatical metaphors in the mental clauses that construe material processes (see Table 3-29 below, rightmost column), and also some other types of ideational metaphors in the remaining clauses (see Table 3-30 next page, rightmost column), one of which construes relational meaning in congruent form. Clause Mental Clause No. grammatical metaphors underlined) Process (process in bold, relevant Grammatical Type metaphors with shifted material processes 23 Affect As a result, we are experiencing the painful “reappraisal” downward reappraisal of the value of (Proc Ent) virtually every asset in the world. 27 Cognition But with the huge expansion of the global “expansion” economy in the 1990s, which produced an (Proc Ent), ocean of new capital, the bankers came up “securitization” with an idea called securitization. (Proc Ent) No one knew the location of the toxic waste. 39 Cognition 59 Perception Apart from the economic pain resulting from “shrinking” shrinking credit markets, we are about to see (Proc Qual), 68 Cognition an earthquake in the relationship between “earthquake” government and financial markets. (Proc Ent) Our policy leaders in Washington are thinking domestically 77 Cognition The bankers don't trust each other. Table 3-29: Minority transitive process types I – mental (6 clauses) 71 Clause Process Clause (process in bold) Type of metaphor No. Type 10 Behavioural looking through the rearview mirror. Ideational lexical metaphor over a (clause 9: Those [=policymakers] clause complex with some knowledge are driving (clause 9-10) policy) 44 Verbal The markets are telling us Ideational metaphor of transitivity over a 63 Existential (clause 45: the world will face a clause complex serious credit crunch in 2009) (clause 44-47) A similar situation exists in many Euro zone countries. Table 3-30: Minority transitive process types II – behavioural/verbal/existential (3 clauses) The processes in clause 10 & 44 are not simple ideational metaphors (such as the examples discussed in Halliday 1994: 344) which operate over a single clause, but are more complex metaphors which function over a clause complex, in that a rewording to a non-metaphorical process would involve neighbouring clauses. This is illustrated in Table 3-31 below. Clause Original clauses, with metaphorical process Alternate version of clause No. in bold 44 & 45, with nonmetaphorical process in bold 43 and why have financial markets collapsed so These conditions of the brutally? financial markets imply 44 The markets are telling us [[that the world will face a 45 the world will face a serious credit crunch in serious credit crunch in 2009 2009]]. Table 3-31: Metaphorical verbal clause 44 interpreted as relational meaning 72 3.3.4 Some Individual Clauses Now we examine more closely particularly interesting individual examples of relational clauses in the text. Clause No. Process Type Clause (bold: ranking process, underlined: grammatical metaphors shifting process to entity) 21 Relational Therefore, the housing crisis was a mere trigger [for a collapse [of trust [in paper]]], Paraphrase 1 Material The housing crisis triggered a collapse of trust in paper. Paraphrase 2 Material The housing crisis destroyed trust in paper instruments. Paraphrase 3 Mental Because of the housing crisis, investors stopped trusting paper instruments. Table 3-32: A relational clause which ultimately construes material and mental processes Throughout the text there is extensive use of relational processes for the abstract construal of physical events and conscious experience. For instance, in clause 21 in Table 3-32, although the explicit process is relational, the relational meaning generated by the sentence is not the ‘final product’, the final experiential meaning conveyed by the sentence as a whole, but an intermediate ‘stepping stone’ used to represent material processes, which are in turn used construe a mental process. This is achieved by use of grammatical metaphors ( “trigger”, “collapse”, “trust”) and preposition group expansion (“for”, “of”, “in”). Table 3-32 also shows some experiential paraphrases of clause 21. Each paraphrase conveys more or less the same overall experiential meaning as clause 21, but from paraphrase 1 to paraphrase 3, the grammatical metaphors are gradually unpacked. The end-result, paraphrase 3, shows that there is no substantial relational or material meaning in clause 21. There is 73 nothing physical that is triggered or that is collapsing. The author seems to enjoy portraying the crisis (which is an event with many material processes) in terms of the mental states it evokes, by packaging the material processes into grammatical metaphors that become participants of a mental process. Here is another example: Clause Clause (process in bold, grammatical metaphors underlined) No. 23 As a result, we are experiencing the painful downward reappraisal [of the value [of virtually every asset [in the world]]]. Table 3-33: A mental clause which portrays material events in grammatical metaphors Grammatical Shift Congruent expression metaphor in clause 23 painful7 Circumstance Quality We are painfully experiencing... downward Circumstance Modifier Its value is going downwards reappraisal Process value, asset Entity Entity Modifier (within prepositional phrase) The assets are reappraised We are reappraising its value We are reappraising our assets Table 3-34: Congruent usage of grammatical metaphors in clause 23 7 It is difficult to argue for a congruent grammatical function for pain, as noted by Halliday (1998a) (Ent: we felt pain, Proc: we were pained) but I’ve chosen Circ here because the original wordings can be preserved, and pain is more a feature of “experiencing” than a feature of “reappraisal”. 74 Level of Thing & Modifier embedding 0 1 Thing Modifier reappraisal [of the value [of virtually every asset [in the world]]]. Thing Modifier value [of virtually every asset [in the world]]. 2 Thing Modifier asset [in the world] Thing 3 world Table 3-35: Word group expansion in clause 23 It seems the author is being playful with the grammatical structure of clause 23, the prepositional group expansion is literally “going downwards” in terms of the level of embedding (Table 3-35), as the clause progresses. This is a fine example of the interplay between the logical metafunction and the experiential metafunction. The experiential construction of the dynamic process of downward reappraisal is assisted by the intra-clausal logical dynamics. In addition, it also seems the inter-clausal logical structure is going downwards, as we move towards clause 23, as well (Figure 3-13). 75 Figure 3-13: Logical structure leading to clause 23 – a “downward” progression Clause 23 appears at the bottommost position of the first major logicocomplex in the text, embedded 4 levels. The semantic pattern continues even as we move up the scale of this semiotic system from the level of logico complex to the discourse level. The idea/experience of going “upwards” and “downwards” is a dominant trend throughout the text, as the following lexical string shows: 76 No. n Dir 13 ↘ ↘ 15 19 ↗ In recent years, our banks, borrowing to maximize the 21 ↘ 23 ↘ 27 ↗ ↗ 32 ↗ 40 ↗ 41 ↗ 43 Clause which, if gulped down , would quickly restore financial stability. Instead, stock markets collapsed leverage of their assets at unheard-of levels , produced mountains of financial paper instruments (called assetbacked securities) with little means of measuring their value Therefore, the housing crisis was a mere trigger for a collapse of trust in paper, As a result, we are experiencing the painful downward reappraisal of the value of virtually every asset in the world. But with the huge expansion of the global economy in the 1990s, which produced an ocean of new capital, the bankers came up with an idea called securitization. and {the bank} sliced them up into a big, beautiful tossed salad. This distrust heightened when global interest rates began to rise . ↘ and why have financial markets collapsed so brutally? ↘ the world will face a serious credit crunch in 2009 After all, the U.S. banks alone so far during the crisis have lost upwards of $2 trillion from their collective asset base. The cost of money is rising 54 ↘ and the availability {is} shrinking . 59 ↘ 67 ↘ Apart from the economic pain resulting from shrinking credit markets, we are about to see an earthquake in the relationship between government and financial markets. The collapse of, say, a major European bank would hardly leave American workers immune. it is that the world's money is sitting on the sidelines 45 50 ↗ 53 ↗ 71 72 -- more than $6 trillion in idle global money markets alone. Figure 3-14: Lexical string of ups and downs – all words associated with rising/falling in the text The lexical string in Figure 3-14 demonstrates how the author plays around with the representation of financial entities/indicators going upwards, and downwards. To begin with, the very theme of the text is the “credit crisis”, alternately referred to as “credit crunch”, hence it is about things in the global financial system e.g. “stock markets”, “credit markets”, “economy”, which are on the decline. 77 The “Dirn” column in Figure 3-14 visualizes the direction of movement. There are almost equal numbers of words associated with ups (eight) and downs (nine). It might seem a bit strange that there is almost a 50-50 balance between the two directions of movement, given that it is an article about the credit crunch. Perhaps the author wants to provoke humour in his representation of the state of financial affairs with these seemingly random ups and downs, making the journey seem erratic and meaningless. Or perhaps the writer wants his financial news article to mimic the behavior of a financial market itself – having ups and downs. A closer look at the words denoting rising movements show that many are really about negative things going upwards, e.g. clause 40: “distrust”, 41: “global interest rates”, 50: “amount of money lost”, 53: “cost of money”. There are two words associated with a lack of movement - “sitting”, clause 71 and “idle”, clause 72. An interesting observation is that the word “up” in clause 27 is not used to mean anything going upwards, but is part of a compound verb “came up with” that has altogether a different meaning from the individual words. However, the choice of having the word “up” there could be deliberate. Similarly, in clause 32, “up” seems to be deliberately added to the process of “sliced”, when its omission would not change the meaning of the sentence alone. The “up” words in clauses 27, 32 follow from the “expansion of the global economy in the 1990s” and seem to reinforce the portrayal of the optimism surrounding the making of the “salad” – the toxic financial instruments that eventually led to the crisis. In other words, the “up” words in clause 27, 32 serve to construct a parody at the discourse level. With these observations, we find that six out of the eight “↗” words have a 78 negative function. Almost all the “↘” words are referring to negative situations, e.g. stock market collapse, collapse of trust, collapse of bank. There is one notable and very interesting exception: the first clause in the list, clause 13, where the magic pill (bailout package) is gulped down – this is supposed to represent a positive scenario. However, the author’s usage of the expression “gulped down” is again likely a parody, because soon after, the author writes “stock markets collapsed”. It is possible that the experiential effects of rising and falling throughout the text are not deliberately created by the author, but are unconsciously accomplished effects resulting from the sum total of his lexical choices, across the clauses in the text. Whether consciously or unconsciously constructed, the experiential dynamics of Figure 3-14 are inevitably an important feature of the text. 79 3.4 ERGATIVITY Ergativity is a system of semantic description whereby the ideation of each clause, brought about by the Process, centers upon a Medium, the key figure in the actualization of the Process - Halliday (1994: 163). The Medium is the only compulsory element in the clause apart from the Process. The other ergative elements in the clause are organized around the Medium and Process, for example, a clause may have (1) an Agent, which is a participant that acts on the Medium, and is interpreted as the cause of the Process, (2) a Range, which construes the scope or domain over which the Process occurs, or (3) Circumstances, which express optional information about the Process. Ergativity analysis is required for the determination of Voice (defined in Halliday 1994: 168-169) for a clause. Voice analysis is important for the analysis of Agency. A clause can either be in middle Voice (The glass broke) or effective Voice (Tom broke the glass), which is the case when an Agent is present (Agent = Tom), possibly implicitly (The glass was broken). In this thesis, non-finite clauses were not analysed for Voice, and coded as Voice-none, because nonfinite clauses have indeterminate participants. 3.4.1 Medium To get a sense of the center stage of ideation in the discourse, we examine all the instances of Medium in the ranking clauses. 80 No. Medium (inferred) Clause (with Medium highlighted in yellow) 7 8 9 10 11 12 13 14 clause 8 policymakers policy 15 16 17 markets markets clause 15, 16 18 Global investors 19 our banks 20 paper instruments 21 the housing crisis 22 23 collapse of trust in paper we 24 paper instruments 25 26 27 I bankers bankers 28 29 30 31 32 33 loans loans loans loans The idea 34 The only problem 35 36 37 38 39 40 41 salad you salad global investors No one distrust global interest rates this salad boycott financial markets markets the world government bank balance sheets Policymakers financial system banks At this point in the credit crisis, at least one thing is certain : most policymakers lack a clue of what is really at stake. Those with some knowledge are driving policy looking through the rearview mirror. Begin with the U.S. Treasury's $700 billion bailout package. This was presented as some magic pill which, if gulped down, would quickly restore financial stability. The "shock and awe" of the sheer size of the taxpayer-funded bailout would somehow restore confidence. Instead, stock markets collapsed and credit markets remained frozen. This is because the credit crisis reflects something more fundamental than a serious problem of mortgage defaults. Global investors, now on the sidelines, have declared a buyers' strike against the sophisticated paper assets of securitization that financial institutions use to measure and offload risk. In recent years, our banks, borrowing to maximize the leverage of their assets at unheard-of levels, produced mountains of financial paper instruments (called asset-backed securities) with little means of measuring their value Incredibly, these paper instruments were insured by more dubious paper instruments. Therefore, the housing crisis was a mere trigger for a collapse of trust in paper, {which was} followed by a de-leveraging of the entire global financial system. As a result, we are experiencing the painful downward reappraisal of the value of virtually every asset in the world. So what are these paper instruments, these asset-backed or mortgagebacked securities? I like to use a salad analogy. Before the last decade, bankers simply lent in the form of syndicated loans. But with the huge expansion of the global economy in the 1990s, which produced an ocean of new capital, the bankers came up with an idea called securitization. Instead of making simple loans and holding them until maturity, a bank collected all its loans together, then {the bank} diced {its loans} and {the bank} sliced them up into a big, beautiful tossed salad. The idea was to sell (for huge fees) individual servings of diversified financial salad around the world. The only problem: under an occasional piece of lettuce was a speck of toxic waste in the form of a defaulting subprime mortgage. Eat that piece of salad, and you 're dead. The overall salad looked delicious, but suddenly global investors were no longer ordering salad. No one knew the location of the toxic waste. This distrust heightened when global interest rates began to rise. 42 43 44 45 46 47 48 49 50 bailout package bailout package bailout package So what does this salad boycott mean for the future and why have financial markets collapsed so brutally? The markets are telling us the world will face a serious credit crunch in 2009 regardless of how much money government spends to remove the toxic salad from bank balance sheets. Policymakers have no means of forcing the banks to start lending short of nationalizing the entire financial system. After all, the U.S. banks alone so far during the crisis have lost upwards of $2 trillion from their collective asset base. 81 No. Medium (inferred) Clause (with Medium highlighted in yellow) 51 52 banks Translation of clause 50-51 53 54 58 The cost of money the availability of money banks fear people such as Warren Buffet uncertainty Most banks are leveraged by more than 10 to 1. Translation: The U.S. financial system will have a whopping $15 trillion to $20 trillion less credit available next year than was around a year and a half before. The cost of money is rising and the availability {is} shrinking. 59 we 60 uncertainty 61 66 embedded clause in clause 60 the collected assets of the major banks clause 62 clause 61, 62 government guarantees we 67 American workers 68 69 73 74 75 policy leaders solution to the crisis the world the world's money more than $6 trillion The challenge global capital The first step 76 77 78 We bankers financial system 79 80 clause 78 governments 55 56 57 62 63 64 65 70 71 72 True, the banks will still lend ‐‐ but the fear is they will do it only to people such as Warren Buffett, who don't need loans. What is uncertain is the amount of lending to borrowers engaged in entrepreneurial risk, the center of business reinvention and job creation. Apart from the economic pain resulting from shrinking credit markets, we are about to see an earthquake in the relationship between government and financial markets. The great uncertainty is whether government has the power to rescue the financial system in times of crisis. It seems doubtful. In the United Kingdom, for example, the collected assets of the major banks are four times the nation's gross domestic product (GDP). A similar situation exists in many Euro zone countries. This means government cannot bail out the system even if it wanted to. Given such massive exposure, government guarantees in a time of crisis become meaningless. Yet because of the interconnected web of global financial relationships, we are all vulnerable to the threat. The collapse of, say, a major European bank would hardly leave American workers immune. Our policy leaders in Washington are thinking domestically when the solution to the credit crisis will be global. It is not that the world lacks money; it is that the world's money is sitting on the sidelines ‐‐ more than $6 trillion in idle global money markets alone. The challenge will be to reform our financial system quickly to draw that global capital back into more productive uses. The first step should be efforts to make the market for future asset-backed paper more transparent and credible. We need a private/public global bank clearing facility. The bankers don't trust each other. The central banks, working with the private institutions in providing enhanced data, need to begin to refashion the world's financial architecture. And while that is happening, the major governments of the world, including the Chinese, should begin major fiscal efforts to stimulate their weakening economies. Figure 3-15: All Mediums in the text, and their clausal context Participants which belong to the same semantic categories are grouped together and their frequency of occurrence in Mediums is counted in Table 3-36: 82 Cat. Frequency No. Clauses with Medium Participant, with examples in containing participant brackets 8 (modifier/rankshifted) 1 9 19, 26, 27, (47), 50, 51, “banks”, “bankers” 55, (62), 77 2 3 8 7 (19), 23, 25, 36, 59, 66, Personal pronouns (“we”, “you”, “I”, (68), 76 “our”) 29-32, 41, 47, 62 Other financial entities (“loans”, “interest rates”, “assets”) 4 7 7, 17, (52), 61, 63, 64, 79 Anaphoric elements referring to preceding text/clauses (“one thing”, “this”, “that”, “it”) 5 6 15, 16, 43, 44, 49, 78 “financial system”, “markets” 6 6 8, 46, 48, (65), 68, 80 “government”, “policymaker” 7 6 20, (22), 24, 35, 37, (42) “paper instruments”, “salad” 8 5 18, 38, 39, 57, 67 People (“investors”, “workers”) 9 5 33, 34, 52, 73, 75 Abstract grammatical metaphors (“the idea”, “the challenge”) 10 5 (22), 40, 56, 58, 60 Feelings (“trust”, “distrust”, “fear”, “uncertainty”) 11 5 (53), (54), 71, 72, 74 Money 12 5 (18), (41), 45, 70, (71) “the world”, “global” Table 3-36: Categories of frequency occurring Mediums in the text The most frequently occurring participant is the bank/banker. They play the important roles in every stage of events in the financial crisis – they produced a mountain of paper instruments (clause 19), essentially “toxic salad”, and served it 8 A clause in brackets denotes the participant is not the entity in the Medium, but a modifier of the entity. e.g. in clause 65, the Medium is “government guarantees”, this Medium belongs to Category 4 but the government is only a modifier of “guarantee”. 83 around the world (clause 26-41). The banks have lost trillions of dollars (clause 50), and their potential collapse would threaten the world (clause 67), and they are also responsible for restoring the financial system (clause 76-78). Apart from banks, the second most frequently occurring category is the personal pronouns. They function to create a tighter bond between the reader and the author, and make the crisis seem something we can all experience, not just an abstract phenomenon affecting financial institutions, e.g. (clause 35-36) “Eat that piece of salad, and you’re dead”. This interpersonal effect is also achieved with participants from category 8: people and category 10: feelings. The appearance of “we” as Medium in three strategically located clauses within the discourse: clause 23 and 66 at the ends of the 1st and 3rd major logico complexes, and clause 59 which begins a new sub-complex, dramatically portrays the people as being in the center of experience, in the financial disaster that unfolds. Clause Clause (bold: Medium) No. 23 As a result, we are experiencing the painful downward reappraisal [of the value [of virtually every asset [in the world]]]. 59 Apart from the economic pain [[resulting from shrinking credit markets]], we are about to see an earthquake [in the relationship [between government and financial markets]]. 66 Yet because of the interconnected web [of global financial relationships], we are all vulnerable to the threat. Table 3-37: Medium “we” in clauses at/near boundaries of logico complexes The next most frequently occurring category in Table 3-36 is category 3: other 84 financial entities, which really represents phenomena orchestrated by human agents. Category 4 consists of anaphoric elements which take part in processes that really are more reflective of the textual & logical metafunction than the ideational metafunction. Next in frequency is category 5: financial system & markets – these (like category 2) occur in clauses which function to show things “happening to” them, rather than them “doing” anything. E.g. clause 15, 16, 43 have markets collapsing/staying frozen. The same goes for category 7: paper instruments/salad. Of particular interest to us, then, is category 6: governments/policymakers. Though they do not occur as frequently as banks or financial entities, the author chooses to begin the article with policymakers (clause 7-10) and end with governments (clause 80). It is possible the term “policymakers” is used to refer to the government (it would be too disrespectful for Smick to say in clause 8: “Most governments lack a clue of what’s really at stake”). In any case, the government takes a far more important role than what the frequency rankings in Table 3-36 suggest. Moving beyond just the Medium, the following reference chains track the participants of categories 1, 2, 5, 6 in all ergative roles throughout the discourse and show that the chains representing banks, personal pronouns, and financial systems contain items distributed evenly throughout the discourse, but the governments and policymakers are clustered near the boundaries of the text. 85 Figure 3-16: Reference chains 1, 2, 5, 6 corresponding to all participants in the discourse belonging to categories 1, 2, 5, 6 of Table 3-36 Note: The system network for reference chain analysis is in Appendix 7.3, p. 419. 