Iberica 13 ibérica 44 (2022): 127-154 iSSN: 1139-7241 / e-iSSN: 2340-2784 https://doi.org/10.17398/2340-2784.44.127 Abstract This study analysed the information-argument structure of the Introduction sections of sociology research papers, to identify differences across three types of sociology research: Investigative research, Development and Evaluation research, and Descriptive research. The information-argument analysis framework bears some resemblance to rhetorical structure frameworks following Swales’ carS model, but focuses on the argumentative aspect of text and how information is used to support argument claims. The coding scheme specifies information types, subdivided into those that imply an argument claim and those that play the role of argument support. Seventy papers were sampled from ten sociology journals for analysis. Sequential association rule mining was used to identify sequential information-argument patterns. The study identified significant differences in information-argument profile across the three types of research papers, as well as differences in sequential patterns. Methodology contributions of the paper include the coding scheme for information-argument types in research papers, and the method of analysing sequential patterns. Keywords: academic writing, argument structure, argument structure analysis, argumentation, academic argument. Resumen Patrones de información y argumentos en las secciones de Introducción de los trabajos de investigación en sociología El presente estudio analiza la estructura de la información y argumentos en la introducción de los trabajos de investigación en sociología, con el objetivo de Information and argument patterns in the Introduction sections of sociology research papers Wei-Ning Cheng & Christopher S. G. Khoo Shanghai University (China) & Nanyang Technological University (Singapore) wcheng009@e.ntu.edu.sg & chriskhoo@pmail.ntu.edu.sg 127 Ibérica 44 (2022): 127-154 WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO identificar diferencias entre los siguientes tres tipos de investigación: la investigación “investigadora”, la investigación de desarrollo y evaluación, y la investigación descriptiva. El marco para el análisis de la información y los argumentos guarda cierta semejanza con los marcos de estructura retórica que siguen el modelo de Swales, carS, pero, a diferencia de ellos, se centra en la argumentación del texto y las maneras en que respaldar las afirmaciones. El esquema de codificación implementado diferencia entre dos tipos de información: los que incluyen una afirmación de los argumentos y los que contienen un apoyo. como muestra de análisis se escogieron setenta artículos de diez revistas de sociología. Se han utilizado diferentes reglas de asociación secuencial para poder identificar posibles patrones secuenciales formadas por una estructura de información y argumentos. El estudio ha identificado diferencias significativas en el perfil de la estructura información-argumento entre los tres tipos de trabajos de investigación y entre los patrones secuenciales. Las contribuciones metodológicas del artículo incluyen un esquema de codificación para los tipos de argumentos de información en artículos de investigación y un método de análisis para patrones secuenciales. Palabras clave: escritura académica, estructura del argumento, análisis de estructura argumental, argumentación, argumento académico. 1. Introduction This is part of a broader study of the information and argument structure of research papers, focusing initially on sociology research papers. This paper reports our analysis of the Introduction section of sociology research papers, using our information-argument analysis framework and coding scheme. The framework is derived partly from rhetorical structure analysis frameworks following Swales’s (1990) creating a research Space (carS) model. rhetorical structure analysis is a kind of discourse analysis focusing on the author’s persuasive strategy and tactics (i.e., rhetorical steps or rhetorical functions). However, the rhetorical purpose must be to persuade or convince the reader about something, namely information and arguments. Our perspective is that discourse analysis of academic writing focusing on its argumentative aspect and how information is used in argument claims will yield a deeper understanding of the nature of a research study, that is, how it constructs new knowledge using a network of argument claims, supports and information. Our coding scheme for information-argument types overlaps with the list of rhetorical steps in Swales’s (1990) carS model (listed in Table 1), and 128 subsequent elaborations by researchers in genre studies. Many of the steps in the carS framework lack an obvious persuasive function. for example, research objective is a type of information and statement of research intent, rather than a persuasive strategy. Of the eleven rhetorical steps in the carS framework, five clearly indicate argument claims: Claiming centrality, Counter-claiming, Indicating a gap, Making topic generalisations and Announcing principal findings. The other six rhetorical steps appear to refer to types of information, but may be considered to contain embedded arguments. for example, Reviewing items of previous research may include