86 Figure 3-17: Items in Reference chains 1, 2, 5, 6 highlighted in the text 87 3.4.2 Agency Agency is a significant feature because it assigns responsibility. Agentive clauses in the text are rare – only 17 out of 74 clauses are in effective Voice: Effective Voice Middle Voice Voice - none Total 17 50 7 74 Table 3-38: Voice in the text (ranking clauses) Agent Ellipsed Agent Agent Ellipsed Agent Active Passive Total 4 3 9 1 17 Table 3-39: Types of Effective Voice (ranking) clauses in the Text Taking into account that four out of the 17 effective Voice clauses have ellipsed Agents, this leaves only 13 clauses in the text with an explicit Agent. Ten of these clauses are relational. Agents in relational clauses are not useful for examining responsibility because agency in relational clauses is an abstract notion and may not correlate with responsibility as much as Agents in material clauses. So turning to material clauses, we find there are only three material clauses in the text with a nonellipsed Agent: Clause Clause (Agent underlined, Process in bold) No. 9 Those with some knowledge are driving policy (Those = Policymakers) 30 a bank collected all its loans together, 88 78 The central banks, , need to begin to refashion the world's financial architecture. Table 3-40: Material clauses in the text with non-ellipsed Agent From this we gather that the two main participants with some responsibility for plunging us into or rescuing us from the crisis are policymakers and banks. The agentivity of clause 30 in particular is important for it marks the beginning of the description of making the “salad”, the root cause of the entire financial crisis. Throughout the text the author takes great care (either intentionally or unintentionally) in what he overtly assigns agentive status. Firstly, there are relatively few effective clauses in the entire text. Most of the Agents in relational effective clauses are grammatical metaphors, so the agentivity is not used to congruently describe action, but to construct relations - an example is clause 33, shown in Table 3-41: Clause Clause (Agent underlined, {} denote ellipsis) No. 30 a bank collected all its loans together, 31 then {the bank} diced {its loans} 32 and {the bank} sliced them up into a big, beautiful tossed salad. 33 The idea was to sell (for huge fees) individual servings of diversified financial salad around the world. Table 3-41: Agency in the salad analogy The Agent “bank” in clause 30 is ellipsed in the subsequent clauses 31 and 32. Then, in clause 33, it disappears from the grammar of the clause entirely, but ideationally it really means the same as clause 33’: 89 Clause Clause (underlined: Agent, italics: replaced material cf clause 33) No. 33’ The banks were going to sell (for huge fees) individual servings of diversified financial salad around the world. Table 3-42: Hidden Agent in the salad analogy However, the author has avoided directly stating the agency of the banks as the salad analogy proceeds to describe more severe consequences. Evidently, the author takes professional care not to play the ‘blame game’ too obviously. The same, however, cannot be said for the comments posted by internet readers. The author also goes to great lengths to hide the agentivity of governments/policymakers by rankshifting the actions of the government and expressing them as Range: Clause Clause (bold: rankshifted clause, underlined: Range) No. 48 Policymakers have no means [[of forcing the banks to start lending]] ( = policymakers can’t force the banks...) 60 The great uncertainty is [[whether government has the power [[to rescue the financial system]] in times of crisis.]] ( = the government cannot rescue the financial system...) 80 the major governments [of the world], , should begin major fiscal efforts [[to stimulate their weakening economies]]. ( = the government should stimulate their weakening economies...) Table 3-43: Rankshifted processes with government/policymaker as potential Agent The government has also been ellipsed as Agent: Clause Clause Remarks No. 12 This was presented as some magic pill by the government? Table 3-44: Ellipsed Agent: government 90 The government only appears once in the text as an explicit Agent, and this is rankshifted: Clause Clause No. 64 This means [[government cannot bail out the system || even if it wanted || to {bail out the system}]]. Table 3-45: Rankshifted Agent: government Despite these attempts, it is evident that the governments are positioned at the top of the hierarchy of agency. 91 3.5 MOOD Mood analysis is performed according to chapter 4 of Halliday (1994). Clauses in the text are primarily declarative (Table 3-46 below), knowledge-statement (Table 3-47 below), non-modalised (Table 3-52, p. 97), present tense (Table 3-56, p. 101), and positive polarity (Table 3-59, p. 104). 3.5.1 Mood and Speech Function Declarative Interrogative Imperative Non-finite Total 62 3 2 7 74 Table 3-46: Mood of clauses in the text Knowledge/Statement Action/Command Action/Offer Total 70 3 1 74 Table 3-47: Speech Function in the text In Table 3-46 and Table 3-47, only ranking clauses are analysed. The seven non-finite ranking clauses in the text are considered as inheriting the speech functions and moods from the finite clause on which they are dependent. The primary purpose of Smick’s story is to provide information about the financial crisis, hence almost all the clauses are in declarative mood, and almost all speech function is knowledge/statement. The exceptions to declarative mood are three interrogatives and two imperatives, but all of these five exceptions are actually mood 92 metaphors which serve to facilitate the presentation of information. Clause Mood Clause No. 24 Interrogative So what are these paper instruments, these asset-backed or mortgage-backed securities? 42 Interrogative So what does this salad boycott mean for the future 43 Interrogative and why have financial markets collapsed so brutally? 11 Imperative Begin with the U.S. Treasury's $700 billion bailout package. 35 Imperative Eat that piece of salad, Table 3-48: Clauses with non-declarative mood in the text The interrogatives in Table 3-48 are not instances of the author asking for information from the reader, as the author poses himself throughout as an authority figure on the subject of the financial crisis. Neither do the imperatives in Table 3-48 expect any form of action from the reader. Clauses 24, 42, 43 and 11 can also be interpreted as a form of mental command to the reader, asking the reader to think about the issues: the paper instruments, the salad boycott, the reason for the collapse of markets, and the bailout package. But undeniably, they are textual devices that are used to organise the text; they signal changes in the development of the text. They are in fact functioning as high level themes at the discourse level: macro themes for portions of the text. From the clause complexing analysis in Figure 3-18, Figure 3-19 and Figure 3-20, we see that clauses 11, 24, and 42 & 43 mark the start of new complexes in the logical structure of the discourse. This is representative of the style of the author. He begins each major logical complex in the text with a mood metaphor, a non-declarative clause functioning as knowledge/statement. Seen from another perspective, the author himself is responding to his own questions in clause 24, 42, 43 and the command clause 11. 93 In particular, it seems the text itself is responding to the command clause 11 “Begin with...”. Elements such as “Let me begin by...” are analysed by Martin (1992) p. 416 as grammatical metaphors functioning in the textual metafunction, because they do not represent a congruent experiential process of “beginning”, but function to negotiate texture in a discourse. Figure 3-18: First major logico complex in the text (underlined in red: nondeclarative mood) Figure 3-19: Second major logico complex in the text (underlined in red: nondeclarative mood) 94 Figure 3-20: Third major logico complex in the text (underlined in red: nondeclarative mood) The only non-declarative clause in the text that does not begin a new complex is clause 35, which forms part of the following nexus: Clause Clause No. 35 Eat that piece of salad, 36 and you're dead. Table 3-49: Nexus of clause 35 and 36 The nexus in Table 3-49 shows that clause 35 is not a congruent imperative functioning as a command, but a mood metaphor that functions as knowledge statement as part of the overall message, “if you eat that piece of salad, you will die.” The purpose of realising the function knowledge/statement in the form of a command is interpersonal: it makes Smick appear as the person in control, and reduces the 95 distance between Smick’s story and the reader to a minimum, because it takes the reader into the transitivity of the text itself. The tense in clause 36 “and you’re dead” is also metaphorical (congruent: you’ll be dead), and the purpose of this is also to reduce the distance of the ideation of the event to the reader. All of this distancereduction serves as a shock tactic to the reader, and in fact, makes this the most interpersonally dramatic moment in the article. Moving back to Table 3-47, almost all clauses in the text are statements of knowledge; three of the four exceptions are commands which the author gives towards the end of the text: Clause Clause (process in bold) No. 75 The first step should be efforts to make the market for future asset-backed paper more transparent and credible. 78 The central banks, working with the private institutions in providing enhanced data, need to begin to refashion the world's financial architecture. 80 the major governments of the world, including the Chinese, should begin major fiscal efforts to stimulate their weakening economies. Table 3-50: Clauses with Speech Function = Command And the last exception being the author’s offer to the reader to use a salad analogy to describe the creation of paper instruments of securitization: Clause Clause No. 25 I like to use a salad analogy. Table 3-51: Clause with Speech Function = Offer Again, these exceptions to the rule are really closely related to the norm. The 96 clauses with speech function as command can really be considered knowledge statements, encoding Subjects (governments, banks) with obligation, and the offer clause 25 can be read as a statement of the author’s preference, rather than an offer to the reader. 3.5.2 Modality and Mood Adjuncts Now we look at the clauses in the text with modality. Subjective/Implicit Subjective/Explicit Objective/Explicit Total 6 1 2 9 Table 3-52: Modality Orientation in the text There are only nine clauses with modality, out of the 67 finite ranking clauses in the text. Not reflected in Table 3-52 is the fact that there are three rankshifted clauses with modality. Modality is a rarity in this text, hiding the fact that the views in the text are the author’s own (though it is stated explicitly in the website, at the end of the article, that “The opinions expressed in this commentary are solely those of the writer”), and to make him appear authoritative and sure of his facts. Some examples of clauses with modality follow: Clause Modality No. Type 13 Probability Clause (modal Finite in bold) which, , would quickly restore financial stability. 14 Probability The "shock and awe" [of the sheer size [of the taxpayer funded bailout]] would somehow restore confidence. 97 67 Probability The collapse of, say, a major European bank would hardly leave American workers immune. 75 Obligation The first step should be efforts [[to make the market [for future asset -backed paper] more transparent and credible]] 80 Obligation the major governments [of the world], , should begin major fiscal efforts [[to stimulate their weakening economies]]. Table 3-53: Clauses with Modal Finites (examples) In clause 13 and 14, the modalised Finite “would” conveys not the author’s judgement, but the judgement of those who the author detracts from. This is a definite trend in the text, where the author extensively employs modality to represent his imagined opponent’s viewpoint. Another example of this comes from clause 55, one of the clauses in the text with the most Mood Adjuncts (two Mood Adjuncts): Figure 3-21: Mood Adjunct analysis of clause 55 The positive polarity conveyed by the Mood Adjunct of polarity “true” does not represent an assertion of the author’s view, because clause 56 says “but the fear is they will do it only to people such as Warren Buffet”. When “true” is used with the Mood Adjunct of time “still”, it makes the statement in clause 55 appear as if its truth is being insisted upon, but the author never uses such excessive interpersonal force to make his own points anywhere else in the text. The effect of these Mood Adjuncts is 98 only to make the statement, which really is a representation of an opponent’s argument, weaker than his own arguments. Clause 55 is a significant point from the perspective of the interpersonal metafunction, and this is elaborated in Chapter 4: Recurrence Plots, under section 4.3.1: Mood, p. 144. There is another clause in the text with two Mood Adjuncts: Figure 3-22: Mood Adjunct analysis of clause 50 In the case of clause 50, the Mood Adjuncts do represent the author’s view, but they do not serve to convince the reader of the truth of an argument; they only convey the intensity of the proposition according to the reference frame of the author. As we shall see, the two Mood Adjuncts of clause 50 also function in the transitivity of the clause, because clause 50 joins the discourse in portraying the scale of the crisis (Chapter 4: Recurrence Plots, under section 4.3.3: Transitivity, p. 155-157). 99 3.5.3 Modality in Transitivity and Grammatical Metaphor Clause Grammatical Clause (process in bold, grammatical metaphor No. metaphor shift underlined) 7 Modality At this point [in the credit crisis], at least one thing is Quality certain (clause 8: : most policymakers lack a clue of what is really at stake.) 58 Modality [[What is uncertain]] is the amount of lending [to Quality borrowers [[engaged in entrepreneurial risk, the center of business reinvention and job creation.]]] 60 Modality The great uncertainty is [[whether government has the Entity power [[to rescue the financial system]] in times of crisis.]] 61 Modality It seems doubtful. Quality Table 3-54: Relational clauses that construe modality Although modality appears only in a small minority of the clauses in the text, there are a few clauses (shown in Table 3-54) which can be considered modalised, though in ways that are not immediately recognizable. This is achieved with the interpersonal grammatical metaphors “uncertainty” , “uncertain”, “certain”, and “doubt”, which are connected to the proposition by relational processes, and also employing the use of reference words “it” (clause 61) and “one thing” (clause 7). The modality in clause 7 is really meant to be brought forward to clause 8, and the modality in clause 61 is brought backwards to clause 60. In fact it can even be argued that the statements of clause 7, 60, 61 are nonstandard forms of objective explicit modality, given that their interpersonal meanings 100 match closely the standard objective explicit forms given by clause 7’, 60’, 61’ in Table 3-55. Clause Modality Clause (process in bold, grammatical metaphor No. Orientation underlined) 7’ Objective/Explicit At this point in the credit crisis, it is certain [[that most policymakers lack a clue of what is really at stake.]] 60’ Objective/Explicit It is uncertain [[whether government has the power to rescue the financial system in times of crisis.]] 61’ Objective/Explicit It seems doubtful [[whether government has the power to rescue the financial system in times of crisis.]] Table 3-55: Some modalised versions of statements in Table 3-54 Notice, however, the versions in Table 3-55 appear far more arguable and less certain than their counterparts in the original text. This is achieved by distancing the modality from the proposition, either textually, by reference (clause 7: “one thing”, clause 61: “it”) or ideationally, by abstraction (clause 60: “uncertainty” – shift from modality to quality “uncertain”, then shift from quality “uncertain” to entity “uncertainty”). 3.5.4 Tense and Polarity Present Past Future Total (finite clauses) 36 20 8 67 Table 3-56: Tense in the text The most representative examples of each tense type are given in Table 3-57 next page: 101 Clause Tense Clause (tense element in bold) No. 19 past In recent years, our banks, , produced mountains [of financial paper instruments ([[called asset backed securities]])] with little means [[of measuring their value]] 23 present As a result, we are experiencing the painful downward reappraisal [of the value [of virtually every asset [in the world]]]. 45 future the world will face a serious credit crunch in 2009 Table 3-57:: Clauses with present, past, future tenses (examples) Most of the 67 finite ranking clauses in the text are in present tense, reflecting the fact that the crisis was a very current event – the article was written near the height of the financial crisis, in Oct 2008. A piano roll diagram can be generated by Systemics to visualize changes in systemic choices from clause to clause. This is particularly useful in the analysis of Tense. There are three possible choices in Tense: past, present and future. In the piano roll, a shaded box appears when a choice is present. A piano roll for Tense is generated for clause 7 to 80: 80 Figure 3-23: Piano roll for Tense The same piano roll enlarged and split: 102 Figure 3-24: Piano roll for Tense (enlarged) The first four and last six of the finite clauses of the text are in the present tense, so the author begins the story in the present, shuttles back and forth between the past and future, but finally ends in the present time again. The second most common tense is past tense. The clauses in the text in past tense tend to occur as a continuous conti group: e.g. clause 18-21 21 (history (h of “paper instruments”), 26-33 33 (making of “salad”), “salad 37-41 41 (consequences of “salad”). Beyond clause 50, past tense no longer appears. This is expected because the author moves from explaining past events leading to the th crisis, to highlighting the consequences. Future tense is the rarest, and almost al all of it (six out of eight)) occurs in the second half of the text (after clause 44). This analysis reveals that the author’s orchestration of Tense is well-organized well on a discourse level. The irregularities irregularities in the unfolding of tense, therefore, signal important changes in the ideational metafunction, for example, future tense only occurs twice in the first 37 clauses of the text, shown in Table 3-58:: Clause Tense Clause (tense element in bold) No. which, if gulped down, would quickly restore financial stability. 14 future The "shock and awe" of the sheer size of the taxpayertaxpayer funded bailout would somehow restore confidence. Table 3-58:: Clauses with future tense at the start of the text 13 future 103 Clause 13 and 14 are involved in the ideation of a fictional future event, which the author brings up only temporarily – to be dismissed immediately with clause 15: “Instead, stock markets collapsed”. The effect of this irregularity is similar to that in the earlier discussion of clauses with an unusual presence of modality and mood adjuncts (p. 98). Evidently, anomalies in mood choices (modality, mood adjunct or tense) point to moments in the discourse where the viewpoint opposite to the author’s is represented awkwardly, for ease of rebuttal. The analysis of Polarity is summarised in Table 3-59: Positive Negative Total (finite clauses) 64 3 67 Table 3-59: Polarity in the text These polarity statistics are defined by the polarity of the Finite element in the ranking clauses. The text is overwhelmingly positive in polarity (the three exceptions have already been discussed in Table 3-22 on p. 61). The author is presenting his own argument, and is not arguing against anyone. However, the overall tone of the text is still pessimistic, but the negativity is not encoded through the Finite element. Clause Polarity Clause (Finite element in bold) No. 21 positive Therefore, the housing crisis was a mere trigger [for a collapse [of trust [in paper]]], 77 negative The bankers don’t trust each other Table 3-60: Clauses with positive/negative polarity (examples) 104 3.5.5 Rankshifting Rankshifted clauses are found in 17 of the 74 ranking clauses. This is a significant number. Of the 17 rankshifted clauses, three are doubly embedded (having an embedded clause within an embedded clause): Clause Process type Clause (process in bold) No. 17 Relational This is because [[the credit crisis reflects something [[ {which is} more fundamental than a serious problem [of mortgage defaults] ]]]]. 52 Relational Translation : [[The U.S. financial system will have a whopping $15 trillion to $20 trillion less credit available next year [[than {what}was around a year and a half before]].]] 60 Relational The great uncertainty is [[whether government has the power [[to rescue the financial system]] in times of crisis.]] Table 3-61: Clauses with double embedding in the text The significance of rankshifting clauses is that rankshifted clauses often have alternate modes of realization as ranking clauses. The choice by the writer to use rankshifted clauses results in a different interpersonal presentation of the message. For example, one could use the following clause (containing no rankshifted clauses) to replace clause 17: Clause Process type Clause (process in bold) No. 17’ Relational The credit crisis is more fundamental than a serious problem of mortgage defaults. Table 3-62: Alternate version of clause 17, with no embedding 105 The important difference between clause 17 and 17’ is interpersonal – the linguistic distancing achieved by doubly embedding the proposition in clause 17 presents the information more subtly and gradually, and hence makes the information less arguable and more likely acceptable to a reader than clause 17’. 106 3.6 GRAMMATICAL METAPHORS The analysis of grammatical metaphors is performed according to Halliday (1998b) and Martin (1992). Grammatical metaphors are identified where there is a shift in the grammatical function of word(s) in a clause from a hypothetical “congruent clause”, which is the grammatically simplest possible re-expression of the original words preserving the original meaning, e.g. (Halliday 1998b: 191) “brake failure” is a grammatical metaphor because it can be re-expressed, with shifts in the grammatical functions of the words, as “the brakes failed”. We define the following grammatical functions: Process, Entity, Modifier, Circumstance, Quality/Attribute9, Relator, Modality and None; examples of these are given in Table 3-63: Example clauses (grammatical metaphors in Shift in grammatical bold, congruent rewordings italicized) function Example 1 John’s sudden call surprised me. Mod Qual Ent John’s (congruent version) John suddenly called me and surprised me. Ent Circ Proc Example 2 I am certain the reason the bridge collapsed Proc Qual Ent sudden was the fact that it was sabotaged. Ent (congruent version) The bridge must have collapsed because it Modal Rel Ent Mod Circ Qual call Proc Ent am Modal Proc certain Modal Qual reason Rel Ent fact None Ent was sabotaged. Table 3-63: Grammatical metaphor examples demonstrating all the grammatical functions 9 Attribute is very closely related to Quality. The difference is that an Attribute is a more basic aspect/property of an entity and is more quantifiable (non-qualitative). E.g. importance is Qual Ent; price is Attrib Ent. Examples of shifts in the Smick text involving Attrib are value, height, size, cost. 107 Grammatical metaphors are also classified according to the metafunction in which they primarily function: experiential, logical, interpersonal or textual. In Table 3-63 (previous page), all the grammatical metaphors (rightmost column) are experiential except for “am” & “certain”, which are interpersonal, and “reason”, which is logical. In the CNN main story, grammatical metaphors are very frequently used – 143 of them; considering there are 74 ranking clauses, that’s an average of about two grammatical metaphors per ranking clause. Grammatical metaphors are found almost continuously from clause to clause, moving from the beginning to the end of the text, and are equally pervasive in the rankshifted clauses (the figure of 143 includes 37 grammatical metaphors in 25 rankshifted clauses). Experiential Logical Interpersonal Textual Total 111 8 14 10 143 Table 3-64: Types of grammatical metaphors in the text Most of the grammatical metaphors are experiential, as we expect, because the article needs to direct substantial resources to construct an adequate representation of the complexity of the financial crisis. As the article represents the field of finance, many financial terms that would be normally analysed as grammatical metaphors are considered as congruent for this article. For instance, in clauses where “exposure”, “loan”, “mortgage” are used as entities, their usage is treated as congruent, though they could possibly be analyzed as 108 grammatical shifts from Process to Entity. In fact these financial processes, e.g. of loaning, mortgaging are often formalized on paper; this means the grammatical metaphors have a physical entity to refer to and become congruent. It is even possible that this human linguistic ability to construct grammatical metaphors, i.e. conceive processes as things, is the root of the financial crisis, because the “toxic salad” could not be created without being able to “slice” and “dice” the loans and sell them. As a final (or foremost) note, the word “crisis” itself in the text (it appears a total of eight times) can be analysed as a grammatical metaphor too, because a crisis does not exist as an entity by itself. It always depends on other entities, as it invokes questions like “Crisis of what? Who is affected?” Clause Sample clauses showing metaphorical and congruent usage of “crisis” No. 21 (“crisis” used as a grammatical metaphor: Circ Ent) Therefore, the housing crisis was a mere trigger [for a collapse [of trust [in paper]]], 50 (“crisis” used congruently, to construe circumstantial meaning) After all, the U.S. banks alone so far have lost upwards of $2 trillion from their collective asset base. Table 3-65: Metaphorical and congruent use of the word “crisis” However, the title of the article “Why there’s a crisis – and how to stop it” introduces “crisis” into the text as an Entity, so I will treat all appearances of “crisis” as congruent, in respect of its status in the title. 109 3.6.1 Experiential Shifts The following table outlines the main types of experiential grammatical metaphors which appear in the story: Shift from to Process Entity Entity Modifier Process Quality Circumstance (other) Quality (other) Others Total Count 36 25 21 8 7 14 111 Table 3-66: Statistics of Experiential grammatical metaphors in the text The majority of the most frequent shift types are shifts to the more concrete, e.g. Process to Entity, and Process to Quality, signifying that in the discourse, the shift is predominantly from relations to entities. In the next few pages, each of the types of shifts listed in Table 3-66 are discussed, in order of frequency. Now let’s examine the instances of the most frequent type of shift: 110 Process Cl. Entity (total: 36) 14 shock 22 de-leveraging 58 lending 69 solution awe 23 reappraisal risk 73 challenge bailout 27 expansion reinvention 74 uses 17 defaults securitization creation 75 efforts 18 strike 33 servings 59 earthquake step securitization 40 distrust relationship 80 efforts risk 42 boycott 65 guarantees 21 trigger 45 crunch 66 relationships collapse 52 Translation threat trust 56 fear 67 collapse Table 3-67: Experiential grammatical metaphors shifting Process to Entity Many of the metaphorical entities have unpleasant connotations, namely, strike, collapse, de-leveraging, distrust, boycott, crunch, fear, threat, collapse. These shifts of negative material processes into entities have allowed the processes in ranking clauses to appear more positive in the form of relations, e.g. see clause 21, 56 in Table 3-68 below: Clause Sample clauses (relevant grammatical metaphors in bold) No. 14 The "shock and awe" [of the sheer size [of the taxpayer-funded bailout]] would somehow restore confidence. 18 Global investors, , have declared a buyers' strike [against the sophisticated paper assets [of securitization] [[that financial institutions use {the sophisticated paper assets [of securitization10]} || to measure || and offload risk]]]. 21 Therefore, the housing crisis was a mere trigger [for a collapse [of trust [in paper]]], 56 -- but the fear is [[they will do it only to people ,]] Table 3-68: Examples of clauses with experiential grammatical metaphors shifting Process to Entity 10 Ellipsed metaphors are not double counted. 111 The next most frequent type of experiential shift is from Entity to Modifier, displayed in Table 3-69 below. Compared to the categories of Medium discussed in the section 3.4 Ergativity p. 80, the participants acting as Modifiers are mostly inanimate, the most frequent being asset, appearing as a Modifier four times. The only human Modifiers are few, and all of them are people not representing the banks or government: taxpayer, buyer, borrowers. The linguistic backgrounding of them as modifiers within nominal groups reflects Smick’s backgrounding of the relative importance of these groups of people. Entity Cl. 11 14 17 18 19 21 22 23 U.S. Treasury's $700 billion taxpayer mortgage buyers' asset their (= our banks) paper global financial system every asset in the world Modifier (total: 25) 24 asset mortgage 27 global economy 39 toxic waste 42 salad 45 credit 53 money 58 business job borrowers 65 67 69 75 76 government major European bank credit crisis asset bank Table 3-69: Experiential grammatical metaphors shifting Entity to Modifier Clause Sample clauses (relevant grammatical metaphors in bold) No. 11 Begin with the U.S. Treasury's $700 billion bailout package. 14 The "shock and awe" [of the sheer size [of the taxpayer -funded bailout]] would somehow restore confidence. 24 So what are these paper instruments, these asset -backed or mortgage backed securities? Table 3-70: Examples of clauses with experiential grammatical metaphors shifting Entity to Modifier The third most frequent experiential metaphorical shift is from Process to 112 Quality. Most of the processes involved describe circumstantial meanings e.g. – backed, tossed, enhanced, and restricted to the noun under modification, but some e.g. frozen, interconnected, shrinking, weakening are qualities applicable to all participants involved in the crisis at large. Cl. Process Quality (total: 21) 11 14 16 19 bailout 34 defaulting -funded 50 collective frozen 59 shrinking unheard-of 62 collected -backed 66 interconnected 21 housing 74 productive 24 -backed 75 -backed -backed 76 clearing 26 syndicated 78 enhanced 32 tossed 80 weakening 33 diversified Table 3-71: Experiential grammatical metaphors shifting Process to Quality Clause Sample clauses (relevant grammatical metaphors in bold) No. 16 and credit markets remained frozen. 59 Apart from the economic pain [[resulting from shrinking credit markets]], we are about to see an earthquake [in the relationship [between government and financial markets]]. 80 the major governments [of the world], , should begin major fiscal efforts [[to stimulate their weakening economies]]. Table 3-72: Examples of clauses with experiential grammatical metaphors shifting Process to Quality The rest of the other experiential shifts occur less than ten times each. Next on the list: circumstantial shifts. Most of them construe spatial/temporal meaning. 113 Cl. Circumstance * (total: 8) Quality (6) Entity (2) 23 painful 7 point downward 39 location 33 individual 34 occasional 75 first future Table 3-73: Experiential grammatical metaphors with shifts from Circumstance The congruent versions of painful, individual, occasional are taken to be painfully, individually, occasionally (e.g. individual servings Clause served individually) Sample clause (relevant grammatical metaphors in bold) No. 23 As a result, we are experiencing the painful downward reappraisal [of the value [of virtually every asset [in the world]]]. Table 3-74: Clause showing shifts from Circumstance Next are the shifts from Quality. They denote mostly positive qualities: ‘stable’, ‘confident’, ‘mature’, ‘available’ and ‘maximum’. Cl. Quality * (total: 7) Entity(4) Circumstance (2) 13 stability 43 brutally 14 confidence 68 domestically 29 maturity Process (1) 54 availability 19 maximize Table 3-75: Experiential grammatical metaphors with shifts from Quality Clause Sample clause (relevant grammatical metaphors in bold) No. 43 and why have financial markets collapsed so brutally? Table 3-76: Clause showing shift from Quality 114 The remaining types of experiential grammatical metaphors are listed in Table 3-77. Experiential Grammatical Metaphors – all other shifts (total: 14) Cl. 14 19 23 53 Attribute size value value cost Entity 40 Attribute Process heightened 11 None Entity package 79 None Process happening 49 Entity Process nationalizing 58 59 65 Entity Quality entrepreneurial economic meaningless 60 65 Relator Circumstance in times of in a time of 67 Relator leave Process Table 3-77: Metaphors representing the least frequent experiential shifts Clause Sample clause (relevant grammatical metaphors in bold) No. 65 Given such massive exposure, government guarantees [in a time of crisis] become meaningless. Table 3-78: Clause showing infrequent type experiential shifts A more congruent version of clause 65 could be “Given such massive exposure, government guarantees during a crisis mean nothing”. This shows the author dresses up his sentences to make them grammatically fuller, which is a form of linguistic distancing occurring at the word/word group level. 115 3.6.2 Logical, Interpersonal and Textual Shifts The logical metaphors in the text help in constructing temporal sequences or causal chains of events of the crisis. Logical shifts (8) 19 22 33 34 48 59 63 66 (from Relator to ... ) Entity In recent years, our banks, , produced mountains [of financial paper instruments ([[called asset -backed securities]])] with little means [[of measuring their value]]. Process {which was} followed by a de-leveraging [of the entire global financial system]. Entity The idea was [[to sell individual servings [of diversified financial salad] around the world]]. Entity The only problem : [[[under an occasional piece of lettuce] was a speck of toxic waste [in the form of a defaulting subprime mortgage]]]. Entity Policymakers have no means [[of forcing the banks to start lending]] Process Apart from the economic pain [[resulting from shrinking credit markets]], we are about to see an earthquake [in the relationship [between government and financial markets]]. Quality A similar situation exists in many Euro zone countries. Circumstance Yet because of the interconnected web [of global financial relationships], we are all vulnerable to the threat. Table 3-79: Clauses with logical grammatical metaphors (underlined) Many of the above words are identified as logical grammatical metaphors by reference to word lists in earlier published work. The words “similar” and “means of” are logical grammatical metaphors due to their clause-relating property, as part of the Vocabulary 3 items in Winter (1977: 20). The analysis of “problem”, “result”, “because of” follows Martin (1992: 409). “idea” is a grammatical metaphor because it is an “anaphoric noun” of cognition, defined by Francis (1986: 15), but the experiential function of “idea” in clause 33 is not as crucial as its logical function; it mainly serves to link clause 33 as a ‘purpose clause’ to the preceding clauses 116 describing the making of “salad”. Some grammatical metaphors in the main story are interpersonal in function (see Table 3-80 below), representing the points in the discourse where the author steps in to assert his views and position himself to the reader. Interpersonal shifts (14) 7 Modal 20 Modal Modal 58 Modal Modal Modal 60 Modal Modal 61 Modal 70 None Modal 71 None Modal 75 Modal Qual Circ Qual Proc Qual Proc Ent Proc Qual Ent Proc Ent Proc Qual At this point [in the credit crisis], at least one thing is certain Incredibly, these paper instruments were insured by more dubious paper instruments. [[What is uncertain]] is the amount of lending [to borrowers [[engaged in entrepreneurial risk, the center of business reinvention and job creation.]]] The great uncertainty is [[whether government has the power [[to rescue the financial system]] in times of crisis.]] It seems doubtful. It is not that the world lacks money ; it is that the world's money is sitting on the sidelines The first step should be efforts [[to make the market [for future asset -backed paper] more transparent and credible]]. Table 3-80: Clauses with interpersonal grammatical metaphors (underlined) The main categories for these interpersonal grammatical metaphors are certainty (7: certain, 58: uncertain, 60: uncertainty), doubt (20: dubious, 61: doubtful) and credibility (20: incredibly, 75: credible). The majority (five of seven) of these words are negative in modality. This reflects the lack of confidence and trust of the author in the government and the financial institutions, which was representative of the mood of the people during those times, that brought the markets crashing and created the financial crisis. 117 In the main story, there are also textual metaphors that help to negotiate texture in the discourse, bringing parts of the text itself into the transitivity of clauses. Textual shifts (10) 7 At this point [in the credit crisis], at least one thing is certain 11 Begin with the U.S. Treasury's $700 billion bailout package. 17 This is because [[the credit crisis reflects something [[ {which is} more fundamental than a serious problem [of mortgage defaults] ]]]]. 23 As a result, we are experiencing the painful downward reappraisal [of the value [of virtually every asset [in the world]]]. 61 It seems doubtful. 62 In the United Kingdom, , the collected assets [of the major banks] are four times the nation's gross domestic product (GDP). 63 A similar situation exists in many Euro zone countries. 64 This means [[government cannot bail out the system || even if it wanted || to {bail out the system}]]. 65 Given such massive exposure, government guarantees [in a time of crisis] become meaningless. 66 And while that is happening, Table 3-81: Clauses with textual grammatical metaphors (underlined) Many textual grammatical metaphors function strongly in the logical metafunction, e.g. “situation”11, “example”, “result”, see Martin (1992: 409, 416417). An outstanding feature revealed by the textual grammatical metaphor analysis in Table 3-81 above is the high level of cohesion from clause 62 to 66 created by the textual metaphors. This cohesiveness was not captured in the earlier analysis of thematic progression in section 3.2: Theme in Table 3-4, p. 39, because the referenced items of the metaphors (e.g. 63: “situation”, 64: “this”, 66: “that”) include both thematic and rhematic elements, or have no clear boundaries. 11 The analysis of “situation” as textual grammatical metaphor here departs from Martin (1992: 409, 416-417) who classifies it as a logical grammatical metaphor, but he also notes the significant overlap between logical and textual grammatical metaphors. Another justification is that the logical aspect of “similar situation” in clause 63 is already encoded in “similar”, see Table 3-79 p. 114. 118 3.6.3 High Density Clauses There are a small number of clauses with a large number of grammatical metaphors each. There are five clauses with six or more metaphors, and the clause with the most has 11. Clause No. of No. Clauses with many grammatical metaphors ( ≥6 ), metaphors grammatical in bold metaphors 14 7 The "shock and awe" [of the sheer size [of the taxpayerfunded bailout]] would somehow restore confidence. 19 7 In recent years, our banks, , produced mountains [of financial paper instruments ([[called assetbacked securities]])] with little means [[of measuring their value]]. 23 6 As a result, we are experiencing the painful downward reappraisal [of the value [of virtually every asset [in the world]]]. 58 11 [[What is uncertain]] is the amount of lending [to borrowers [[engaged in entrepreneurial risk, the center of business reinvention and job creation.]]] 75 7 The first step should be efforts [[to make the market [for future asset -backed paper] more transparent and credible]]. Table 3-82: Clauses with the most grammatical metaphors The clause with the most grammatical metaphors is clause 58: 119 Figure 3-25: Grammatical metaphor analysis of clause 58 Clause No. Clause (bold: grammatical metaphors, underlined: congruent rewording) 58 What is uncertain is the amount of lending to borrowers engaged in entrepreneurial risk, the center of business reinvention and job creation. 58’ ( possible congruent version of clause 58: ) Banks might not lend enough to entrepreneurs who risk their money to start businesses and create jobs. Table 3-83: Unpacking grammatical metaphors in clause 58 Clause 58 is a thematic equative clause, the only such clause in the text, giving 120 it textual prominence. The fact that it is also the clause with the most grammatical metaphors adds further to this textual prominence. The processes of reinventing businesses and creating jobs become the entities “business reinvention” and “job creation”, which make them appear more textually tangible than if they were expressed as congruently as processes – this makes them appear substantial and important, which is probably the intention of the author. A congruent version of clause 58 would require agentive entities such as “banks” and “entrepreneurs”, but doing this would eclipse the prominence of the entities “business reinvention” and “job creation”. By realizing the Modality as a Quality “uncertain” and placing it in the same clause with the entities “business reinvention” and “job creation”, the author creates the following effect: an association is created between the entities, i.e. that business reinvention is uncertain and job creation is uncertain, even though the only direct relation that clause 58 states is that the lending is uncertain. Another effect created by the writer loading clause 58 with such high density of grammatical metaphors is that this creates a kind of textual opposition of high/low density, between clause 58, and clause 55, which represents an opponent’s argument (see the logical context in Figure 3-26). This textual opposition of grammatical metaphors seems to reinforce the author’s interpersonal opposition. The relative complexity of clause 58 establishes him in a position of greater knowledge. Figure 3-26: Opposition between clause 55 and clause 58 121 3.6.4 Distribution of Grammatical Metaphors A comparison of the patterns of grammatical metaphor density with the logical structure of the discourse is made in Figure 3-28 on page 123. The table shows the counts of each type of grammatical metaphor per clause, for all clauses in the text, and clauses with a higher density are shaded darker according to the colour code: 1 to 2 3 to 5 6 to 7 11 Figure 3-27: Colour code for Figure 3-28 The categories of grammatical metaphor listed in the table in Figure 3-28 consist of the six categories of experiential metaphors discussed previously in Table 3-66 (p. 110) collectively under the heading EXP, followed by the three other metafunctions: LOG (Logical), INT (Interpersonal), and TEX (Textual). The figure to the right of the table represents all the logico complexes in the text, with the same vertical scale. The three major complexes are highlighted in red in the table. 122 CLAUSE 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 SUM Total No. 3 0 0 0 5 0 1 7 0 1 3 4 7 2 5 3 6 4 0 1 3 0 1 0 0 1 4 3 0 0 0 0 2 2 0 2 1 0 2 0 0 1 1 1 0 1 2 1 0 1 0 11 5 3 2 2 2 1 5 4 3 1 2 2 2 0 1 2 7 2 0 1 2 2 143 EXP Proc Ent Ent Mod Proc Qual Circ →* 1 2 1 3 1 1 1 3 1 1 2 Qual →* Other LOG INT TEX 1 1 1 1 1 1 1 1 1 1 1 2 1 2 3 1 1 1 1 1 2 1 1 2 1 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 4 2 3 1 1 1 1 3 1 2 1 1 1 1 2 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 2 1 1 2 1 1 1 1 1 2 1 1 1 1 36 25 1 19 8 7 16 1 8 14 10 Figure 3-28: Grammatical metaphor density against logical structure 123 The shading reveals a disproportionate concentration of high density clauses in the first major logico complex of the text (even though it is the smallest, only comprising 13 clauses). Hence grammatical metaphors seem to be crucial in the article’s initial “building up” of the representation of events of the financial crisis. In fact the shading reveals that clauses of moderate density (three or more grammatical metaphors) occur in clusters close to the start and end of the discourse. In particular, the interpersonal and textual grammatical metaphors are completely absent in the middle of the discourse (clause 24 to 57). This makes the mid-part of the text more oriented towards ideational content. But in contrast, the experiential metaphors generally spread evenly in the beginning, the middle and the end. An exception to this is the experiential metaphors Entity to Modifier, there are 25 of these and about half of them (ten of them) are concentrated in the first major complex, see Table 3-84 below. The reason for this seems to be that the author is trying to pack as many participants into the introduction as possible. Clause No. Experiential grammatical metaphor (Ent Mod) 11 U.S. Treasury’s, $700 billion 14 taxpayer 17 mortgage 18 buyer’s 19 asset, their (= paper instruments) 21 paper 22 entire global financial system 23 virtually every asset in the world Table 3-84: Experiential grammatical metaphors shifting Entity to Modifier, at the start of the discourse 124 3.6.5 ‘Void’ of Grammatical Metaphors The widespread presence of grammatical metaphors throughout the text suggests that grammatical metaphors are an expected feature of the text, a reflection of the nature of the writer’s style as well as the register of formal financial news reporting. What is significant, then is not so much the presence of grammatical metaphors, but their absence in certain places, and their most glaring absence is a ‘void’ comprising clauses 35-38 (four clauses), highlighted in red in Table 3-85 next page. Its logical context is also presented in Figure 3-29 below: Figure 3-29: Grammatical metaphor ‘void’ in second major logico-complex 125 CLAUSE 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 EXP Proc Ent Total No. 3 0 0 0 5 0 1 7 0 1 3 4 7 2 5 3 6 4 0 1 3 0 1 0 0 1 4 3 0 0 0 0 2 2 0 2 1 0 2 0 0 1 1 1 0 1 2 1 0 1 0 11 5 3 2 2 2 1 5 4 3 1 2 2 2 0 1 2 7 2 0 1 2 2 Ent Mod Proc Qual Circ →* 1 2 1 3 1 1 1 3 1 1 2 Qual →* Other LOG INT TEX 1 1 1 1 1 1 1 1 1 1 1 2 1 2 3 1 1 1 1 1 2 1 1 2 1 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 4 2 3 1 1 1 1 3 1 2 1 1 1 1 2 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 2 1 1 2 1 1 1 1 1 2 1 1 1 1 1 1 Table 3-85: Grammatical metaphor ‘void’ in red 126 The ‘void of grammatical metaphors’ occurs at the end of a major logico complex in the text and describe the conclusions of the author at the end of a long chain of arguments. In contrast, in the earlier stages within the same logico complex (as shown in Figure 3-29, p. 125), the author uses many grammatical metaphors: Clause No. Clauses leading up to the ‘void’, with grammatical metaphors in bold 24 So what are these paper instruments, these asset-backed or mortgagebacked securities? But with the huge expansion of the global economy in the 1990s, 27 , the bankers came up with an idea [[called securitization]]. The only problem: under an occasional piece of lettuce was a speck of 34 toxic waste in the form of a defaulting subprime mortgage. Table 3-86: Grammatical metaphors in the second major logico complex leading to the ‘void’ The contrastive absence of grammatical metaphors towards the end of the logico complex makes the argument more convincing, making it appear as a straightforward matter-of-fact. The sobering reality of the toxic salad is brought forward to the reader more dramatically. However, this congruency is an illusion because all the abstraction and complexity is hidden by the lexical metaphor of the “salad”, which appears in three of the four clauses of the ‘void’: Clause No. Clauses of the grammatical metaphor ‘void’ 35 Eat that piece of salad, 36 and you’re dead 37 The overall salad looked delicious, 38 But suddenly investors were no longer ordering salad. Table 3-87: Occurrences of “salad” in the grammatical metaphor ‘void’ 127 3.6.6 Parallel Patterns with Rankshifted and Interrupting clauses In the previous section we’ve examined the distribution of grammatical metaphors in relation to the logical structure of the discourse. It appears as if the author uses grammatical metaphors to build up some kind of “tension” in the text, due to the packing up of processes into entities, together with other shifts (e.g. Proc Qual, etc). Now, grammatical metaphors are resources that the author can choose to use in a text. With that view, rankshifting and interrupting clauses also represent another kind of resource that can package extra processes into a ranking clause. As the juxtaposition of the distribution of grammatical metaphors, rankshifted clauses and interrupting clauses in Table 3-88 12 shows, the author orchestrates these three groups of resources almost in parallel. Two regions of discourse where these resources are most intensely exercised are clause 17-19, and 58-60. It is striking that, at a discourse level, the overall textual patterns of gradual build-up and release of meanings are very similar. 