conclusions synthesised or generalised from previous research papers. Continuing a tradition may include justifications of why that is useful. Thus, the rhetorical steps mix persuasive functions with information types and argument claim types. Our coding scheme (listed in Table 2) starts with a list of information types, but groups them into those that carry or imply an argument claim (group 1), those that typically take the role of argument support (group 2), those that are not argumentative but indicate types of descriptive information (group 3), and those that indicate types of context information (group 4). focusing on the argumentative aspect of a text allows us to analyse the text in terms of support => claim links, bearing in mind that a claim can function as a support to another claim (i.e., claim => claim), and furthermore a support statement can occur in the text after the claim statement (claim <= support). Our analysis identified frequently occurring sequences of argument claims in the text (referred to as sequential argument patterns). We also analysed logical links between argument claims and argument supports that are independent of the sequence in which they appear in the text, thus conflating support => claim and claim <= support sequences We refer to these as logical argument patterns.1 We attempt to show in this paper that by foregrounding the argumentative aspect of discourse and its relation to information types, we gain - by linking information-argument elements into a network structure of discourse. In an earlier study (cheng, 2020), we identified five types of sociology research represented in our corpus. as three of the research types were found in only a few papers and most of these studies employed qualitative methods, we combined them into one Descriptive research category. Thus, we carried out information-argument structure analysis for three types of sociology research papers, to identify discourse differences between them: Investigative research, development and Evaluation research, and descriptive research (defined later in Table 4). INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 129 The objective of this study was therefore to analyse the information- argument structure of the Introduction sections of sociology research papers, to identify differences across three types of sociology research. More specifically, we sought to find out: 1. How their information-argument profiles differ—in terms of the relative proportions of the different argument claim, support and information types. 2. How the sequential and logical argument patterns found in the papers are different. One innovation in our analyses was the use of sequential pattern mining (a data mining technique) to identify frequent sequential patterns of argument claim, support and information types. Table 1. Swales’ cars model for Introduction sections of research papers. WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154130 ' 2 ' Moves Steps Move 1: Establishing a territory Step 1: Claiming centrality Step 2: Making topic generalisations Step 3: Reviewing items of previous research Move 2: Establishing a niche Step 1A: Counter-claiming Step 1B: Indicating a gap Step 1C: Question-raising Step 1D: Continuing a tradition Move 3: Occupying the niche Step 1A: Outlining purposes Step 1B: Announcing present research Step 2: Announcing principal findings Step 3: Indicating the Structure of the Research Article ! ' 2 ' ! Coding scheme Code Group 1. Information types that carry/imply an argument claim ArgC01 Research issue ArgC02 Research gap ArgC03 Research question ArgC04 Research objective ArgC05 Research method ArgC06 Research hypothesis ArgC07 Research result ArgC08 General result ArgC09 Research contribution/recommendation ArgC10 Topic centrality ArgC11 General statement ArgC12 Research idea/approach ArgC13 Address research limitation Table 2. Coding scheme for information-argument analysis framework. INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 131 ' 2 ' ! ArgC14 Develop something new ArgC15 Concept/theory/model–apply ArgC16 Concept/theory/model–compare/relate ArgC17 Concept/theory/model–derive ArgC18 Concept/theory/model–define ArgC19 Concept/theory/model–evaluate/critique ArgC20 Literature–summarise/generalise ArgC21 Literature–derive/infer ArgC22 Literature–highlight claim ArgC23 Literature–concession ArgC24 Literature–counter argument ArgC25 Literature–evaluate/critique ArgC26 Literature–difference to current research ArgC27 Literature–compare/relate ArgC28 Literature–highlight scholar/concept ArgC29 Concession/counter argument ArgC30 Novelty Code Group 2. Information types playing an argument support role ArgS01 Research method ArgS02 Practical problem ArgS03 Data analysis ArgS04 Example ArgS05 Established knowledge ArgS06 Inference chain ArgS07 Important implications ArgS08 Theoretical mechanism ArgS09 Explanation/Elaboration ArgS10 Cited author’s claim/opinion ArgS11 Research motivation/justification ArgS12 Theoretical framework ArgS13 Extension of previous research Code Group 3. Descriptive information types InfD01 Research scope InfD02 Research area InfD03 Outline the structure InfD04 Data context InfD05 Single study description InfD06 Own previous work description InfD07 Literature signpost InfD08 Concept/theory/model-details Code Group 4. Context information types InfC01 Practical background InfC02 Historical background InfC03 Personal background InfC04 Theoretical background InfC05 Topic classification structure ! 