12 For the count statistics in Table 3-88: -a single interrupting complex with x clauses is counted as x interrupting clauses -a single rankshifted complex with x clauses is counted as x rankshifted clauses 128 NUMBER OF GRAMMATICAL METAPHORS CLAUSE 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 SUM 3 0 0 0 5 0 1 7 0 1 3 4 7 2 5 3 6 4 0 1 3 0 1 0 0 1 4 3 0 0 0 0 2 2 0 2 1 0 2 0 0 1 1 1 0 1 2 1 0 1 0 11 5 3 2 2 2 1 5 4 3 1 2 2 2 0 1 2 7 2 0 1 2 2 143 NUMBER OF RANKSHIFTED CLAUSES 0 1 0 0 0 0 0 0 0 0 1 3 2 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 2 0 0 0 1 0 2 1 2 0 0 0 3 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 1 25 LEVEL 1 LEVEL 2 NUMBER OF INTERRUPTING CLAUSES 1 1 1 3 2 1 2 1 1 1 1 1 1 1 1 2 1 1 1 3 1 1 2 1 23 2 1 8 Table 3-88: Grammatical metaphors versus rankshifted and interrupting clauses 129 _____________________________________________________________________ Chapter 4: RECURRENCE PLOTS _____________________________________________________________________ 4.1 CLAUSES AS VECTORS In the previous sections we have covered the results of traditional SF analysis on text, facilitated by the software Systemics. In this chapter we take an exploratory step towards mathematicizing the linguistic analysis. From a mathematical standpoint, even traditional SF analysis involves a mapping of natural language into some simple kind of mathematical structure: system networks, dependency relations and constituent structures. This section explores in depth the utility of a different mathematical representation: a vectorical perspective, which simplifies the SF analysis of a clause as a vector. A vector is simply a set of numbers, which can be written in a row or column form: ሺ2 1 0ሻ 2 ൭1൱ 0 Figure 4-1: A vector, in row and column form. To demonstrate how to represent a clause as a vector, consider the first clause of the main story, and its analysis in terms of ergativity: Figure 4-2: Mapping a clause into tags (clause 7) 130 For clause 7 we have Circumstance, Medium, Process and Range tags, but this does not represent the full range of possible ergativity tags. In Systemics, the exact coding of all such possible tags are: Cl/E1/Agent Cl/E1/Benef Cl/E1/Circ Cl/E1/Medm Cl/E1/Proc Cl/E1/Range Table 4-1: All possible ergativity tags in a clause There can be more than one tag of a kind in a clause, e.g. 2 or more “Cl/E1/Circ” tags. By simply counting the number of tags of each kind, we arrive at the following six dimensional vector to represent the ergativity meaning in clause 7: Cl7 = (0 0 1 1 1 1) Figure 4-3: A vector representing the ergativity analysis of clause 7 From here we will sometimes use the term “tags” to refer to types of tags, which means dimensions of meaning. The ordering of the numbers in the vector in Figure 4-3 follows the alphabetical order of the tag names in Table 4-1 for convenience, but this choice of ordering is arbitrary; it does not denote the relative importance of the tags. While much information of the original analysis (Figure 4-2) is lost in such a vast simplification, the utility of this vectorical formulation lies in the ease with which we can perform comparisons of a clause with other clauses, thereby revealing the comparative dynamics of meaning throughout the discourse. Applied linguists have long concerned themselves with studying processes in language use. Larsen-Freeman & Cameron (2008) have written at length on the 131 conceptualisation of discourse as a complex dynamical system. This thesis takes an exploratory step forward in defining the dynamics of discourse, by using precise mathematical parameterization of meanings. The formulation of a clause as a vector effectively characterises the state of the discourse, treating the clause as a unit of measurement, a moment of constancy. In Systemic Functional Theory, a clause is the outcome of simultaneous choices across multiple semantic domains and can therefore be represented as a vector/point in a multi-dimensional semantic space, where each dimension corresponds to a potential choice, which is formalised as a node in a system network. The multidimensional nature of the vector represents the complexity of the space of possible meanings. In reality there are hundreds of dimensions. The vector space spanned by the complete set of tag vectors can be said to represent this “semantic space”. The usage of vector spaces to represent meaning in text is not new, see Landauer (2007). In the past two decades, extensive research has been conducted on Latent Semantic Analysis (LSA), which is a theory of meaning based on a vectorical representation and analysis of a text document, relying on word/phrase frequencies in part or whole of the document. Computer automated techniques based on LSA has enjoyed tremendous success in mimicking up to 90% of human abilities in answering language MCQ tests and grading essays (Landauer 2007: 20-21, 26-27). LSA also enjoys widespread, multi-lingual application in internet search engines’ ranking of website relevancy relative to search queries, see Berry & Browne (2005). However, the limitations of LSA are that pragmatic context, word order, and the dynamic process of creating text are ignored, all of which are addressed in Systemic Functional Grammar. 132 Systemic Functional linguists Halliday (1993), Matthiessen (1992), and Butt & O'Toole (2003) have used the term “semantic space” in very general ways, but here, for the first time, we present a precise mathematical definition of a semantic space, based on the SF theory of meaning. In fact, even in this present study, the term “semantic space” has been used differently, for instance, in Chapter 3 (p. 35), it was used to describe the potential of discourse progression based on lexical associations around the words “credit crisis”. For disambiguation, we shall call the vector space derived from tags Functional Semantic Space, or more simply, “Semantic Space” (capitalised). The higher the degree of delicacy of the SF analysis, the higher the number of dimensions the Semantic Space becomes. For our analysis of the CNN story, the number of dimensions across all metafunctions has surpassed one hundred - see section 4.5: Combined Metafunctions (p. 172). However, a higher dimensional space is impossible to visualise directly. One indirect way to visualise higher dimensional data would be to project them into two or three dimensions, but this will result in great loss of information and distortion. Recurrence Plots were introduced by Eckmann et al. (1987), to facilitate the visualisation and analysis of dynamical systems. They are named after their ability to show the tendency of many deterministic dynamical systems to recur, i.e. periodically return to a previous state. Recurrence plots are useful because they afford a means of visualising high dimensional systems on a two dimensional plot, without losing information. In the sections that follow we will define the operations and calculations we use to derive the plots, before presenting the plots. 133 The reason for our use of recurrence plots in discourse analysis is not so much to show recurrence in the discourse, because discourse is not exactly periodic (except for theme) or even deterministic, but it is certainly not random, hence the plots can be useful for extracting patterns that have significant functional meaning. 4.2 TYPES OF RECURRENCE PLOTS 4.2.1 XOR count: Difference in Meaning One way of comparing clauses is to measure the difference between their vector representations. Consider comparing the ergative analysis of clause 7 (covered in the previous section, in Figure 4-2, p. 130) with that of clause 8: Figure 4-4: Ergativity analysis of clause 8 Cl8 = (0 0 0 1 1 1) Figure 4-5: A vector representing the ergativity analysis of clause 8 Comparing the vectors for clause 7, 8, we find only one difference, in red: Figure 4-6: Difference in ergativity meaning between clause 7 and 8 This corresponds to the presence of a “Circ” (circumstance) in clause 7 and the 134 absence of any “Circ” in clause 8. Systemics version 1.2.7 (released 28 Nov 2009) is built-in with tools to automatically compute “XOR count”: the number of cases in which a tag is present in one clause, but absent in the other clause. The name “XOR” denotes the logical algebraic operation that is performed between the two vectors to generate the count. XOR gives a positive output only when a tag is eXclusively present in either vector1 OR vector2. Input Output Vector 1 1 1 0 0 Table 4-2: XOR Definition Vector 2 1 0 1 0 XOR count 0 1 1 0 For clause 7 and 8, XOR count = 1. 4.2.2 AND count: Similarity in Meaning XOR count measures the difference between clauses. However, sometimes we are interested in measuring similarity instead, which we can do by defining an AND operation that returns positive when a tag is present in vector 1 AND present in vector 2 as well: Input Vector 1 1 1 0 0 Table 4-3: AND Definition Output Vector 2 1 0 1 0 XOR count 1 0 0 0 135 Using this AND comparison on clause 7 and 8, Figure 4-7: Similarity in ergativity meaning between clause 7 and 8 For clause 7 and 8, AND count = 3. 4.2.3 OR count: Sum of Meanings And finally, we also define an OR operation that returns positive when a tag is present in vector 1 OR vector 2 (possibly both). In effect, this sums up the presences of tags in both vectors, or measures the sum of meanings in both clauses. Input Vector 1 1 1 0 0 Table 4-4: OR Definition Output Vector 2 1 0 1 0 XOR count 1 1 1 0 Note that by definition, OR count = XOR count + AND count. Figure 4-8: Sum of ergativity meaning between clause 7 and 8 For clause 7 and 8, OR count = 4. 136 4.2.4 Reduced Vectors In the previous examples of XOR, AND, OR calculations, we’ve only encountered vectors with entries up to a value of “1”. However it is perfectly possible to encounter a vector with values larger than “1”, for example, when a clause has two circumstances of transitivity. For the purposes of XOR, AND, OR calculations, only reduced vectors are used – if a raw vector has entries larger than “1”, the larger values are all reduced to “1”, and the result is a reduced vector. This is because the comparison is only based on the presence/absence of tags, not on the difference in their numbers. E.g. If vector 1 and vector 2 are similar in all respects except that vector 1 has one “Circ” and vector 2 has two “Circ”, their XOR count is zero. 4.2.5 Tag Decomposition The vector representations we have shown for clause 7 and 8 in Figure 4-3 and Figure 4-5 have only taken into account the end-node selections of the system network. However, a selection of the end-node often means selecting all the superordinate nodes as well. For example, in the transitivity analysis of clause 7, Figure 4-9: Transitivity analysis of clause 7 the selection “Cl/T1/Proc/Reln/Attr/Int” is really made up of six selections (or from another perspective, the tag name is made up of six names): 137 Cl Cl/T1 Cl/T1/Proc Cl/T1/Proc/Reln Cl/T1/Proc/Reln/Attr Cl/T1/Proc/Reln/Attr/Int Table 4-5: Decomposition of a tag representing intensive relational attributive process Performing this hierarchical breakdown of tags produces a larger set of dimensions of meaning with which compare clauses. This describes more fully the choices made in the instantiation of a clause, and the meaning within. This can be useful, for example, comparing only the end nodes of Cl/T1/Proc/Reln/Attr/Int Cl/T1/Proc/Reln/Attr/Poss gives an AND count = 0. But comparing all nodes, AND count = 5. A nonzero AND count does more justice to the similarity between two closely related selections that only differ at the end nodes. But some superordinate tags may be less useful or meaningful, e.g. “Cl”, “Cl/T1”, and “Cl/T1/Proc” are present in all clauses, so they are of no consequence in making clause comparisons. Nevertheless, we will apply the tag decomposition universally in all our analyses in this study, unless stated otherwise. 138 4.3 PLOTS OF INDIVIDUAL METAFUNCTIONS Most of our analysis will be based on XOR comparisons, due to their ease of interpretation. XOR count can be calculated for each clause compared against all the clauses, producing a square matrix of numerical values. The matrix can be visualised as a two dimensional array of colour-coded pixels, which is called a recurrence plot. So far in this chapter, we have only demonstrated vector representation and analysis for ergativity. In Systemics, the user can select any combination of tags he wishes to analyse recurrence for, e.g. whether to compare the clauses in terms of a limited number of tag types, or all tags in an entire metafunction, or tags across metafunctions. In this section, we will examine some interesting patterns in selected recurrence plots of all the 74 clauses of the CNN main story. These recurrence plots represent a synoptic perspective of the differences in meaning through the entire discourse. Not only are the patterns and features of these plots the result of dynamics of similarities and differences in meaning between successive clauses as the discourse unfolds, but also the relations between faraway clauses. 4.3.1 Mood We examine the differences in choices of Mood between the clauses, with a recurrence plot of XOR counts in all 35 Mood dimensions as listed in Table 4-6 next page. The numbers to the left of each tag type show the number of its occurrences in the text. 139 286 286 10 2 2 6 2 2 39 36 3 3 56 54 2 2 65 6 Cl Cl/M1 Cl/M1/Adj Cl/M1/Adj/Comm Cl/M1/Adj/Comm/Mood Cl/M1/Adj/Mood Cl/M1/Adj/Pol Cl/M1/Adj/Pol/Mood Cl/M1/Circ-Adj Cl/M1/Circ-Adj/Residue Cl/M1/Circ-Adj/WH Cl/M1/Circ-Adj/WH/Residue Cl/M1/Comp Cl/M1/Comp/Residue Cl/M1/Comp/WH Cl/M1/Comp/WH/Residue Cl/M1/Fin Cl/M1/Fin/Modl 6 35 18 18 6 6 3 3 45 45 1 67 60 4 4 3 3 Cl/M1/Fin/Modl/Mood Cl/M1/Fin/Mood Cl/M1/Fin/Pred Cl/M1/Fin/Pred/Mood-Residue Cl/M1/Fin/Temp Cl/M1/Fin/Temp/Mood Cl/M1/Modal-Met Cl/M1/Modal-Met/Mood Cl/M1/Pred Cl/M1/Pred/Residue Cl/M1/Residue Cl/M1/Subj Cl/M1/Subj/Mood Cl/M1/Subj/TH Cl/M1/Subj/TH/Mood Cl/M1/Subj/WH Cl/M1/Subj/WH/Mood Table 4-6: Tags (35) used in the analysis of Figure 4-10 Note that here, only level 1 tags (i.e. only “M1” tags, not “M2”, “M3”) are analysed for recurrence. This prevents the embedded clauses, which contain many embedded tags, from dominating the patterns in the recurrence plot. 140 Figure 4-10: Differences in Mood between all clauses (M1 tags –XOR) LEGEND (for Figure 4-10) Darkest Red: XOR count = 0 No difference e.g. (1, 1) (2, 2) Darkest Blue: XOR count = 15 Largest difference e.g. (20, 64) (55, 64) In the colour scale of the recurrence plot, all lighter shades denote numerical values (in whole numbers) in a continuous linear spectrum between the two opposite extreme values represented by red and blue. In other words, a white or “blank” pixel 141 does not denote zero, but a value between zero and the largest difference. For example, in Figure 4-10, a blank pixel represents the midpoint between 0 and 15, i.e. 7 or 8. Each pixel in the recurrence plot represents the XOR count for a certain pair of clauses. The location of the pixel in the plot, according to the horizontal and vertical axes, determines which particular pair of clauses are under comparison. This plot is 80 by 80 pixels. We will use the terms “clause” and “row/column of pixels” interchangeably. The horizontal and vertical scale numberings in the plot label the pixels before the markings, not after; see Figure 4-11: Figure 4-11: Scale numbering in Figure 4-10 Clause 1 to 6 are not clauses. They’re just zero vectors, but have been included in the plot nonetheless, as reference – because they easily show the vector magnitudes of the “real” clauses, i.e. the clauses that appear bluest in the left column have a large number of total tags, the clauses that appear slightly red have almost no tags at all. All XOR recurrence plots, such as Figure 4-10, will have a prominent “Red Diagonal” from top left to bottom right. The pixels in this diagonal denote the result of comparing the clauses to themselves, i.e. comparing clause 1 with clause 1, clause 2 with clause 2, etc. Hence the difference is always zero, and the diagonal is 142 uniformly dark red. Note that the recurrence plot is symmetric along the Red Diagonal, because a comparison of clause i with clause j yields the same result as a comparison of clause j with clause i. Interpretation of Figure 4-10 The most prominent feature in the figure is undoubtedly the blue row (or column) at clause 55, representing the comparison of clause 55 with every other clause in the text. The lack of red pixels in this row shows the clause to be consistently very different from the rest in terms of Mood choices, and the outstanding intensity of the blue colouring means no other clause in the entire text generates as much XOR count. Figure 4-12: Mood analysis of clause 55 From Figure 4-12, it initially seems that clause 55 is just a simple short clause that does not appear unique. Herein lies the power of the recurrence plot, in revealing what is not obvious. Clause 55 is unique because of what uncommon tags it has and what common tags it lacks. The following are excerpts from Table 4-6 to show how common/uncommon these tags are: Clause 55 lacks: 56 54 39 36 35 Cl/M1/Comp Cl/M1/Comp/Residue Cl/M1/Circ-Adj Cl/M1/Circ-Adj/Residue Cl/M1/Fin/Mood 143 Clause 55 has: 6 2 2 6 6 Cl/M1/Adj/Mood Cl/M1/Adj/Pol Cl/M1/Adj/Pol/Mood Cl/M1/Fin/Temp Cl/M1/Fin/Temp/Mood The significance of the uniqueness of clause 55 extends beyond that of mere numbers of tags, when we examine the discourse context. Figure 4-13: Logical complexing context around clause 55 In Figure 4-13, we can recognize that clause 55 executes a significant turn in interpersonal meaning within the discourse. Clause 55 is a marked choice, because the point raised in clause 55 doesn’t exactly represent the author’s view. The overall argument of the entire sub-complex clause 48-58 was that “banks would not lend”. The author’s real argument began with clause 48. Clause 55 represented a kind of 144 false counterargument, which the author immediately knocks down in the next clause, clause 56, beginning with an adversative “but”. Clause 55 thus represented an opponent’s point of view, which the author loads with two Mood Adjuncts. It is hardly ever the case that the author uses even a single Mood Adjunct when presenting his own point of view. The lack of a complement and circumstantial adjuncts result in an unusually short clause, with immediate impact. The absence of a complement, in particular, is significant because the author frequently makes use of rankshifted complements to make his own propositions, e.g. in many of the relational identifying clauses, see Table 3-23 on p. 62. 4.3.2 Theme The purpose of this section is to demonstrate the usefulness of the XOR recurrence plot on another metafunction, as well as to assess the applicability of the AND recurrence plot. All “TH1” tags are extracted for analysis: 164 164 4 2 2 2 2 2 71 23 23 Cl Cl/TH1 Cl/TH1/Int Cl/TH1/Int/Mod Cl/TH1/Int/Mod-Met Cl/TH1/Int/Mod-Met/Theme Cl/TH1/Int/Mod/Adj Cl/TH1/Int/Mod/Adj/Theme Cl/TH1/Rheme Cl/TH1/Text Cl/TH1/Text/Conj 9 9 14 14 66 3 3 56 7 7 Cl/TH1/Text/Conj/Adj Cl/TH1/Text/Conj/Adj/Theme Cl/TH1/Text/Conj/Str Cl/TH1/Text/Conj/Str/Theme Cl/TH1/Topic Cl/TH1/Topic/TH Cl/TH1/Topic/TH/Theme Cl/TH1/Topic/Theme Cl/TH1/Topic/WH Cl/TH1/Topic/WH/Theme Table 4-7: Tags (21) used in the analysis of Figure 4-14 and Figure 4-22 145 Figure 4-14: Differences in Theme between all clauses (TH1 tags –XOR) The XOR recurrence plot for Theme is dominantly red. This is probably because of the simplicity of the system networks for Theme in Systemics, so there isn’t potential for large differences in “TH1”. The checkerboard structure in the plot reflects some extent of periodicity in the discourse, a good demonstration of the wavelike nature of the textual metafunction, theorized by Pike (1982), reiterated in Martin (1992). We examine the following patterns: red blocks of pixels, corresponding to clauses with similar selections in theme, and light blue strips, corresponding to clauses with unique Theme selections. 146 Figure 4-15: Patterns in thematic differences (from Figure 4-14) Figure 4-16: “Red Block 1”: clause 28-35 – consistent TH1 choices All six clauses in Red Block 1 (Figure 4-16) are thematically consistent with each other. The textual consistency is largely contributed to by the tight internal logical relations between the pair of clauses 28, 29, which forms a marked dependent 147 beta clause complex stemming from the main triplet clause complex 30, 31, 32. In Systemics, this dependency is also encoded in the textual metafunction under “TH1”, where clause 28, 29 are encoded entirely as Theme, and clause 30, 31, 32 are encoded entirely as Rheme. This similarity in “TH1” encoding is reflected in the recurrence plot. It also happens that following this five-clause complex, the next three clauses 33-35 also have similar simple Themes (each clause consisting of only two tags “Topic/Theme” and “Rheme”), therefore adding to the thematic consistency of Red Block 1. But this is likely no accident: the author makes use of this thematic consistency as a platform to develop his “salad analogy”. But what really defines the existence of Red Block 1 is the fact that the clauses before and after it, clauses 27 and 36, have structural conjunctions as textual Themes, coloured light blue-green in Figure 4-16. The absence of these structural conjunctions would make Red Block 1 merge with the surrounding landscape of red. Figure 4-17: “Red Block 2”: clause 58-63 – consistent TH1 choices 148 Red Block 2 is demarcated by clauses 57 and 64, which have thematic substitution “who” and “this”. In Systemics, they are not coded with the usual “Cl/TH1/Topic/Theme”, but instead coded differently to take into account their effect of cohesion and anaphora, as follows: Cl/TH1/Topic/WH/Theme Cl/TH1/Topic/TH/Theme A question might arise as to why clause 58, which is a thematic equative and the only such clause in the text, is considered part of the Red Block. Its thematic uniqueness should make it a blue strip. However in Systemics, “TH1” does not have the option of coding thematic equatives differently. This is a reflection of the software tagging limitations, as well as the user’s choice to select what patterns to visualize. Another question is why clause 59 and 62, which have marked Topical Themes, is part of this block which contains unmarked clauses. The reason is that in Systemics the “TH1” tag information does not state whether a Topical Theme is marked or not; this information is coded at the “Tex1” Keys. Figure 4-18: “Red Block 3”: clause 73-78 – consistent TH1 choices 149 Like Red Block 1, Red Block 3 seems to be demarcated by clauses with structural conjunctions. It is interesting that the Red Blocks occur in different major complexes. They represent parts of the discourse where the author accumulates text before some kind of conclusion, possibly to increase its impact. Now we shall examine the blue strips. Figure 4-19: “Blue Strip 1”: clause 24 – unique TH1 choices Figure 4-20: “Blue Strip 2”: clause 42, 43 – unique TH1 choices Figure 4-21: “Blue Strip 3”: clause 71 – unique TH1 choices The significance of Blue Strips 1 and 2 is immediately evident. They correspond not only to unique changes in the textual metafunction (owing to their textual Themes - “so”: continuative, “and”: structural conjunction) and Theme substitutions), but in the interpersonal metafunction as well, as these are the only clauses that have interrogative mood, and mark major turning points in the logical metafunction. Blue Strip 3 is more of a puzzle. It seems that clause 71 is almost the same as clause 70, but clause 70 is not part of the blue strip at all. The key happens to be the 150 semi-colon that begins clause 71. I have analysed it as a structural conjunction on the basis that it behaves like “but”. This is a small point, but the recurrence plot for theme is very sensitive to small differences. On top of this, I have also analysed “it is that” as Interpersonal Modal Metaphor, so that makes clause 71 very unique in theme indeed. To conclude the discussion of using recurrence plots in theme analysis, we observe that most of the Red Blocks (1 and 2) are found in the middle of large logico complexes, and most of the Blue Strips are found at the very start of them (1 and 2). This seems to be an accurate visualization of the textual development of a discourse: textual changes correspond to changes in logical metafunction. So far we’ve only looked at XOR plots. Now we consider a AND plot in Figure 4-22 next page, showing similarities in Theme between clauses. The AND plot uses a scale similar to the XOR plot (described on p. 141), which shows large numbers in blue and small numbers in red, so large AND counts are shown in blue in Figure 4-22. This means, in an AND plot, red corresponds to little similarity, and blue corresponds to high similarity. The uniform Red Diagonal for XOR plots does not exist here, because a pair of identical clauses can have a different AND count from another pair, depending on the number of tags. 151 Figure 4-22: Similarities in Theme between all clauses (TH1 tags –AND) The patterns in the AND plot for theme are noticeably different from the XOR plot we’ve seen earlier, though the input data are identical. Most prominent in Figure 4-22 is the five-pixel-wide red band comprising clause 28-32. The redness arises from the small number of tags in the clauses – each clause is labeled by only a single tag at TH1 (Theme or Rheme, not both), and this makes them different from every other clause (usually having both Theme and Rheme). Even if there were similarities, the AND count would be limited by the low number of tags. In fact, even the pixels representing the comparison of clauses 28-32 with themselves appear red. 152 There are also several single-pixel-wide red strips in the plot. The darkest ones are clause 10, 47, 74 and 80. This is a meaningful pattern captured by the recurrence plot because clauses 10, 47 and 74 are the only three non-finite ranking clauses in the text, consisting of only Rheme. Clause 80, too, only consists of Rheme, but it is a higher level Rheme – it is a finite clause acting as the alpha clause for the marked beta clause 79. The next most notable feature in the AND plot is a near-perfect “Blue Square” of clauses 69-72. Figure 4-23: “Blue Square” clauses in Figure 4-22 This represents clauses that are very similar to each other, but not only that, the paucity of blue in the recurrence plot means that clauses 69-72 must be very different from the rest. One important reason is that three of the four clauses have structural conjunctions. Another is that clause 70, 71 have interpersonal modal metaphors as Theme. Though these are only two tag types, the hierarchical decomposition of tags (see section 4.2.5, p. 137), together with the low AND count values everywhere else in the recurrence plot, make the blue square stand out in the plot. 153 4.3.3 Transitivity In this section the XOR recurrence plot will be used on the ideational metafunction. The OR recurrence plot, which has not been used in the previous two sections: 4.3.1 Mood and 4.3.2 Theme, will also be explored here. All “T1” tags are extracted for analysis: 249 249 3 2 1 22 21 4 7 1 21 7 2 2 2 2 1 5 1 4 10 16 6 10 2 15 10 2 5 13 5 Cl Cl/T1 Cl/T1/Accom Cl/T1/Accom/Add Cl/T1/Accom/Comt Cl/T1/Actor Cl/T1/Att Cl/T1/Att/Circ Cl/T1/Att/Possd Cl/T1/Attbr Cl/T1/Carr Cl/T1/Carr/Possr Cl/T1/Caus Cl/T1/Caus/Reas Cl/T1/Contg Cl/T1/Contg/Cond Cl/T1/Ex Cl/T1/Ext Cl/T1/Ext/Spat Cl/T1/Ext/Temp Cl/T1/Goal Cl/T1/Idd Cl/T1/Idd/Token Cl/T1/Idd/Value Cl/T1/Idd/Value/-Cl/T1/Idr Cl/T1/Idr/Token Cl/T1/Idr/Token/Circ Cl/T1/Idr/Value Cl/T1/Locn Cl/T1/Locn/Place 8 8 3 5 5 5 83 1 1 36 6 1 4 1 38 22 4 11 7 16 3 13 1 10 1 3 1 2 1 6 Cl/T1/Locn/Time Cl/T1/Mann Cl/T1/Mann/Means Cl/T1/Mann/Qual Cl/T1/Phen Cl/T1/Phen/Range Cl/T1/Proc Cl/T1/Proc/Behl Cl/T1/Proc/Exist Cl/T1/Proc/Mat Cl/T1/Proc/Ment Cl/T1/Proc/Ment/Aff Cl/T1/Proc/Ment/Cogn Cl/T1/Proc/Ment/Perc Cl/T1/Proc/Reln Cl/T1/Proc/Reln/Attr Cl/T1/Proc/Reln/Attr/Circ Cl/T1/Proc/Reln/Attr/Int Cl/T1/Proc/Reln/Attr/Poss Cl/T1/Proc/Reln/Iden Cl/T1/Proc/Reln/Iden/Circ Cl/T1/Proc/Reln/Iden/Int Cl/T1/Proc/Verbl Cl/T1/Range Cl/T1/Rec Cl/T1/Role Cl/T1/Role/Guise Cl/T1/Role/Prod Cl/T1/Sayer Cl/T1/Sens Table 4-8: Tags (61) used in the analysis of Figure 4-24 and Figure 4-27 154 Figure 4-24: Differences in Transitivity between all clauses (T1 tags –XOR) The most salient feature in Figure 4-24 is a blue strip at clause 50: Figure 4-25: Transitivity analysis of clause 50 155 Clause 50 is a relational attributive clause, which is not a very unique transitive type, considering there are 36 relational clauses in the text, and 21 of them are relational attributive processes. What is responsible for generating the large XOR count, then, is the presence of four circumstances of transitivity, each of which are only present in few other clauses: 2 4 5 8 Cl/T1/Accom/Add Cl/T1/Ext/Temp Cl/T1/Locn/Place Cl/T1/Locn/Time And after node-splitting, we get three more rare tag types: 3 Cl/T1/Accom 5 Cl/T1/Ext 13 Cl/T1/Locn This means a comparison of clause 50 with most clauses would give an XOR count of at least 7. All the circumstances of transitivity in clause 50 ultimately function to express the magnitude of the crisis. The circumstance “from their collective asset base” serves to magnify the severity of the banks’ losses. The circumstances “alone”, “so far”, and “during the crisis” are limiting the scale of the losses in time and geographic location, but this limiting effect only applies to the direct ideation of the clause – indirectly, it has the effect of magnifying the potential of losses elsewhere in the world, and in future. The ideation of the immensity of the financial crisis is a dominant stream of meaning maintained throughout the article, evidenced by the existence of the lexical string displayed in Figure 4-26 next page. Clause 50 is one of the 27 clauses which contributes to the lexical string. 156 Figure 4-26: Lexical string of words amplifying scale 157 Figure 4-27: Sum of similarities and differences in Transitivity between all clauses (T1 tags –OR) We finally turn to the OR plot in Figure 4-27. Since OR count = XOR count + AND count, we interpret OR plots as displaying “the sum of similarities and differences” in meaning. Though this interpretation is not exactly intuitive, the resulting plot does appear interesting: the most salient feature in Figure 4-27, is the same feature in Figure 4-24, but appearing even more prominent. But apart from the blue strip, there is also another interesting aspect about the plot. It seems to be divided into two phases. 158 Figure 4-28: Two Phases in Transitivity (T1 tags –OR) Legend Black arrows denote material clauses (29 of them) The OR plot seems to reveal the existence of two phases of transitivity in the text, outlined in Figure 4-28 above. The phases are distinguished by the abundance of red pixels (distributed evenly in a checkerboard fashion) in Phase I and their relative absence in Phase II. This pattern was not conspicuous in the XOR plot (Figure 4-24, 159 p.155). Closer observation reveals that material clauses (see black arrows in Figure 4-28) are generating the red pixels. • Phase I contains mostly material clauses (25 out of 49) • Phase II contains hardly any material clauses (4 out of 25) In fact, Phase II contains mostly relational clauses, which make up most of the nonmaterial clauses in the text. The phase change from material to relational is an interesting logogenetic pattern which we will return to in Chapter 5: Singular Value Decomposition (SVD) in p. 194-194. A possible explanation for this material/relational distinction appearing in the OR recurrence plot is that material tags are hierarchically limited, e.g. Cl/T1/Proc/Mat Cl/T1/Actor Material clauses are also very similar to each other, so they generate lots of red pixels when compared amongst themselves. However, relational clauses have tag names with greater depth of hierarchy, e.g. Cl/T1/Idr/Token/Circ Cl/T1/Proc/Reln/Attr/Circ and therefore generate large XOR counts when compared with other different clauses, and large AND counts when compared amongst themselves. Recalling that OR count = XOR count + AND count, this means OR counts will always be large for relational clauses, hence they appear blue in the recurrence plot. As for why the phase distinction in the OR plot is clearer than in the XOR 160 plot, this is because in the XOR plot, the relational clauses generate red pixels when compared with themselves, giving rise to some red pixels in the Phase II region of the XOR plot. But for the OR plot, the relational clauses will have large AND counts which erase or lighten out the red pixels (in the perspective of moving from Figure 4-24 to Figure 4-27). 4.4 THREE LEVEL INVESTIGATION Lemke (2000) has advanced the view of semiotic systems as multi-scale dynamical systems. Lemke’s notion of dynamics refers to the relations between different scales in a semiotic system – at every level N scale, there is a lower scale, (N-1 scale), and a higher scale, (N+1). The units and interactions in the (N-1) scale are constitutive for the N scale, and those of the (N+1) scale are constraining for the N scale. Drawing inspiration from his concepts, I have conducted a three level investigation of my discourse data. However, the analogy is not direct. While the scales Lemke (2000) discusses are physical scales, the “levels” in my investigation corresponds to the level of delicacy of analysis of one text alone. In Systemics terms, the level of delicacy is the number of tag types that is extracted from the database for visualization. Three XOR recurrence plots are made for the analysis of the CNN main story: 161 < First Level > Speech Function, Mood, Mood Metaphor tags < Second Level > All Int1 tags (i.e. all First Level tags, plus Modality, Modalityorientation, Mood Adjunct, Tense and Polarity tags) < Third Level > All Int1 and TH1 tags The recurrence plots in this section are different from the recurrence plots we’ve seen previously – the pixels are color-coded in a rainbow coloured spectrum, not just blue and red. The colour scale is shown to the right of the recurrence plot, which also shows the highest XOR count value represented by the darkest pixel. Also, the tags are not hierarchically decomposed – only end-nodes are used to construct the vector dimensions. The “false clauses” clause 1 to 6 are not shown here. Technical limitations also prevent the last clause, clause 80, from appearing – only 73 of 74 clauses are shown in the plots. 4.4.1 First Level of Delicacy The tags in Table 4-9 below are used to generate the First Level recurrence plot in Figure 4-29 next page. 8 3 3 59 2 7 3 1 70 An/Int1/Mood metaphor An/Int1/Mood/WH-interrogative/full An/Int1/Mood/declarative/ellipsed An/Int1/Mood/declarative/full An/Int1/Mood/imperative/full An/Int1/Mood/non-finite An/Int1/SPEECH-FUNCTION/Action/command An/Int1/SPEECH-FUNCTION/Action/offer An/Int1/SPEECH-FUNCTION/Knowledge/statement Table 4-9: < First Level > tags (9) analysed 162 Figure 4-29: < First Level > Speech Function, Mood, Mood Metaphor tags –XOR plot The colour scale on the right shows the XOR count values range from 0 to 5 in this plot. The seven strips of darkest blue representing greatest change are clauses 11, 24, 35, 42, 43, 75, 78, with the following Int1 analyses: 163 Figure 4-30: Analysis of clauses with greatest changes in First Level Note that all the seven clauses are mood metaphors, representing almost all of the total of eight mood metaphors in the text (the eighth being clause 80, which is not included in the recurrence plot in Figure 4-29). 164 4.4.2 Second Level of Delicacy The tags in Table 4-10 below are used to generate the Second Level recurrence plot in Figure 4-31 next page. 2 1 6 3 1 3 2 3 1 3 8 3 3 59 2 7 5 62 3 1 70 8 20 36 An/Int1/MODALITY-orientation/objective/explicit An/Int1/MODALITY-orientation/subjective/explicit An/Int1/MODALITY-orientation/subjective/implict An/Int1/MODALITY/Modalization/probability/median An/Int1/MODALITY/Modulation/inclination/median An/Int1/MODALITY/Modulation/obligation/median An/Int1/MOOD-ADJUNCT/degree An/Int1/MOOD-ADJUNCT/intensity An/Int1/MOOD-ADJUNCT/polarity/positive An/Int1/MOOD-ADJUNCT/time An/Int1/Mood metaphor An/Int1/Mood/WH-interrogative/full An/Int1/Mood/declarative/ellipsed An/Int1/Mood/declarative/full An/Int1/Mood/imperative/full An/Int1/Mood/non-finite An/Int1/POLARITY/negative An/Int1/POLARITY/positive An/Int1/SPEECH-FUNCTION/Action/command An/Int1/SPEECH-FUNCTION/Action/offer An/Int1/SPEECH-FUNCTION/Knowledge/statement An/Int1/TENSE/future An/Int1/TENSE/past An/Int1/TENSE/present Table 4-10: < Second Level > tags (24) analysed The Second Level recurrence plot analysis includes more dimensions of interpersonal meaning in addition to the dimensions in First Level. The strips of greatest change of meaning in the First Level still fairly visible in the Second Level, representing the fact that First Level features have an impact on the recurrence landscape at the Second Level. 165 Figure 4-31: < Second Level > Int1 tags –XOR plot The most obvious difference as we move from the First to the Second Level is that the pervasive sea of red is no more. Change is everywhere. An amazing feature of this recurrence plot however, is the existence of an “island” of stability within this “sea of change”, represented by the Red Square of clauses 60-65. The six clauses have the same analysis under “Int1”: Figure 4-32: Analysis of clauses in “Red Square” Analysis of the Red Square reveals a configuration of interpersonal meanings 166 that represents the most direct, common way of presenting information to an audience, without modality; presenting opinions as a matter of fact. Figure 4-33: Logico-complexing context of “Red Square” The six clauses within this Red Square are actually part of the logical subcomplex comprising clause 60-67, which elaborates on clause 59, clarifying what will lead to the “earthquake” between the government and financial markets. Clause 59 is in future tense and clause 66 has a Mood Adjunct “all”, so they depart from the Red Square configuration. But what is significant is that within the Red Square, many clauses actually present different interpersonal meanings e.g. modality is present in clause 60, 61 in a tokenized grammatical metaphor and in clause 64, negative polarity “cannot” is rankshifted. It is only by disguising these deviations, using the grammatical resources of rankshifting and grammatical metaphor, that the Red Square is possible, and with it, the author creates a more persistent, consistent and positive 167 “information giving mode” that adds credibility to his opinions. 4.4.3 Third Level of Delicacy The tags in Table 4-11 below are used to generate the Third Level recurrence plot in Figure 4-34 next page. 2 1 6 3 1 3 2 3 1 3 8 3 3 59 2 7 5 62 3 1 70 8 20 36 2 2 71 9 14 3 56 7 An/Int1/MODALITY-orientation/objective/explicit An/Int1/MODALITY-orientation/subjective/explicit An/Int1/MODALITY-orientation/subjective/implict An/Int1/MODALITY/Modalization/probability/median An/Int1/MODALITY/Modulation/inclination/median An/Int1/MODALITY/Modulation/obligation/median An/Int1/MOOD-ADJUNCT/degree An/Int1/MOOD-ADJUNCT/intensity An/Int1/MOOD-ADJUNCT/polarity/positive An/Int1/MOOD-ADJUNCT/time An/Int1/Mood metaphor An/Int1/Mood/WH-interrogative/full An/Int1/Mood/declarative/ellipsed An/Int1/Mood/declarative/full An/Int1/Mood/imperative/full An/Int1/Mood/non-finite An/Int1/POLARITY/negative An/Int1/POLARITY/positive An/Int1/SPEECH-FUNCTION/Action/command An/Int1/SPEECH-FUNCTION/Action/offer An/Int1/SPEECH-FUNCTION/Knowledge/statement An/Int1/TENSE/future An/Int1/TENSE/past An/Int1/TENSE/present Cl/TH1/Int/Mod-Met/Theme Cl/TH1/Int/Mod/Adj/Theme Cl/TH1/Rheme Cl/TH1/Text/Conj/Adj/Theme Cl/TH1/Text/Conj/Str/Theme Cl/TH1/Topic/TH/Theme Cl/TH1/Topic/Theme Cl/TH1/Topic/WH/Theme Table 4-11: < Third Level > tags (32) analysed 168 Figure 4-34: < Third Level > Int1 and TH1 tags –XOR plot At the Third Level, we move beyond looking solely at the interpersonal metafunction, and combine the dimensions of interpersonal meanings we’ve seen earlier with some dimensions of textual meaning. This results in a recurrence landscape with even more variation. But it still retains some features that have been identified in the First and Second Levels: the blue strips from the First Level at clauses 11, 24, 35, 42, 43, 75, 78, and a more diminished Red Square at clauses 61-63. There are features in the Third Level, however, that are arguably new, i.e. not attributable to the First and Second Levels, e.g. the double dark blue strips at clause 169 42 and 43 - though they also exist in the First Level, they almost vanish by the Second Level, but now they are the most prominent of the blue strips, see Figure 4-35. Figure 4-35: Change in meaning at clause 42, 43 compared across three levels of delicacy 170 The text of clauses 42, 43 and their analysis in TH1 are: Figure 4-36: Clause 42 & 43 TH Analysis Looking back at the TH1 XOR recurrence plot Figure 4-15 on p. 147, we find that clauses 42 and 43 are prominent strips. So it is their uniqueness in thematic meaning that cause them to reappear as blue strips in the Third Level (in Figure 4-35), despite having disappeared in moving from First to Second Level. This reappearance shows how changes in meaning (in a small selection of dimensions) can cause a clause to stand out, despite the fact that a massive quantity of meanings being simultaneously conveyed in multitudes of other dimensions. This disappearance and reappearance of features in the recurrence landscape goes to show how markedness in meaning depends on the perspective, and how perspectives across metafunctions can be combined. 171 4.5 COMBINED METAFUNCTIONS Finally we combine the analysis across textual, ideational and interpersonal metafunctions, extracting TH1, Tex1, T1, E1, Exp1, M1, Int1 tags. 4 3 9 1 50 7 2 1 6 3 1 3 2 3 1 3 8 3 3 59 2 7 5 62 3 1 70 8 20 36 3 5 21 1 4 3 38 9 3 8 52 1 1 16 1 36 71 83 42 2 6 2 36 3 54 2 6 35 18 An/Exp1/VOICE/effective/active/agent An/Exp1/VOICE/effective/active/ell-agent An/Exp1/VOICE/effective/passive/agent An/Exp1/VOICE/effective/passive/ell-agent An/Exp1/VOICE/middle/medium An/Exp1/VOICE/none An/Int1/MODALITY-orientation/objective/explicit An/Int1/MODALITY-orientation/subjective/explicit An/Int1/MODALITY-orientation/subjective/implict An/Int1/MODALITY/Modalization/probability/median An/Int1/MODALITY/Modulation/inclination/median An/Int1/MODALITY/Modulation/obligation/median An/Int1/MOOD-ADJUNCT/degree An/Int1/MOOD-ADJUNCT/intensity An/Int1/MOOD-ADJUNCT/polarity/positive An/Int1/MOOD-ADJUNCT/time An/Int1/Mood metaphor An/Int1/Mood/WH-interrogative/full An/Int1/Mood/declarative/ellipsed An/Int1/Mood/declarative/full An/Int1/Mood/imperative/full An/Int1/Mood/non-finite An/Int1/POLARITY/negative An/Int1/POLARITY/positive An/Int1/SPEECH-FUNCTION/Action/command An/Int1/SPEECH-FUNCTION/Action/offer An/Int1/SPEECH-FUNCTION/Knowledge/statement An/Int1/TENSE/future An/Int1/TENSE/past An/Int1/TENSE/present An/Tex1/Theme/beta-clause An/Tex1/Theme/multiple/interpersonal-1 An/Tex1/Theme/multiple/textual-1 An/Tex1/Theme/multiple/textual-2 An/Tex1/Theme/none/alpha-clause An/Tex1/Theme/none/non-finite An/Tex1/Theme/simple An/Tex1/Topic/marked/-An/Tex1/Topic/marked/dependent An/Tex1/Topic/none An/Tex1/Topic/unmarked/-An/Tex1/Topic/unmarked/ellipsed An/Tex1/Topic/unmarked/rankshift Cl/E1/Agent Cl/E1/Benef Cl/E1/Circ Cl/E1/Medm Cl/E1/Proc Cl/E1/Range Cl/M1/Adj/Comm/Mood Cl/M1/Adj/Mood Cl/M1/Adj/Pol/Mood Cl/M1/Circ-Adj/Residue Cl/M1/Circ-Adj/WH/Residue Cl/M1/Comp/Residue Cl/M1/Comp/WH/Residue Cl/M1/Fin/Modl/Mood Cl/M1/Fin/Mood Cl/M1/Fin/Pred/Mood-Residue 6 3 45 1 60 4 3 2 1 22 10 4 7 1 14 7 2 2 1 1 4 10 6 8 2 8 2 5 5 8 3 5 5 1 1 36 1 4 1 4 11 7 3 13 1 10 1 1 2 1 6 2 2 71 9 14 3 56 7 Cl/M1/Fin/Temp/Mood Cl/M1/Modal-Met/Mood Cl/M1/Pred/Residue Cl/M1/Residue Cl/M1/Subj/Mood Cl/M1/Subj/TH/Mood Cl/M1/Subj/WH/Mood Cl/T1/Accom/Add Cl/T1/Accom/Comt Cl/T1/Actor Cl/T1/Att Cl/T1/Att/Circ Cl/T1/Att/Possd Cl/T1/Attbr Cl/T1/Carr Cl/T1/Carr/Possr Cl/T1/Caus/Reas Cl/T1/Contg/Cond Cl/T1/Ex Cl/T1/Ext/Spat Cl/T1/Ext/Temp Cl/T1/Goal Cl/T1/Idd/Token Cl/T1/Idd/Value Cl/T1/Idd/Value/-Cl/T1/Idr/Token Cl/T1/Idr/Token/Circ Cl/T1/Idr/Value Cl/T1/Locn/Place Cl/T1/Locn/Time Cl/T1/Mann/Means Cl/T1/Mann/Qual Cl/T1/Phen/Range Cl/T1/Proc/Behl Cl/T1/Proc/Exist Cl/T1/Proc/Mat Cl/T1/Proc/Ment/Aff Cl/T1/Proc/Ment/Cogn Cl/T1/Proc/Ment/Perc Cl/T1/Proc/Reln/Attr/Circ Cl/T1/Proc/Reln/Attr/Int Cl/T1/Proc/Reln/Attr/Poss Cl/T1/Proc/Reln/Iden/Circ Cl/T1/Proc/Reln/Iden/Int Cl/T1/Proc/Verbl Cl/T1/Range Cl/T1/Rec Cl/T1/Role/Guise Cl/T1/Role/Prod Cl/T1/Sayer Cl/T1/Sens Cl/TH1/Int/Mod-Met/Theme Cl/TH1/Int/Mod/Adj/Theme Cl/TH1/Rheme Cl/TH1/Text/Conj/Adj/Theme Cl/TH1/Text/Conj/Str/Theme Cl/TH1/Topic/TH/Theme Cl/TH1/Topic/Theme Cl/TH1/Topic/WH/Theme Table 4-12: Combined metafunctions - tags (118) analysed 172 Figure 4-37: Combined Metafunctions XOR plot The first observation is how the whole plot is almost uniformly blue, except for some scattered lone yellow/orange pixels. This shows how unique each clause is with respect to all the other clauses. Yet even within this dense jungle of 118 dimensions, there are discernible patterns – light blue bands and dark blue bands. The most prominent dark band and light band are labeled in Figure 4-38 next page. Gone are the single strips of blue, or checkerboard patterns of red squares. Remarkably, some vestiges of the Red Square (clauses 60-65) of Figure 4-31, p. 166 remain in the Combined Metafunctions plot, in the form of light blue and yellow 173 pixels, in a “Light Square” labeled in Figure 4-38. Figure 4-38: Patterns in Combined Metafunctions XOR plot The most interesting feature in the combined plot is actually the Light Band. In most of the recurrence plots in this chapter, we discussed strips of differences, but this light blue band represents the similarity of these clauses with every other clause in the text. This similarity is only a kind of “relative similarity” against the background of dark blue, because light blue actually denotes a rather high absolute value of XOR count, of about 16. Figure 4-39: Clauses in the “Light Band” 174 The factors contributing to the existence of the Light Band’s colouration (reflecting relatively low XOR count) are many – textual, interpersonal and ideational. A remarkable property of the five strips comprising the Light Band is that the individual strips are fairly continuous in lightness from clause 7 to 79. Many other light blue strips are found elsewhere in the plot, but they are usually not as continuously light, but prone to interruption by darker pixels. The best possible explanation is that these five clauses have universal properties, i.e. they generally have unmarked simple Themes, common transitive processes (relational/material), no circumstances of transitivity, and lack modality or marked mood choices. While these are choices which are individually unremarkable, their simultaneous occurrence continuous over five clauses - brings the Light Band into existence. Therefore, the Light Band represents the most typical clausal configurations of functionality in Smick’s story. From the discourse context, the Light Band is a point where the author brings out the sobering reality of the crisis using facts and figures, e.g. “The U.S. financial system will have a whopping $15 trillion to $20 trillion less credit…” (clause 52) and does not invoke playful lexical metaphors (“driving policy”, “magic pill”, “mountains of financial paper instruments”, “ocean of new capital”, “earthquake”) and emotions (“painful”, “brutally”) or employ modality or change in mood, as is the case elsewhere in the text. The Dark Band of clauses 28-32 is actually a feature we’ve seen earlier in the TH1 AND recurrence plot, Figure 4-22, p. 152. We have also pointed it out as the largest clause complex in the article in the section 3.1: Logical Complexes, p. 27. 175 Figure 4-40: Clauses in the “Dark Band” The complete analysis of the clauses in the Dark Band is located in Appendix 2, p. 265. The five clauses’ appearance as a feature in the combined metafunctions recurrence plot show that either their uniqueness in theme has been carried over to the other metafunctions, or their uniqueness in theme alone has generated such large XOR counts that makes them emerge as features above the changes in other metafunctions in other clauses. It is possible that the significance of this Dark Band is just a by-product of coding conventions in Systemics at TH1 and TH2, for clause complexes with marked dependent clauses. Regardless of any analysis or coding convention, these clauses undoubtedly do stand out from the rest of the article in terms of meaning because they introduce the “salad” into the text. Finally moving to the “Light Square”, which isn’t a very discernible feature of Figure 4-37, but recognizable because it appeared in clear form in the “Second Level” recurrence plot (Figure 4-31). The remnants of that Red Square in this combined metafunction plot demonstrates how features of interpersonal metafunction can impact the features in a combined metafunctional view. 176 Conclusion on Recurrence The advantage of the recurrence plots for systemic functional research is probably the ability to combine the functional meanings across different metafunctions and display the results in a mathematical way. One result of this is that we find the units at which changes in meaning occur are different under different scales of perspective. The combined metafunction recurrence plot has features that comprise of at least four clauses. Unlike in the other plots of lower dimensions, there are no singleclause sized features evident: there are many dark strips, but they are so numerous that they cannot be considered outstanding. This seems to suggest that the more the dimensions of meaning we examine, the larger the units of discourse at which significant features occur will be. We have also observed how the OR plot of Figure 4-27 smoothens out small scale variations in the landscape of recurrence, and makes large scale phases in the text clearer. Since the OR plot is generated by summing meanings in both clauses, the patterns in an OR plot in fact show patterns arising from pairs of clauses, not single clauses. The results of the experiments in this chapter illustrate that the patterns of meaning in discourse occur at all scales of measurement: with single clauses or pairs of clauses, at low dimensionality or high dimensionality, within individual metafunctions, and across metafunctions. 