2. Analysis framework: Information-argument analysis framework in relation to rhetorical structure frameworks Our coding scheme for information-argument types was developed bottom- up from an analysis of 20 sociology research papers, and confirmed and refined using an additional 50 papers. The focus of the coding was on identifying types of information and types of argument claim and support that appear to be important to social science research, especially in arguing the validity of the research results and that a significant research contribution was made. Thus, Research result and Research contribution are two important information types, which carry the implicit claim that they are valid. “claim” suggests that the propositions in the Research result and Research contribution statements are not self-evidently true; the author has to adduce support for the claims to show that the claims are valid or at least reasonable. Thus, other information types are used as argument supports for the claims. We consider the main information-argument types in research papers to be Research issue, Research gap, Research objective, Research result, and Research contribution. These coding categories indicate information or content types, but they also embody certain implicit claims. The research objective statement implies the following claims: 1. it is well-formed: the statement is a clear research objective, and is researchable 2. it is well-founded: founded on theory, the literature or common knowledge (i.e., prior knowledge) 3. it is well-worth researching: addresses a research gap, research issue or practical problem. These embedded claims in a research objective statement are, we assume, common knowledge of experienced researchers, often explained in research method textbooks (e.g., Munyua, 2021), and taught in phd programs. That the research objective is “well-formed” (i.e., clear, unambiguous and researchable) is supported by offering definitions, explanations, and elaborations. Thus, the following sequence of information-argument elements indicate support that the research objective is well-formed: Explanation/Elaboration => Research objective. WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154132 The following sequences provide support for the research objective’s well- foundedness (i.e., founded on theory, the literature or common knowledge): Theoretical framework => Research objective Literature–summarise/generalise => Research objective Established knowledge => Research objective. The following sequences argue that the research objective is well-worth researching: Topic centrality => Research objective Research gap => Research objective. Examples of these support => claim sequences from our corpus are available from dr-NTu (data)—the data repository of the Nanyang Technological university.2 It is clear from the above support => claim sequences and the examples that the argument support statement clarifies or makes explicit which claim is supported, for example, whether the research objective is well-formed, well-founded or well-worth researching. In other words, the argument support statement does not merely support the claim, but also highlights which claim it supports, if there are multiple implied claims. In fact, for some researchers in the field of argumentation, the term argument refers to the argument support (van Eemeren et al., 2014; Tindale, 1999). The research result statement carries the implied claim that it is valid and, for quantitative research studies, replicable. These claims are often supported by research method and data analysis statements: Research method & Data analysis => Research result. In the coding scheme, Research method is the only category that appears in both group 1 (a claim) and group 2 (a support). This is because research method can be used to support the research result, but may itself need to be defended as appropriate to address the research objective. Most of the other group 1 information-argument types carry the implication that the claim is valid or reasonable. although our coding scheme was developed bottom-up based on analysing sample texts, our prior experience in rhetorical structure coding has INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 133 obviously influenced us to create several categories that are similar to rhetorical steps, for example Topic centrality (sometimes called Centrality/Importance of the topic) and General statement (or Topic generalisation). Indeed, several authors in the field of argumentation have pointed out the interrelatedness of argumentation and rhetoric. Hinton (2019) noted that the “distinction [between persuasion and argument] is difficult to maintain (…) and it is clear that any investigation into how language is used to put across arguments cannot remain aloof from considerations of rhetorical impact” (p. 96). a comparison of our information-argument coding scheme with extant rhetorical structure coding schemes (i.e., Kanoksilapatham, 2005; Kathaplia & Khoo, 2020; Swales,1990; Zhang et al., 2011) can usefully highlight the characteristics