177 _____________________________________________________________________ Chapter 5: SINGULAR VALUE DECOMPOSITION (SVD) _____________________________________________________________________ Systemic functional analysis decomposes text into clauses which are composed of different combinations of tags. The tags can be conceived as orthogonal dimensions that define an abstract multidimensional Semantic Space. A clause can be formulated as a vector, visualised as a point located in this Semantic Space. A text, being a collection of clauses, can hence be considered as a collection of vectors, which can be expressed as a matrix. The matrix can be visualised as a collection of points (like a cloud) in Semantic Space. However a problem with such a representation is that all the tags used to describe the cloud are treated as equal (equally different), but some tags are more closely related than others, according to theory and depending on context. The mathematical technique of Singular Value Decomposition (SVD) applied to the matrix representation of discourse, offers a perspective that addresses this issue. Application of SVD in linguistics is not new; it has been used in LSA to analyse lexical distributions for the past two decades (see Landauer 2007). SVD is a matrix decomposition technique that decomposes the discourse matrix to describe the clauses as being compositions of Features instead of tags. Mathematically, a Feature (capitalised for disambiguation) is a vector defined as a linear combination of tags13. This can be interpreted as: a Feature is defined as a combination of various degrees of 13 Vectors are like words: they can be defined in terms of each other. Tags, clauses, Features are all vectors, and any one of them can be defined in terms of either of the other two. 178 the importance of absence/presence of certain tags. These Features are what we can use to better characterise the discourse than tags, because SVD is sensitive to the distribution of clauses in Semantic Space. Each Feature combines tags that are related to each other, based on the associations and oppositions of clauses in Semantic Space. Though it is theoretically more difficult to think of a clause as a combination of Features, Features provide a more economical means of characterising a particular text as a whole. Figure 5-1: Visualization of SVD in two dimensions14 Figure 5-1 provides a simple visualization of the process of SVD, where the Semantic Space is only two-dimensional. The clauses are the dots and the blue ellipse represents the shape of their distribution. SVD generates a new coordinate system comprising Feature 1 and Feature 2, which are aligned with the major and minor axes of the ellipse. The Features are more suited to the description of the distribution of clauses than the tags. A complete formal mathematical account of SVD is provided by Trefethen & 14 The figure does not use the actual SF data; the positions of the dots are for illustrative purposes only. In the actual data, there should not be any clause in the negative region of tagspace, and clauses should only occupy whole number values along each tag vector. 179 Bau (1997). SVD is implemented in Systemics 1.2.7 with algorithms from Press et al. (1994). The primary concern here is not how SVD works, but rather the value of its application to SF data. A brief technical introduction to SVD is non-trivial and will therefore not be given in this chapter, but rather placed in Appendix 5: SVD Definition p. 387. Some aspects of SVD, however, are necessary for understanding how to interpret the SF analysis by SVD – hence this chapter adopts the following approach: SVD will be introduced alongside the discussion of actual linguistic data. In this thesis, the results of SVD of SF data are presented in the form of large visualization plots generated by Systemics (A3 paper size foldouts). This chapter is devoted to the discussion of the SVD visualizations plots, but to avoid disrupting the flow of the chapter, all the SVD plots are not contained within this chapter but compiled in a separate volume, Volume II, Appendix 6, p. 390-412. SVD is a technique useful for capturing overall patterns in a system, identifying polarities in a large dataset with many variables, and identifying latent correlations. It is these objectives that my explorations of SF data with SVD are based on. An appropriate starting point, therefore, is transitivity, due to its complexity and ease of interpretation. 5.1 EXPERIMENT 1: MAIN STORY All “T1” tags in the main story are extracted with a “Cl/T1*/*” search command in the Systemics Search Page. A raw matrix, of clauses (rows) against 180 “T1” tags (columns), is then output. The entries of the raw matrix show the number of tags of a particular type in a particular clause. The matrices listed in Appendix 3: SF Analysis Matrices (p. 368) are all raw matrices. The raw matrix is then converted into a reduced matrix by reducing all matrix entries larger than 1, down to 1. This means the reduced matrix will only have two possible entries: “1” denoting the tag type is present in a clause (in whatever numbers), or “0” denoting the tag type is absent. The reduced matrix becomes the input matrix, on which SVD is performed. Table 5-1 and Table 5-2 below summarise the diversity of extracted T1 tags in the reduced matrix. Relational 36 Material 29 Mental 6 Behavioural 1 Verbal 1 Existential 1 Total 74 Table 5-1: Overview of transitive processes in the main story 74 1 1 29 6 1 4 1 36 21 4 10 7 15 3 12 1 Cl/T1/Proc Cl/T1/Proc/Behl Cl/T1/Proc/Exist Cl/T1/Proc/Mat Cl/T1/Proc/Ment Cl/T1/Proc/Ment/Aff Cl/T1/Proc/Ment/Cogn Cl/T1/Proc/Ment/Perc Cl/T1/Proc/Reln Cl/T1/Proc/Reln/Attr Cl/T1/Proc/Reln/Attr/Circ Cl/T1/Proc/Reln/Attr/Int Cl/T1/Proc/Reln/Attr/Poss Cl/T1/Proc/Reln/Iden Cl/T1/Proc/Reln/Iden/Circ Cl/T1/Proc/Reln/Iden/Int Cl/T1/Proc/Verbl 181 7 2 4 1 Cl/T1m/Proc Cl/T1m/Proc/Behl Cl/T1m/Proc/Mat Cl/T1m/Proc/Verbl Table 5-2: Frequency distribution of all T1 process types in reduced matrix 74 74 3 2 1 22 21 4 7 1 21 7 2 2 2 2 1 5 1 4 10 15 5 10 2 15 10 2 5 12 5 8 8 3 5 5 5 74 1 Cl Cl/T1 Cl/T1/Accom Cl/T1/Accom/Add Cl/T1/Accom/Comt Cl/T1/Actor Cl/T1/Att Cl/T1/Att/Circ Cl/T1/Att/Possd Cl/T1/Attbr Cl/T1/Carr Cl/T1/Carr/Possr Cl/T1/Caus Cl/T1/Caus/Reas Cl/T1/Contg Cl/T1/Contg/Cond Cl/T1/Ex Cl/T1/Ext Cl/T1/Ext/Spat Cl/T1/Ext/Temp Cl/T1/Goal Cl/T1/Idd Cl/T1/Idd/Token Cl/T1/Idd/Value Cl/T1/Idd/Value/-Cl/T1/Idr Cl/T1/Idr/Token Cl/T1/Idr/Token/Circ Cl/T1/Idr/Value Cl/T1/Locn Cl/T1/Locn/Place Cl/T1/Locn/Time Cl/T1/Mann Cl/T1/Mann/Means Cl/T1/Mann/Qual Cl/T1/Phen Cl/T1/Phen/Range Cl/T1/Proc Cl/T1/Proc/Behl 1 29 6 1 4 1 36 21 4 10 7 15 3 12 1 10 1 3 1 2 1 6 7 2 2 2 5 3 2 7 2 4 1 2 1 1 1 1 Cl/T1/Proc/Exist Cl/T1/Proc/Mat Cl/T1/Proc/Ment Cl/T1/Proc/Ment/Aff Cl/T1/Proc/Ment/Cogn Cl/T1/Proc/Ment/Perc Cl/T1/Proc/Reln Cl/T1/Proc/Reln/Attr Cl/T1/Proc/Reln/Attr/Circ Cl/T1/Proc/Reln/Attr/Int Cl/T1/Proc/Reln/Attr/Poss Cl/T1/Proc/Reln/Iden Cl/T1/Proc/Reln/Iden/Circ Cl/T1/Proc/Reln/Iden/Int Cl/T1/Proc/Verbl Cl/T1/Range Cl/T1/Rec Cl/T1/Role Cl/T1/Role/Guise Cl/T1/Role/Prod Cl/T1/Sayer Cl/T1/Sens Cl/T1m Cl/T1m/Actor Cl/T1m/Behav Cl/T1m/Goal Cl/T1m/Locn Cl/T1m/Locn/Place Cl/T1m/Locn/Time Cl/T1m/Proc Cl/T1m/Proc/Behl Cl/T1m/Proc/Mat Cl/T1m/Proc/Verbl Cl/T1m/Range Cl/T1m/Role Cl/T1m/Role/Guise Cl/T1m/Sayer Cl/T1m/Verb Table 5-3: Frequency distribution of all T1 tags in reduced matrix -77 tags 182 Table 5-3 (previous page) shows counts from the reduced matrix of all the “T1” tag types, and these reduced matrix counts can be interpreted as showing the number of clauses each tag type appears in. The purpose of Table 5-1, Table 5-2 and Table 5-3 is to facilitate the interpretation of the SVD results later in the chapter. These three tables show count statistics from the reduced matrix; alternatively, count statistics from the raw matrix have been shown earlier in Table 4-8 in Chapter 4: Recurrence Plots, p. 154. Some counts here in Table 5-3 differ from Table 4-8 (p. 154). This is because the reduced matrix counts the clauses, but the raw matrix counts the tags. In the raw matrix, erroneous double counting of tags sometimes occurs. For example, the process tag is split into two tags in “sliced them up” (clause 32). There are seven clauses with split “Mat” processes and two with split “Reln” processes. These problems do not occur in the reduced matrix. Another difference is that “T1m” tags are included here, but were excluded in Table 4-8. The Features generated by SVD have different “strengths”. The strength of a Feature is the measure of the overall contribution of the Feature to all the clauses. Features are named in descending order of their strength: F0, F1, F2, etc. We shall discuss only a handful of the strongest Features. Each Feature is essentially just a set of decimal numerical values (positive or negative), but for ease of interpretation, the numbers are visualised in a variety of plots in Systemics 1.2.7. The visualisations are colour coded according to the numerical values. Three types of SVD tinted plots are presented: Tag Wheel Plots, Text Plots and Neighbourhood Plots – all are compiled in Appendix 6: SVD Plots p. 390-412. 183 Tag Wheel Plots define a Feature – they are radially branching nodes representing the SF grammar which are colour tinted according to the contribution of tags to the Feature. Text Plots show the parts of the discourse exhibiting a Feature – they are lists of the full text of the clauses, colour tinted in whole and in parts to show the contributions of the clause and the clause constituents to the Feature. Features can also be described in two types of Neighbourhood Plots. Clause neighbourhood plots show the associations of clauses in Semantic Space as clusters of clauses in a 2D diagram, tinted according to the contribution of clauses to a Feature. Alternatively, neighbourhood plots can also be made of tags, showing clusters of closely related tags, based on their tendency to co-occur in clauses. The tints in tag neighbourhood plots show the contribution of tags to a Feature. The following five sections 5.1.1 to 5.1.5 will explain the significance of each Feature F0 to F4 derived from the SVD of the matrix of all the main story clauses, in all the “T1” tag dimensions. SVD plots will be used to describe each Feature. 5.1.1 Feature F0 F0 represents the “average clause”, the average position of all the clauses in Semantic Space. F0 is often not a Feature that we’re interested in, because an idea of the statistical average can be obtained from simple statistics like Table 5-1, without the need for a procedure as sophisticated as SVD. But the simplicity of F0 means F0 is ideal for introducing the SVD plots. Figure A-1 (please see Volume II, Appendix 6: SVD Plots p. 391) represents 184 F0 in a Tag Wheel Plot. A Tag Wheel Plot defines a particular Feature, as combinations of presence or “absence”15 of certain tags, by colouring the tags to different degrees. The right column in Figure A-1 represents the same information as the central wheel, with tags in a list instead of a hierarchical plot. The left column is a clause list that represents F0 in a different way, showing the strength of the Feature in the clauses. Key (for all SVD plots in Appendix 6, p. 391-412) -Blue indicates presence and Red indicates “absence” (or vice versa)16 -intensity of colouring, font size and the size of node/circle denote strength (for a tinted tag, strength refers to strength of tag’s contribution to the Feature; for a tinted clause, strength refers to the strength of the Feature in the clause) All F0 plots in Appendix 6 are only ever blue tinted, never red, but plots for F1, F2, onwards have a mix of red & blue tints. In the case of Figure A-1, blue (clearly) denotes presence (and not absence). The fact that “T1”, “Cl” and “Proc” are the most prominent (largest and most intensely coloured) tags is trivial, since every clause has these three tags. Rather, the important thing to note in Figure A-1 is that “Reln” and “Mat” are intense, and this is what we expect based on Table 5-1 (p. 181), which states that relational and material processes are the two most frequent process types. Though this is not a new result from SVD, it’s a satisfactory check. 15 Red does not simply mean absence in the sense of a zero in the vector, because most tags are absent in a Feature anyway. “absence” is explained further in p. 201. 16 In general, a Feature is actually a set of positive and negative numbers, and blue represents positive (which means presence) and red represents negative (which means “absence”), or vice versa. The colours can be swapped - see p. 201. 185 Figure A-2 (please see Volume II, Appendix 6, p. 392) is another visualisation of F0, displaying the tags in a different format. In this Tag Neighbours plot, tags most similar to each other are grouped into clusters. The grouping is not done by SVD; it is done by relying simply on the information in the reduced matrix. In the reduced matrix, clauses are described as vectors comprising tags. Looking at the matrix in another way, it also tells us how to view each tag as a combination of clauses. Tags that have the tendency to occur in the same clauses are considered “similar” and grouped together. The resulting configuration is what we expect: tags associated with the same transitive process tend to cluster together e.g. tags associated with relational identifying processes (“Idd”, “Idr”, “Proc/Reln/Iden”, etc) are grouped together in the top center. The results of the SVD as shown by the tinting here tell us that the three strongest contributors to F0 (“Cl”, “Cl/T1”, “Cl/T1/Proc”) occur in the same cluster. The next tinted tag nearest to these “strongest three” is “Cl/T1/Proc/Reln”, which is what we expect, given that “Reln” is the most frequent transitive process type. Apart from this most strongly tinted central cluster, two other clusters are fairly tinted, one above, representing relational identifying processes and one below, representing material processes. The strongest tints apparently occur in clusters. Again, SVD is not revealing anything new; but verifying what is already known. Figure A-3 (please see Volume II, Appendix 6, p. 393), a Clause Neighbours Plot, visualises the contribution of F0 in clauses in the text. This is really another way of defining F0, but instead of using tags as a basis, it uses clauses. This is a form of nearest neighbour plot, like Figure A-2, but of clauses rather than tags. Almost every 186 clause is tinted intensely (the numbers 1 to 6 are not clauses). Since F0 represents the average clause, this is exactly as expected. Note the difference in the extent of colouring in Figure A-2 and Figure A-3. In Figure A-2, only about 3 tags are coloured intensely and all the rest are white, while in Figure A-3 the converse is observed almost every clause is dark blue and a handful is white. This means that only a handful of tags are sufficient to define F0, but F0 is powerful enough to describe virtually all of the clauses. This is a welcome result – it shows SVD has captured the few universal properties of the text, with F0. In Figure A-3 we notice there are three clusters of clauses that are especially tight, and these are also the most intensely coloured. The clustering turns out to be very meaningful – each cluster groups together clauses of a particular transitive process type. The clusters are highlighted in Figure 5-2: Figure 5-2: Tightest clusters in Figure A-3 187 The clauses within these three clusters are as follows: (1) Relational attributive possessive (6) - top left cluster 8 47 48 57 70 76 : most policymakers lack a clue of what is really at stake. to remove the toxic salad from bank balance sheets. Policymakers have no means of forcing the banks to start lending who don't need loans. It is not that the world lacks money; We need a private/public global bank clearing facility. (2) Relational attributive intensive (6) - top right cluster 16 21 36 37 61 67 and credit markets remained frozen. Therefore, the housing crisis was a mere trigger for a collapse of trust in paper, and you're dead. The overall salad looked delicious, It seems doubtful. The collapse of, say, a major European bank would hardly leave American workers immune. (3) Relational identifying (8) (intensive & circumstantial) - bottom right cluster 20 33 34 56 58 60 73 75 Incredibly, these paper instruments were insured by more dubious paper instruments. The idea was to sell (for huge fees) individual servings of diversified financial salad around the world. The only problem: under an occasional piece of lettuce was a speck of toxic waste in the form of a defaulting subprime mortgage. -- but the fear is they will do it only to people such as Warren Buffett, What is uncertain is the amount of lending to borrowers engaged in entrepreneurial risk, the center of business reinvention and job creation. The great uncertainty is whether government has the power to rescue the financial system in times of crisis. The challenge will be to reform our financial system quickly The first step should be efforts to make the market for future asset-backed paper more transparent and credible. Table 5-4: Clauses in the three tightest clusters of clause neighbours in T1 It is expected for SVD to pick these clusters as the most intense, because large groups of clauses with very similar tags will have a large influence on the average position of all the clauses in Semantic Space. A good question to ask is why are the closest clusters all relational, given that there is a very large number of material processes too. A possible explanation is that material processes have a tendency to occur with a large variety of circumstantial elements and hence have greater tag diversity. 188 This concludes the exploration of F0. The purpose of studying F0 in detail was to familiarise with the visualisation tools in Systemics for SVD analysis. F0 is probably the easiest to interpret, because its characteristics are obvious. Having done this, we now look at the next few strong Features, which demonstrate the utility of SVD in providing revealing perspectives into the text that would otherwise be difficult or not possible. 5.1.2 Feature F1 F1, shown in Figure A-4 (Volume II p. 394), represents the polarity between material and relational processes – this is as we expect, given that material and relational processes are the two biggest groups of clauses that differ from each other in process type (see Table 5-1 p. 181). The darkest blue: Mat, Actor, Goal The darkest red: Reln, Attr, Carr, Att By “polarity” or “opposition”, I do not mean there is any kind of repulsive force between the clauses, but the point is that the presence of one category of tags correlates with the absence of certain other groups of tags. SVD is able to detect such tendencies. 189 More about SVD Before we continue exploring F1, we’ll now elaborate further on what SVD does: in particular, on the meaning behind the red/blue polarity in a SVD Feature. While it is easy to imagine how the presence of a tag can contribute to a Feature (i.e. the blue tags), it is not obvious how exactly the absence of tags can contribute to a Feature (i.e. the red tags). Nor is it clear why red/blue colours can be interchanged, as mentioned in the “Key to SVD plots” on page 185. Here we hope to clarify these matters. Figure 5-3: Meaning of F0 and F1 in SVD Figure 5-3 visualises the tags (X, Y) and Features of SVD (F0, F1) for a collection of clauses represented as small circles in two dimensions. The clauses really reside in a multi-dimensional Semantic Space - 77 dimensions for transitivity alone (Table 5-3, p. 182) - but for simplicity they are projected down to two dimensions in Figure 5-3 (depending on how the projection is done, X, Y may no longer represent the original tag vectors). The cloud is not a random scatter of points, 190 but has some kind of shape, and the distribution of clauses in space can be described in terms of Features that could have useful linguistic interpretation. It is this “spread” that SVD seeks to characterise – if all clauses represented exactly the same set of choices, then SVD would have nothing to offer us. To describe the distribution of clauses in Figure 5-3, firstly one could say that the clauses are not centred on the origin. This is the first Feature F0 that is picked up by the SVD. F0 points away from the XY origin towards the average position of all the clauses. Secondly, we can say that the clauses are mostly split into two clouds at two “opposing” ends. This polarity is exactly what the second Feature F1 describes. F1 points from the center of one cluster to the center of the other cluster. What SVD is doing is deriving a new set of vectors, F0, F1, from the original set X, Y, which better describes the distribution of the clauses. These two Features have different “strengths” because they contribute differently towards the description of the clauses. While F0 helps to bring you from the origin to the position of every clause, F1 is slightly less useful than F0, because it does not help in describing some clauses (e.g. the few white coloured circles near the tip of the F0 arrow), so F1 is weaker than F0. Note that clause positions described with the original basis vectors X & Y can only have positive values (denoting presence of a tag), or zero (denoting absence). However, another important property of Features is that they can contain negative values, e.g. red tints can be interpreted as “absence” of tags, or more accurately, as a kind of opposition to those tags: opposite to their presence elsewhere in Semantic 191 Space. Now, it does not make sense to define red always as associated with absence, because Features are not numbers; Features are directions (defined by numbers). Since F1 is just a vector used to describe the position of clauses, it does not matter if its direction is reversed, just as it doesn’t matter if we reverse the direction of a ruler that we use to measure distance. Therefore, it is also correct to describe F1 (say, in Figure A-4) as the absence of blue tags and the presence of red tags, this is tantamount to reversing the direction of the green arrow of F1 in Figure 5-3 (p. 190). Now let’s explain how “absence” of tags can be incorporated into the definition of a Feature. Figure 5-4 a, b: Contribution of presence and absence of tags to F1 Suppose blue are material clauses and red are relational clauses Figure 5-4. We introduce two vectors A and B to describe the average position of each cluster in Figure 5-4a. A and B are both positive valued, if defined in terms of the old basis X, Y. A can be interpreted as “travelling from the origin to the blue cluster” and B as “travelling from the origin to the red cluster”. Adding a negative sign to a vector reverses its direction, so “-B” means travelling from the red cluster to the origin, as shown in Figure 5-4b. Now, F1 can be interpreted as “travelling from the red to the 192 blue cluster”. Therefore, we can define F1 in terms of A and B as follows: F1 = A - B This formula is geometrically represented in Figure 5-4b. Figure 5-4b explains how a Feature can include the “absence” of tags, for it is both the presence of A and the “absence” of B that defines F1. This concludes our elaboration on the theory behind SVD. Let’s return to the F1 Tag Wheel Plot, Figure A-4 (p. 394). A more careful look at it reveals that the opposition is between not all the relational processes, but only the relational attributive. The tags associated with relational identifying processes are virtually untinted in Figure A-4. The reason is likely because (from Table 5-2) there are slightly less “Reln/Iden” processes (15) than “Reln/Attr” (21), and F1 picks out the stronger difference, leaving “Reln/Iden” associated characteristics to be picked up by weaker Features. The clause list (the column of coloured circles on the left in Figure A-4) also shows the logogenetic dynamics of this opposition – blue occurs more in the beginning of the text and red occurs more towards the end. The beauty of this visualisation of the clause list is that other clauses that are weak in this Feature simply fade into the background as untinted tiny circles. Particularly interesting are two long continuous stretches of nearly uninterrupted clauses of the same colour (transitive process type), see Figure 5-5 next page: 193 Figure 5-5:: Stretches of constant F1 values (in the clause list of Figure A-4) Material Processes (clause 25 to 32) -blue stretch 25 26 27 28 29 30 31 32 I like to use a salad analogy. Before the last decade, bankers simply lent in the form of syndicated loans. But with the huge expansion of the global economy in the 1990s, which produced an ocean of new capital, the bankers came up with an idea called c securitization. Instead of making simple loans and holding them until maturity, a bank collected all its loans together, then {the bank} diced {its loans} and {the bank} sliced them up into a big, beautiful tossed salad. Relational Attributive Processes (clause 65 to 72) -red stretch 65 66 67 68 69 70 71 72 Given such massive exposure, government guarantees in a time of crisis become meaningless. Yet because of the interconnected web of global financial relationships, we are all vulnerable to the threat. The collapse of, say, a major European bank would hardly leave American workers immune. Our policy leaders in Washington are thinking domestically when the solution to the credit crisis will be global. It is not that the world lacks money; it is that the world's money is sitting on the sidelines -- more than $6 trillion in idle global money markets alone. Table 5-5:: Stretches of text that continuously exhibit similar F1 values 194 The reason for this to happen is likely that material processes are more suitable for construction of past actions and events, hence their appearance early in the text, but relational processes are more useful for representing present and future states of being, hence their appearance towards the end of the text when the author moves to describe the implications of the crisis. This is the same reason for the pattern of phases in the recurrence OR plot for transitivity in Figure 4-28, p. 159. Figure A-5 (Volume II, p. 395), a Text Plot, is a visualisation of the Feature tints on the actual text of the clauses. The tinting is not uniform for the whole clause but sensitive to the different contributions by different tags in the same clause to the Feature. For example, in the first clause, clause 7, “one thing” and “certain” are Carrier (“Carr”) and Attribute (“Att”), the strongest red tags in Figure A-4, hence they appear as the reddest words. However, “is” is a fainter red, meaning that “Proc/Reln/Attr/Int” contributes less to F1 than “Carr” and “Att”17. “At this point in the credit crisis” is completely untinted, as it is a transitive circumstance that does not figure at all in F1. Note that “at least” is of a smaller font size, but tinted quite strongly red. “at least” is a Mood Adjunct that does not form part of the transitive structure of the clause, hence its smaller font, but “at least” is tinted strongly anyway, not because of the tag’s contribution (there’s no transitivity tag there), but because of the clause’s strong contribution to F1. This is perhaps not meaningful, but is a present limitation of the way the visualisation is performed by the software. Throughout Figure A-5, although it is the words that are tinted, the tinting is 17 But how can this be, since in Figure A-4, Reln and Attr look even stronger than Carr and Att? Most probably the tint on “is” is determined by the weak end-node Int in Proc/Reln/Attr/Int, and not the superordinate nodes Reln and Attr. 195 not based on the words in themselves, but on the tags that the words are assigned, and in some cases, the clause in which the words reside. So really, the Text Plot Figure A-5 can be considered as another form of tag plot. In Figure A-6 (Volume II, p. 396), note that the red tints are almost solely concentrated in one cluster, and the blue tints too. Since F1 is the strongest Feature (after F0, which is trivial because it shows universally occurring tags), this means probably that these two most intensely tinted clusters contain the most frequently occurring non-trivial tags: the tags that play the dominant role in distinguishing between clauses within the text from one another. Note that some faint blue tint spills outside the cluster, more importantly, to “Mann/Qual”, of which occurrences are shown in Table 5-6 below. This actually means two things: one, “Mann/Qual” have a tendency to co-occur with the blue (material process) tags, and two, they have a tendency not to occur with the red (relational attributive) tags. Let’s examine the text to see why. 13 30 38 43 68 material material material material mental which, , would quickly restore financial stability. a bank collected all its loans together, but suddenly global investors were no longer ordering salad. and why have financial markets collapsed so brutally? Our policy leaders [in Washington] are thinking domestically Table 5-6: All Mann/Qual occurrences in the text Table 5-6 shows that almost all the clauses in the text with “Mann/Qual” are material processes. This explains why “Mann/Qual” is tinted blue together with the “Mat”, “Actor”, “Goal” tags. The reason for this could be due to both theory and 196 context: “Mann/Qual” by its nature has the greatest potential to occur with material processes, and it just happens in this text that the material clauses do contain a few “Mann/Qual”. Figure A-7 (Volume II, p. 397) shows the clause tinting in a Clause Neighbourhood Plot. Figure A-7 should really be used in tandem with Figure A-6, so that tags associated with the strong tints can be easily identified. A remarkable observation in Figure A-7 is that every clause is tinted, at least to a light degree, even clauses that are neither material or relational. You can see the difference between the circles numbered 1 to 6, which are not clauses and are completely blank, and some clusters of clauses in the bottom right corner which are very slightly tinted. Tinting the Clause Neighbourhood Plot with F1 (as opposed to any other Feature) actually provides the best way to differentiate the clusters. The greatest opposition amongst the clauses is revealed by this tint – material and relational attributive. Counting the number of deep blue clauses in Figure A-7, you will find 29 of them, exactly the number of material clauses given in Table 5-1, p.181. Counting the number of deep red clauses gives 21, as was stated in Table 5-2, p. 181. According to the grammar of transitivity, relational attributive processes are subdivided into circumstantial, intensive and possessive types. In Figure A-7, we can see the deep red clauses are in three separate regions – one cluster in the top left, one cluster in the top right, and a few “renegade” clauses in the bottom. These three regions are highlighted in Figure 5-6. However, these three regions do not perfectly correspond to the three types of “Reln/Attr” clauses, Table 5-7 next page shows. 197 Figure 5-6: Deep red regions of F1 in Figure A-7 (1) Relational Attributive Possessive (7) - top left cluster 8 47 48 50 57 70 76 : most policymakers lack a clue of what is really at stake. to remove the toxic salad from bank balance sheets. Policymakers have no means of forcing the banks to start lending After all, the U.S. banks alone so far during the crisis have lost upwards of $2 trillion from their collective asset base. who don't need loans. It is not that the world lacks money; We need a private/public global bank clearing facility. (2) Relational Attributive Intensive/Circumstantial (10) - top right cluster -Intensive (6) 16 21 36 37 61 67 and credit markets remained frozen. Therefore, the housing crisis was a mere trigger for a collapse of trust in paper, and you're dead. The overall salad looked delicious, It seems doubtful. The collapse of, say, a major European bank would hardly leave American workers immune. -Circumstantial (4) 45 the world will face a serious credit crunch in 2009 198 69 71 72 when the solution to the credit crisis will be global. it is that the world's money is sitting on the sidelines -- more than $6 trillion in idle global money markets alone. (3) Relational Attributive Intensive (4) - bottom few clauses -pair one 65 66 Given such massive exposure, government guarantees in a time of crisis become meaningless. Yet because of the interconnected web of global financial relationships, we are all vulnerable to the threat. -pair two 7 62 At this point in the credit crisis, at least one thing is certain In the United Kingdom, for example, the collected assets of the major banks are four times the nation's gross domestic product (GDP). Table 5-7: Clauses of the three deep red regions of F1 in Figure A-7 The appearance of region (1) as a cluster of all and only the possessive types is not surprising; however, it seems a bit unexpected that the intensive and circumstantial are grouped together as region (2). But at least there is something understandable about the grouping of region (2) – the six intensive ones are the bunch that is most tightly knit together in (2) (see Figure A-7), while the more peripheral “tail” at the bottom within (2) consists of the circumstantial types. Also slightly unexpected is that not all the intensive types are in the same cluster, but split between (2) and (3). Looking at the individual intensive type clauses in (2) and (3) more closely, we see why this is the case, for clauses in (2) have minimal tags – see Figure 5-7 next page, but clauses in (3) have circumstances – see Figure 5-8 next page. 199 Figure 5-7: Transitivity of some clauses in red region (2) in Figure A-7 Figure 5-8: Transitivity of a pair of clauses in red region (3) in Figure A-7 After considering the deep red clause clusters in Figure A-7, let’s move to the deep blue clauses, also arranged in three regions: there’s a massive central cluster of 19 clauses, and two smaller clusters next to it, Figure 5-9 next page highlights these regions: 200 Figure 5-9: Deep blue regions of F1 in Figure A-7 (1) Material process - diverse variety: with Range/with Goal/without Range or Goal, with/without circ (19) – largest cluster 9 11 15 18 25 26 28 31 32 35 40 41 46 49 53 54 74 79 80 Those with some knowledge are driving policy Begin with the U.S. Treasury's $700 billion bailout package. Instead, stock markets collapsed Global investors, now on the sidelines, have declared a buyers' strike against the sophisticated paper assets of securitization that financial institutions use to measure and offload risk. I like to use a salad analogy. Before the last decade, bankers simply lent in the form of syndicated loans. Instead of making simple loans then {the bank} diced {its loans} and {the bank} sliced them up into a big, beautiful tossed salad. Eat that piece of salad, This distrust heightened when global interest rates began to rise. regardless of how much money government spends short of nationalizing the entire financial system. The cost of money is rising and the availability {is} shrinking. to draw that global capital back into more productive uses. And while that is happening, the major governments of the world, including the Chinese, should begin major fiscal efforts to stimulate their weakening economies. 201 (2) Material process with circumstance of extent (4) – cluster on the left 29 38 51 55 and holding them until maturity, but suddenly global investors were no longer ordering salad. Most banks are leveraged by more than 10 to 1. True, the banks will still lend (3) Material process with circumstance of manner (6) – deepest blue cluster 13 14 19 30 43 78 which, if gulped down, would quickly restore financial stability. The "shock and awe" of the sheer size of the taxpayer-funded bailout would somehow restore confidence. In recent years, our banks, borrowing to maximize the leverage of their assets at unheard-of levels, produced mountains of financial paper instruments (called asset-backed securities) with little means of measuring their value a bank collected all its loans together, and why have financial markets collapsed so brutally? The central banks, working with the private institutions in providing enhanced data, need to begin to refashion the world's financial architecture. Table 5-8: Clauses of the three deep blue regions of F1 in Figure A-7 The large diversity of material clauses reflects the immense global scope of the financial crisis and the variety of entities and events that need to be portrayed. Apart from the clusters in Figure A-7 tinted deep red and deep blue, there are the clusters with faint tints to consider. They are highlighted as four groups in Figure 5-10 next page. The first group is one faint blue cluster in the top right corner, consisting of the following clauses (all mental processes): Mental processes (5) – top right cluster 23 Aff 39 Cogn 59 Perc 68 Cogn 77 Cogn As a result, we are experiencing the painful downward reappraisal of the value of virtually every asset in the world. No one knew the location of the toxic waste. Apart from the economic pain resulting from shrinking credit markets, we are about to see an earthquake in the relationship between government and financial markets. Our policy leaders in Washington are thinking domestically The bankers don't trust each other. Table 5-9: Clauses of the faint blue cluster in Figure A-7 202 Figure 5-10: Faint regions of F1 in Figure A-7 The reason why the mental clauses encircled in Figure 5-10 are tinted faint blue (like material clauses) is not immediately apparent, because they don’t share any tags with material clauses, except for the generic “Cl”, “Cl/T1” and “Cl/T1/Proc”, and maybe one or two more, like the manner circumstance “domestically” of clause 68. However the tint is uniform throughout the cluster, not more intense for clause 68. Most probably, F1 does include some tags associated with mental processes, but the extent of this is so slight that the tints on these tags do not show up on the Tag Wheel Plot of Figure A-4. It is not surprising that F1 includes the mental process tags to some degree – the mental clauses are the next most frequently occurring transitivity processes, after material and relational (as was shown in Table 5-1, p.181). 203 The only clauses left to discuss in Figure A-7 (Volume II, p. 397) are those with the very faintest tints. Let’s look at the very faint red clauses in the bottom right region of Figure A-7 in Table 5-10: (1) Relational identifying - encoding (Token=Idr) (10) – bottom right cluster 20 Circ 33 Int 34 Int 56 58 Int Int 60 Int 73 75 Int Int 17 22 Incredibly, these paper instruments were insured by more dubious paper instruments. The idea was to sell (for huge fees) individual servings of diversified financial salad around the world. The only problem: under an occasional piece of lettuce was a speck of toxic waste in the form of a defaulting subprime mortgage. -- but the fear is they will do it only to people such as Warren Buffett, What is uncertain is the amount of lending to borrowers engaged in entrepreneurial risk, the center of business reinvention and job creation. The great uncertainty is whether government has the power to rescue the financial system in times of crisis. The challenge will be to reform our financial system quickly The first step should be efforts to make the market for future asset-backed paper more transparent and credible. (clauses plotted slightly further apart from the main bunch) Circ This is because the credit crisis reflects something more fundamental than a serious problem of mortgage defaults. Circ {which was} followed by a de-leveraging of the entire global financial system. (2) Relational identifying - decoding (Token=Idd) (5) – centre right cluster 12 24 Int Int 42 52 Int Int 64 Int This was presented as some magic pill So what are these paper instruments, these asset-backed or mortgagebacked securities? So what does this salad boycott mean for the future Translation: The U.S. financial system will have a whopping $15 trillion to $20 trillion less credit available next year than was around a year and a half before. This means government cannot bail out the system even if it wanted to. Table 5-10: Clauses of the two faint red clusters in Figure A-7 Table 5-10 shows that all the very faint red clusters in Figure A-7 are all the 15 relational identifying processes in the text. The reason why they are coloured at all is probably because they have the “Reln” tag, hence they share the same colour (red) as the relational attributive clauses. Not only does the neighbourhood plot group the 15 together nicely, but SVD also colours them similarly. Note how well the 204 neighbours plot of Figure A-7 arranges the circumstantial types together (clause 17, 20, 22). The relational identifying clauses are separated into two clusters - “encoding” and “decoding” - these two types were discussed in section 3.3 Transitivity, p. 62. Last of all, we’ll examine the outlier clauses of Figure A-7 in Table 5-11. These clauses are tinted very faintly blue, but the colour is barely discernible. They are found in Figure A-7 beside the relational identifying clauses at the bottom right. Note that some of these clauses are actually part of other larger, more deeply tinted clusters, but are outliers within those clusters, e.g. clause 27, 44. Faint Pair – near bottom right 10 63 Behavl Exist looking through the rearview mirror. A similar situation exists in many Euro zone countries. Faint Tail-end Clauses – near bottom right 27 Ment/Cogn 44 Verbl But with the huge expansion of the global economy in the 1990s, which produced an ocean of new capital, the bankers came up with an idea called securitization. The markets are telling us Table 5-11: The outliers: very faint blue clauses in Figure A-7 These outlier clauses are oddities of transitivity of the text; there’s only one existential, one behavioural and one verbal process in the text. It is a bit of a surprise why the mental clause 27 in Table 5-11 is not in the same cluster as the other five mental clauses of the text (highlighted as a cluster in Figure 5-10, p.203), but in the “tail” of a cluster of mostly relational attributives – of deep red cluster (2) in Figure 5-6, p.198. Similar questions arise as to why the other “tail clauses” 45, 69, 71, 72 (next to clause 27) choose to distance themselves from the main cluster, like clause 27. It turns out that many of these “tail clauses” have 205 metaphors of transitivity: clause 27, 45, 71, 72: Figure 5-11: Clauses with metaphors of transitivity in the “tail” of red cluster (2) in Figure 5-6 The fact that clause 27 shares some “T1m” tags (associated with material processes) with clause 45 explains why clause 27 is in the same cluster as clause 45, though the congruent process “T1” of clause 27 is different from the rest of the 206 cluster. However, not all clauses with “T1m” tags are together in that cluster; some are in other clusters: clause 12, 18, 47. Conclusion on F1 SVD empowers SF analysis by mathematical simplification. The utility of SVD is most evident in a F1 tinted Clause Neighbours Plot, Figure A-7 (Volume II, p. 397). The Feature F1 makes use of characteristics of the text that traditional SF analysis would likely reveal too. However, the utility of F1 in a neighbourhood plot is in achieving an immediate visualisation of the overall associations and oppositions in the text. Traditional SF analysis breaks a text into clauses, but a neighbourhood plot composes the clauses as clusters, which effectively conceives text as being composed of clusters. Clusters provide an intermediate scale for characterising the text, as opposed to describing effects at the clause level or the discourse level. Because clustering and SVD are performed based on functionality, SVD tints on clusters have a tendency to be highly consistent within clusters. Instead of the discrete parameterization of a clause in terms of absolute presence and absence of tags, Feature tinting is continuous, so clusters are tinted to different degrees, differentiating them, as well as relating them to one another. F1 is the best Feature to demonstrate this because it is the strongest Feature that includes both positive and negative original tag vectors, and reveals the most diversity in a neighbourhood plot. As we continue from here to explore further Features, the utility of SVD will be shown not only in re-representing information in different ways, but to reveal patterns that traditional analysis cannot. 207 5.1.3 Feature F2 F2, shown in Figure A-8 (Volume II, p. 398), represents the polarity between attributive and identifying relational processes – this is as we expect, given that relational identifying (15 of them) is the next most frequent process after material (29) and relational attributive (21) (see Table 5-2, p. 182). The darkest blue: Reln/Iden, Idr/Token, Idd/Value The darkest red: Reln/Attr, Carr, Att so interesting to note that “Mat”, “Actor” are moderately moderatel intense red. It is also This places them in F2, together with the relational attributive tags, which they were opposed to in F1. There are also some some very faintly tinted red tags e.g. “Locn”. However, owever, there aren’t any other blue tints apart from the very intense ones. This suggests that F2 is dedicated ded to measuring the distance of the relational identifying processes from the rest of the text. Moving on to Figure A-9 (Volume II, p. 399),, we see how the tints for F2 appear in context. In short, Figure A-9 shows the 15 relational identifying processes proces standing out against the rest of the text, which form a red background. Another interesting pattern this visualisation shows is that the relational identifying processes have a tendency to occur in pairs or triplets close together: Pair One 208 Pair Two Triplet Figure 5-12:: Doublets and triplets t of blue relational identifying clauses in Figure A-9 The effect of close repetition within each doublet/triplet creates is textual; it mutually reinforces the impact of the message in each clause, delivering ideas to the reader in a manner that appears more factual. Another common feature of transitivity amongst the seven relational identifying clauses in Figure 5-12 is that they are all “encoding” encoding” types, where the Identifier is also the Token.. In fact, as discussed in Table 3-23 on p. 62 (section 3.3: Transitivity), these seven clauses constitute all the relational identifying intensive clauses of the encoding type. The layout of these encoding clauses in the text as doublets/triplets is strategic - their semantic tightness (explained in section 3.3: 3.3 Transitivity, p. 65) make the text seem impervious to questioning. To continue ntinue the discussion on these seven special clauses from the section Transitivity p. 62, these hese seven examples have some common thematic and mood structures: a concise Theme against a long Rheme (making the Theme even more prominent); an unmarked Theme (for greater focus, presenting the facts more directly); Value as Subject/Mood Subject and Token as Complement/Residue esidue (placing ( the Tokens,, which are more familiar, in the Residue and hence making them less directly 209 arguable). Also, the Themes tend to be grammatical metaphors: “challenge”, “step” and “fear” are experiential shifts from Process to Entity, “uncertain” and “uncertainty” are interpersonal shifts from Modality, and “idea” and “problem” are logical shifts from Relator to Entity (see section 3.6.2: Logical, Interpersonal and Textual Shifts, p. 116). Having grammatical metaphors as Theme allows the congruent Entities involved: “financial system”, “market”, “people, “government” to be hidden in the Rheme. The result is a more abstract construction of reality. Moving to Figure A-10 (Volume II, p. 400), the Tag Neighbourhood Plot tinted for F2, the blue tints appear most deeply in one cluster (“Reln/Iden/Int”), and faintly in another (“Reln/Iden/Circ”). The red tints, not as intense as the blue, appear mostly in one cluster (“Reln/Attr”), and a little in another cluster (“Mat”). Note that an anomaly appears in the central cluster of mostly red tags: a tag tinted blue, “Cl/T1/Proc/Reln”. According to Table 5-2 there are 36 of these tags, 21 of which are “Attr”, and 15 are “Iden”. If F2 is designed to single out the “Reln/Iden” clauses, then this must mean the “Reln/Iden” tags are tinted more strongly blue than the “Reln/Attr” tags are tinted red, in order to result in a blue tint for “Reln” despite their lesser numbers. Proceeding to Figure A-11 (Volume II, p. 401), the Clause Neighbours Plot for F2, we see what we exactly expect: tight clusters of deep blue clauses against a sea of faint red. This is what F2 is doing, singling out the “Reln/Iden” tags, against the backdrop of the rest of the text. The two blue clusters are tinted to the same extent, reflecting their equal importance to the Feature. These two clusters have already been discussed in Table 5-10, p. 204. Note that unlike the Clause Neighbours Plot for F1 210 (Figure A-7, p. 397), in the same type of neighbours plot for F2 here, there are some clauses that are completely untinted: 23, 39, 59, 68, 77 (mental clauses), 10 (behavioural), 44 (verbal), 63 (existential). 5.1.4 Feature F3 F3, shown in Figure A-12 (Volume II, p. 402), represents mainly the polarity between the mental processes and material processes – this is again not too surprising, given that mental clauses (six of them) are the next most frequent after relational (36) and material (29), (see Table 5-1, p.181), and form a cluster of their own in the Clause Neighbourhood Plot that has not been intensely tinted in any earlier stronger polar Features F1, F2 (a polar Feature is a Feature with both red and blue tints). The darkest blue: Proc/Ment/Cogn, Sens, Phen/Range, Locn/Time (T1m): Proc/Mat, Locn/Time, Range The darkest red: Mat, Actor, Reln/Attr/Int A strange colour appears: “Reln/Attr/Int”, “Carr”, “Att” and “Goal” are faintly red tinted, but it looks like they are faintly blue tinted as well. In the tag neighbours plot Figure A-14 (Volume II, p. 404), however, they are unambiguously red tinted. A fine detail to note is that not all “Reln/Attr” is red, because “Reln/Attr/Circ” is blue, as shown by zooming in to the tag list on the right in Figure A-12: 211 Figure 5-13:: Tints for Reln/Attr tags in tag list in Figure A-12 Moving on to Figure A-13 A (Volume II, p. 403), the text ext tints for F3 bring out the six mental clauses well – they are the clauses that contain the darkest blue tints. Given that “T1m” material process tags appear in Figure A-12 as prominently blue as mental process tags, one expects that such clauses (Table ( 5-12)) show up just as blue as mental clauses in Figure A-13, A , but this is not the case (except for clause 27, which is a mental clause itself too). 12 This was presented as some magic pill 27 But with the huge expansion of the global economy in the 1990s, which produced an ocean of new capital, the bankers came up with an idea called securitization. 45 the world will face a serious credit crunch in 2009 47 to remove the toxic salad from bank balance sheets. Table 5-12:: Clauses with metaphorical material processes in the text (T1m/Proc/Mat) However, Figure A-13 shows a lot of other clauses are partially blue tinted as well. The circumstances of location/time are moderately blue, and attribute/circumstance are faint blue. But this makes the interpretation retation of F3 more confusing because these blue tags are not always associated with blue process tags, i.e. red and blue can occur in the same clause. Examples, excerpt from Figure A-13: Figure 5-14:: Examples of mixed red & blue tinted clauses in Figure A-13 Nevertheless, from Figure A-13 it is clear that the tags “Locn/Time Locn/Time” and “Att/Circ” do play a big role in the definition of F3; F3 is not just defined by mental 212 processes. Let’s then examine the frequencies of these tags, and also tags closely associated with them: 21 4 7 13 5 8 5 3 2 Cl/T1/Att Cl/T1/Att/Circ Cl/T1/Att/Possd Cl/T1/Locn Cl/T1/Locn/Place Cl/T1/Locn/Time Cl/T1m/Locn Cl/T1m/Locn/Place Cl/T1m/Locn/Time Table 5-13: Frequencies of tags not related to mental processes but significant in F3 (in bold), with other associated tags not significant in F3 (not in bold) It appears that the absolute frequency of “Locn/Time” and “Att/Circ” are not high, neither is it particularly high relative to related tags like “Att/Possd” and “Locn/Place”. Therefore it is still unclear why SVD process has picked them out as significant parts of F3. Let’s examine the clauses in which these tags appear: 45 the world will face a serious credit crunch in 2009 69 when the solution [to the credit crisis] will be global. 71 ; it is that the world's money is sitting on the sidelines 72 -- more than $6 trillion *are sitting* in idle global money markets alone. Table 5-14: All Attribute/Circumstance in the text (bold) 7 19 26 27 42 45 50 78 At this point [in the credit crisis], at least one thing is certain In recent years, our banks, , produced mountains [of financial paper instruments ([[called asset -backed securities]])] with little means [[of measuring their value]]. Before the last decade, bankers simply lent in the form of syndicated loans. But with the huge expansion [of the global economy] in the 1990s, , the bankers came up with an idea [[called securitization]]. So what does this salad boycott mean for the future the world will face a serious credit crunch in 2009 After all, the U.S. banks alone so far have lost upwards of $2 trillion from their collective asset base. The central banks, , need to begin to refashion the world's financial architecture. Table 5-15: All Location/Time in the text (bold) 213 A closer look at the clauses with the tags in Table 5-14 and Table 5-15 still does not reveal any obvious reason why they should be tinted blue. “Att/Circ” and “Locn/Time” do not co-occur strongly with blue mental processes, nor do they correlate with the absence of processes tinted red (in fact, many processes in Table 5-15 are “red” in F3). I shall not pursue this matter further now, this could be a topic for further future inquiry. There is one point, though, SVD is a mathematical process that is “blind” to SF theory, (in the sense that it relies on tag distributions, and knows nothing about the semantic relations between tags, aside from their hierarchy), so many Features could have properties that are there because they are the result of mathematical organisation of the grammar tags, and do not have meaningful interpretation in SF theoretical terms. A final clue to the mystery is that “Locn/Time” is the most frequent end-tag amongst all the circumstances of transitivity, as Table 5-16 below shows. This could be the reason why F3 picks it up. But it is still a mystery why “Locn/Time” is part of the “blue pole”, instead of the “red pole”, because “Locn/Time” usually appears in material and relational attributive intensive processes. 3 2 1 2 2 2 2 1 5 1 4 13 5 Cl/T1/Accom Cl/T1/Accom/Add Cl/T1/Accom/Comt Cl/T1/Caus Cl/T1/Caus/Reas Cl/T1/Contg Cl/T1/Contg/Cond Cl/T1/Ex Cl/T1/Ext Cl/T1/Ext/Spat Cl/T1/Ext/Temp Cl/T1/Locn Cl/T1/Locn/Place 8 8 3 5 3 1 2 5 3 2 1 1 Cl/T1/Locn/Time Cl/T1/Mann Cl/T1/Mann/Means Cl/T1/Mann/Qual Cl/T1/Role Cl/T1/Role/Guise Cl/T1/Role/Prod Cl/T1m/Locn Cl/T1m/Locn/Place Cl/T1m/Locn/Time Cl/T1m/Role Cl/T1m/Role/Guise Table 5-16: All circumstances of transitivity, T1 or T1m – 25 tags 214 Let’s proceed to Figure A-14 (Volume II, p. 404), the Tag Neighbourhood Plot for F3. Here the tints are clearer and this makes it probably a better plot to define F3 than Figure A-12. Note that unlike F1 and F2, in F3 the tints for a single colour are no longer dominated by a single cluster. Now we move on to Figure A-15 (Volume II, p. 405), the Clause Neighbourhood Plot for F3. It is notably different from the earlier Neighbourhood Plots, because here, almost all the clauses are untinted – this makes it relatively uninteresting. The only tints are blue tints: the mental clauses: the cluster on the top right: 23, 39, 59, 68, 77, as well as 27, in another cluster. Also tinted blue are all the metaphorical material clauses (except 47): 27, 45, and (very faintly) 12. Clause 27 is the deepest blue because it is both mental and metaphorical material. There are only two more clauses with discernible blue tints: clause 71 and 72; these are relational attributive circumstantial clauses. It is not known why no red tints are showing at all in Figure A-15; perhaps the individual contribution from each clause is insignificant. Upon very careful scrutiny, one observes that the rest of the non-blue clauses are not completely untinted, but are extremely faint red. 5.1.5 Feature F4 F4, defined in Figure A-16 and Figure A-18 (Volume II, p. 406 & 408), represents mainly the polarity between the mental processes and metaphorical processes – however there are a lot of secondary tags dragged into the “blue pole” and 215 “red pole” as well, so each pole becomes harder to interpret. The darkest blue: T1m, T1m/Proc, T1m/Locn The second darkest blue: (T1m) Place, Behl, Behav, Mat (T1) Mat, Actor, Att/Circ, Proc/Attr/Circ The darkest red: Ment, Sens, Phen/Range The second darkest red: Ment/Cogn, Reln/Att/Int (For succinctness, the above tag addresses are approximate. e.g. I sometimes use “A/B” to mean two tags: A and A/B) At first glance it is difficult to distinguish F4 from F3. Both involve mental process tags as one of their poles. In fact, comparing their Tag Neighbourhood Plots Figure A-14 and Figure A-18 demonstrates just how subtle their differences are. The most striking similarity is the polarity between the mental and (congruent) material processes. There are two most striking differences: in F3 the mental tags were associated with the metaphorical tags, but in F4 they are in opposition, and secondly, in F3 mental tags were associated with “Att/Circ” and opposed to “Attr/Int”, but in F4 the situation is reversed: they are opposed to the “Att/Circ” and associated with “Attr/Int”. That sounds as if F4 is mostly just F3 inverse. It does make one wonder if F3 and F4 are truly orthogonal. But SVD is supposed to produce orthogonal Features, i.e. that are semantically mutually exclusive. We proceed to the other SVD plots of F4 – 216 perhaps the text tints and clause neighbours tints will tell us more about F4. Figure A-17 (Volume II, p. 407) is the Text Plot for F4. The six mental clauses that were prominent in F3 as blue, now appear as prominent red. However, there are other prominent clauses in Figure A-17 – about six prominent blue clauses: 12, 18, 45, 47, 71, 72, and it looks like these make up six of the seven total metaphorical transitivity processes in the text. The seventh metaphorical clause is actually clause 27, which also happens to be a mental clause, and should be deep red. Therefore, clause 27 is not prominent, probably as a result of being both deep red and deep blue: the colouring cancels each other out. Also, unlike the F3 Text Plot where the backdrop was virtually completely faint red, the backdrop of the Text Plot for F4 is more diverse – a mix of faint red and faint blue. We move on to Figure A-19 (Volume II, p. 409), the Clause Neighbours Plot for F4. This is slightly more interesting than the same plot for F3, but again, most of the clauses are untinted. Five of the six mental clauses are deepest red, followed by most of the relational attributive intensive in faint red. The deepest blue clauses are 12, 18, 45, 47, 71, 72 - six of the seven clauses in the text with “T1m” tags. Conclusion to SVD of main story The short experiment of SVD on the main story has concluded. It is time now to review the usefulness of the visualisations by SVD and the overall effectiveness of SVD as a tool for SF research. For defining each Feature, I found the Tag Neighbourhood Plots just as necessary as the Tag Wheel Plots. For interpreting the 217 strengths of Features in the clauses, I found the Clause Neighbourhood Plots most useful, but the Text Plots are sometimes necessary too. The most immediately realisable use of SVD is as an organising tool for a SF analysis database. The power of SVD lies not just in spotting correlations of the presence of a meaning-making resource with the presence of another resource, but also unexpected correlations of the presence of a resource and the absence of a resource. The challenge in using SVD is the interpretation of the SVD Features. I found the tints for F1 and F2 the most useful, as they managed to pick up strong linguistically relevant patterns in the text. F1 and F2 visualisations are useful even for clauses which did not exhibit the Features strongly, because the non-discrete nature of SVD allowed almost every clause to be defined, as having a Feature to some degree, in relation to the rest. The continuous variation of SVD tints allowed for more detailed visualisations, such as in the classification of transitivity analyses with Neighbourhood Plots, and the logogenetic changes in Feature strengths as the text unfolds in Text Plots. This “fuzzy” representation of linguistic semantics could be more accurate than discrete tag annotation (in traditional SF analysis) because fuzziness is the nature of natural language and meaning, as pointed out by Halliday (1995). The weaker Features like F3, F4 get progressively harder to interpret, due to confounding factors. The Features weaker than F4, not examined in this study, could still be interesting because they are unlikely to be noise: the Smick text is a carefully sculpted text and is unlikely to contain random meaning. These represent options for further research. 218 5.2 EXPERIMENT 2: COMMENTS Following our investigation of transitivity in the main story, we now perform a SVD analysis for the comments. All “T1” tags in the comments are extracted with a “Cl/T1/*” search command in the Search Page in Systemics (note that this command differs from that for Experiment 1. Here, “T1m” tags are not extracted. This is because, as discussed on p. 206, “T1m” tags tended to mis-cluster clauses) SVD is performed on the resulting reduced matrix. As a note before we begin to examine the results, SVD is designed to describe and characterise the diversity of meanings. Since the comments are made by different people, SVD seems potentially ideal for analysing comments. Please refer to Volume II, Appendix 6.2: SVD Plots - Comments, p. 410-412 for the Tag Wheel plots defining Features F1, F2, F3. The plots show that the strongest few Features in the comments are essentially the same as for the main story. The same following polarities are captured by the SVD of comments: F1: Mat & Reln -Figure A-20 (p. 410) F2: Attr & Iden -Figure A-21 (p. 411) F3: Ment & others -Figure A-22 (p. 412) This is an amazing result, given that the comments can be made by anyone, yet the commenters choose to use the same transitivity patterns as Smick. Smick’s characteristic inverted Token-Value constructions also appear in the comments. It seems that Smick’s attempts to be in control and appear authoritative have succeeded: 219 his commenters are following his styles of construing the crisis. E.g. gmakens says: Figure 5-15: gmakens using Smick’s style Clause 67 (comments) in Figure 5-15 is another one of those encoding clauses with an inverted Token-Value configuration. Conclusion to SVD of comments While the previous chapter on recurrence explored the dynamics of meaning between clauses, SVD has not only shown the dynamics of clauses, but has also unveiled dynamics between two discourses: the main story and the comments. While comparative studies of functional meaning are not new, this is certainly the first time that a mathematical way has been demonstrated to compare the main features of one discourse from another, in terms of systemic functional meaning. These initial results are very promising and these methods have the potential for wider application. 220 _____________________________________________________________________ Chapter 6: CONCLUSION _____________________________________________________________________ In this thesis, we have completed a comprehensive systemic functional analysis of one text, and demonstrated how the resources of logic, theme, mood, transitivity, ergativity, grammatical metaphor combine to make the text, construe the financial crisis, and enact the relationship between the author, the readers, and the participants in the story. The clarity of the logical complexes, the abstract thematic progressions, and the abundance of grammatical metaphors give rise to a complex but well-orchestrated text. The transitivity and ergativity construe the financial crisis as an event of immense proportions with severe impact on the people, and the responsibility was placed on the banks and the government. Theme, mood and transitivity analyses reveal that the author has presented himself as an unquestionable authority figure to the readers, with greater knowledge than the banks or the government. We have also introduced methods of linear algebra, in the form of vector and matrix operations, as a means to visualise patterns of similarities and differences in meaning in discourse across metafunctions and multiple scales. The results of these mathematical excursions are only preliminary, but are confirmations of the applicability of the tools of Dynamical Systems Theory to Systemic Functional Linguistics, and suggestive of possible further work. While this study was focused on textual discourse, the techniques that were used are generalisable to multimodal discourse, employing the system network as the basis of meaning. 221 The effectiveness of the theories and techniques used in this research was best demonstrated in the analysis of one particular section of the CNN main story. The “salad analogy” of the text was not only marked in terms of the lexis (“salad”), but marked in terms of interclausal and clausal grammar: in the analyses of logical complexing, thematic progression, mood, transitivity and grammatical metaphor. This markedness in meaning against the entire discourse and across the metafunctions was captured by the combined metafunctions recurrence plot as a Dark Band, and by the SVD as a continuous stretch of similar F1 values in the Text Plot. These results show the interplay between lexis and grammar creates an impact on the landscape of functional meaning at the discourse level, and the extent of this impact can be measured and visualised. The perspective is that discourse is a multidimensional system of meaning, which can be evaluated with mathematical formalisms for deeper insights. The linguistics world is polarised between formal and functional schools of thought. Mathematics is a language more suited to the description of physical form rather than social function. However, Systemic Functional linguistics encodes functionality in a formal way with system networks, and the Systemics software simplifies the encodings into concise tags. This transforms functional meaning into a form that is amenable to mathematical parameterization, thus bringing some of the best aspects of both schools of thought together – the objectivity of logical analysis used by formal linguistics and the social-cultural relevancy that is preserved in functional linguistic research. Meaning in language is not static but dynamic, and the theme of dynamics in 222 this thesis is most prominent in our discussion of thematic progressions, grammatical metaphor shifts, and recurrence. The existence of meaning on multiple scales has been most evident in our analysis of logico-complexes and recurrence. Meaning is also by nature intangible. The objective of using computerised and mathematical techniques in this thesis was not to measure and define the meanings in discourse in a delimiting manner, but to empower existing theories of meaning and open new perspectives to the way we view and understand discourse. We live in a large and complex world. Things happen all the time around us too quickly, many times we never fully know what happened, and why. Unless we look back and think. Only then, will we have a chance to fully understand ourselves, and the world in which we live. 223 _____________________________________________________________________ REFERENCES _____________________________________________________________________ Berry, M. W., & Browne, M. (2005). Understanding search engines: mathematical modeling and text retrieval (2nd ed.). Philadelphia: Society for Industrial and Applied Mathematics. Butt, D., & O'Toole, M. (2003). Transactions between matter and meaning : a functional theory for the science of text. Paper presented at the second International Conference Studies for the Integrated Text Science, Nagoya, Japan. Collins, P. C. (1991). Cleft and pseudo-cleft constructions in English. London: Routledge. Daneš, F. (1974). Functional sentence perspective and the organization of the text. In F. Daneš (Ed.), Papers on functional sentence perspective (pp. 106-128): The Hague: Mouton Eckmann, J. P., Kamphorst, S. O., & Ruelle, D. (1987). Recurrence plots of dynamical systems. Europhysics Letters, 4(9), 973-977. Firth, J. R. (1948). Sounds and prosodies. Transactions of the Philological Society, 47(1), 127-152. Francis, G. (1986). Anaphoric nouns. In M. Coulthard, M. Hoey & M. Knowles (Eds.), Discourse Analysis Monograph 11: English Language Research, University of Birmingham. Fries, P. H. (1981). On the status of theme in English: arguments from discourse. Forum Linguisticum, 6(1), 1-38. Fries, P. H. (1995). A personal view of Theme. In M. Ghadessey (Ed.), Thematic development in English texts (pp. 1-19). Pinter: London. Halliday, M. A. K. (1967). Notes on transitivity and theme in English, Part 2. Journal of Linguistics, 3(2), 199-244. Halliday, M. A. K. (1968). Notes on transitivity and theme in English, Part 3. Journal of Linguistics, 4(2), 179-215. Halliday, M. A. K. (1978). Language as social semiotic: the social interpretation of language and meaning. London: Edward Arnold. Halliday, M. A. K. (1993). Towards a language-based theory of learning. Linguistics and Education, 5(2), 93-116. 224 Halliday, M. A. K. (1994). An introduction to functional grammar (2nd ed.). London: E. Arnold. Halliday, M. A. K. (1995). Fuzzy grammatics: a systemic functional approach to fuzziness in natural language. In J. J. Webster (Ed.), Computational and Quantitative Studies. London: Continuum. Halliday, M. A. K. (1998a). On the grammar of pain. Functions of Language, 5(1), 132. Halliday, M. A. K. (1998b). Things and relations: regrammaticising experience as technical knowledge. In J. R. Martin & R. Veel (Eds.), Reading science: critical and functional perspectives on discourses of science (pp. 185-235). London: Routledge. Halliday, M. A. K., & Matthiessen, C. M. I. M. (2004). An introduction to functional grammar (3rd ed.). London: Arnold. Harvey, A. (2001). Relational clauses in English technical discourse: patterns of verb choice. Pragmatics, 11(4), 379-400. Judd, K. (2009a). Visualization experiments 1 (unpublished report), 23 Mar. Judd, K. (2009b). Visualization experiments 2 (unpublished report), 17 Aug. Judd, K. (2009c). Visualization experiments 3 (unpublished report), 15 Sep. Landauer, T. K. (2007). LSA as a theory of meaning. In T. K. Landauer, D. S. McNamara, S. Dennis & W. Kintsch (Eds.), Handbook of latent semantic analysis. Mahwah, New Jersey: Lawrence Earlbaum Associates, Inc. Larsen-Freeman, D., & Cameron, L. (2008). Complex systems and applied linguistics. Oxford: Oxford University Press. Lemke, J. L. (2000). Opening up closure: semiotics across scales. Annals of the New York Academy of Sciences 901, 100-111. Martin, J. R. (1992). English text: system and structure. Philadelphia, PA: John Benjamins Matthiessen, C. M. I. M. (1992). Interpreting the textual metafunction. In M. Davies & L. Ravelli (Eds.), Advances in systemic linguistics: recent theory and practice. (pp. 37-81). London: Pinter. O'Halloran, K. L. (2003). Systemics 1.0: software for research and teaching systemic funcitonal linguistics. RELC Journal, 34.2, 157-178. Pike, K. L. (1982). Linguistic concepts: an introduction to tagmemics. Lincoln: University of Nebraska Press. 225 Press, W., Teukolsky, S., Vetterling, W., & Flannery, B. (1994). Numerical recipes in C: the art of scientific computing (2nd ed.). Cambridge, New York: Cambridge University Press. Trefethen, L. N., & Bau, D. (1997). Numerical linear algebra. Philadelphia: Society for Industrial and Applied Mathematics. Winter, E. O. (1977). A clause-relational approach to English texts: a study of some predictive lexical items in written discourse. Instructional Science, 6(1), 1-92. OTHER REFERENCES Statistics of the financial crisis in section 2.1 The CNN Story were taken from the local newspapers in Singapore, The Straits Times (in the month of October 2008). 226 [...]... is the aim of the Socio Cultural Modelling Project led by Kay O’Halloran and Kevin Judd at the Multimodal Analysis Lab, Interactive Digital Media Institute (IDMI), and this thesis represents the beginnings of the project The study of meanings in language1 is a complex problem that challenges theoreticians on all fronts To begin with, we can think of meaning in text as arising from meaning in words This... Tables in the Clause Page can be printed in Systemics The complete set of printouts representing the comprehensive SF analysis of the main story and the comments are compiled in Appendix 2: SF Analysis of Clauses, p 234 The SF analysis of the main story is also presented in the form of matrices in Appendix 3: SF Analysis Matrices, p 368 Printouts from other Pages of Systemics (Interclausal, Discourse) ... biggest investment bank in US) sold itself and US insurer AIG came to the brink of collapse, triggering shockwaves worldwide These events in the US impacted on financial markets in Europe, Asia and all corners of the globe, and plunged economies into recession In the week before 10 October, global markets lost U$6 trillion in panic selling, and governments injected U$6 trillion into the financial system in. .. confined in sentences; it is an ongoing dynamic process of orchestration of functional relations between words, sentences, discourse and social context, see Figure 1-1: Figure 1-1: Meaning in text as arising from a web of functional relations (examples of each relation in italics) There is often the philosophical question of whether linguistics is like a snake trying to swallow itself: language was... clauses in red region (2) in Figure A-7 200 Figure 5-8: Transitivity of a pair of clauses in red region (3) in Figure A-7 200 Figure 5-9: Deep blue regions of F1 in Figure A-7 201 Figure 5-10: Faint regions of F1 in Figure A-7 203 Figure 5-11: Clauses with metaphors of transitivity in the “tail” of red cluster (2) in Figure 5-6 206 Figure 5-12: Doublets and triplets of blue... helpful in many situations, it fails to cover the meaning derived from the 1 The term “language” in this thesis refers to natural language (as opposed to “artificial language” such as those created for programming machines), unless stated otherwise 2 patterns of words within sentences and patterns of sentences within a discourse Meaning in text is not confined in words, nor is it confined in sentences;... Figure 2-5: Analysis of metaphors of transitivity in Systemics Having introduced all the Keys in the tables, we now explain the theoretical significance of the tags The tags in the Clause Table and Analysis Table represent, for the instance of the particular clause/word group/word, the choices made by the writer according to system networks stored in the Grammar Page In Systemic Functional Theory, the... links) • Cohesive (non-structural links) Hypotactic links have arrows which point towards the dependent clause Text Colour Key Conjunctive Adjunct (Theme) • Structural Conjunction (Theme) • Interrupting Clause or Interrupting Clause Complex • Rankshifted Clause Complex • The clause complexing between an interrupting clause and its ranking clause, and between clauses within a rankshifted/ interrupting... Visualization of SVD in two dimensions 179 Figure 5-2: Tightest clusters in Figure A-3 187 Figure 5-3: Meaning of F0 and F1 in SVD 190 Figure 5-4 a, b: Contribution of presence and absence of tags to F1 192 Figure 5-5: Stretches of constant F1 values (in the clause list of Figure A-4) 194 Figure 5-6: Deep red regions of F1 in Figure A-7 198 Figure 5-7: Transitivity of some... also been organised giving a total of 103 clauses, in Appendix 1.2 (p 230) 2.2 SYSTEMICS (THE SOFTWARE) Analysis of the text in the CNN article is performed with advanced versions of Systemics, a software for systemic functional analysis developed by Kevin Judd and Kay O’Halloran (O'Halloran 2003) The interface of Systemics is organized into “Pages” which fulfill different aspects of the analysis • Text ... beginnings of the project The study of meanings in language1 is a complex problem that challenges theoreticians on all fronts To begin with, we can think of meaning in text as arising from meaning in. .. programming machines), unless stated otherwise patterns of words within sentences and patterns of sentences within a discourse Meaning in text is not confined in words, nor is it confined in sentences;... is a theory of language and meanings based on the choices we make in producing each instance of language; the function of each utterance as a reflection of our momentary intention in the immediate