of our framework.3 The extra categories in our coding scheme include: Research issue, Research idea, Develop something new, and Novelty. These types of claims are clearly important to researchers. Our coding scheme has more refined categories for theory: Concept/theory/model–apply, Concept/theory/model–compare/relate, and Concept/theory/model–derive. The coding scheme includes the following categories as argument support: Theoretical mechanism, and Theoretical framework. It has an extensive list of categories related to literature review, including: Literature–derive/infer, Literature–highlight claim, Literature–counter argument, and Literature–difference to current research. The detailed coding scheme for information-argument types, and resources to support xml tagging and display of coded text in a Web browser, have been deposited in dr-ntu (data).4 3. Argument structure Information-argument structure analysis starts with tagging text units (usually sentences) with categories in the coding scheme. a text unit that is tagged with an argument claim or argument support type (i.e., group 1 and group 2 categories in Table 2) is referred to as an argument discourse unit, following peldszus and Stede (2013). We then extract sequences of argument claims and supports, such as claim => claim => claim and support => claim => claim. We refer to these as argument chains. The shortest argument chain is the two-step pattern claim => claim or support => claim called an argument step. WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154134 a frequently occurring sequential argument chain is referred to as a sequential argument pattern. We also identified sequential argument patterns where some claims in the argument chain are skipped over (explained later). as mentioned earlier, the support for a claim may appear after the claim statement in a text (i.e., claim <= support). We attempted to analyse support-claim relations independent of text presentation order, thus conflating support => claim and claim <= support. furthermore, the support for a claim may not appear immediately before or after the claim statement, but may occur some distance away. We refer to a support => claim relation independent of presentation order as a logical argument step. This kind of relation is difficult to code as it has to be inferred based on the coder’s understanding. Nevertheless, it is important to try as this can uncover a deeper argument structure that may be obscured by sequential presentation order. Argument structure can refer to the internal logical structure of propositions (called microstructure). Our concern is with what freeman (2011) termed macrostructure, which is concerned with: how its component statements (together perhaps with other elements) fit together as wholes to allegedly lend support to some claim or claims. Which statements are put forward to support which other statements in the course of an argument and how, if at all, are those claims of support qualified?. (p. 1) We use argument structure to refer to a set of argument elements (i.e., claims and supports) in a section of a document, linked together (by logical argument steps) into a network structure. Argumentation refers to the way of developing an argument structure in a research paper. Table 3 lists our definitions of these terms. INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 135 Table 3. Definitions of argument-related terms. 4. Argument structure in linguistics literature There is substantial interest in argument structure analysis in genre studies: a literature search in the Linguistics and Language Behavior abstracts carried out in July 2022 found 555 peer-reviewed publications published from 2010 to 2022 with the query ab (argument* PRE/0 (structure? OR pattern? OR scheme?)). Surprisingly, we have not identified any paper focusing on argument structure of research papers. Limiting the search to research papers (by adding the query phrase ab((academic or research or journal) PRE/0 (paper? or article? or report?))) retrieved only one paper that focused on hedging strategies. Linguistics scholars have proposed different approaches for analysing argument structure. folli and Harley (2013) proposed an analysis method from a syntactic perspective, using a set of grammatical rules to identify relations between arguments; Moretta, feltracco, Jezek, and Magnini (2018) proposed a method from a semantic perspective, using different signifiers such as words or phrases to indicate semantic roles. Some researchers (e.g., Kohen et al., 2011; Sheinfux, 2017) investigated the structure between the syntactic and semantic level. Some studies focused on the argument flow (i.e., sequence of arguments) in student essays without categorising types of arguments (i.e., just identifying the argument claims) (Jalilifar et al., 2017; Lee, 2013; rusfandi, 2015; Liu & furneaux, 2015). WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154136 Term Definition Argument claim The proposition that the author seeks to convince the reader is true, valid or reasonable. Argument support A proposition that supports the argument claim. It may also clarify or make explicit the claim that is supported (out of a set of possible implied claims). Sequential argument step A claim=>claim or support=>claim sequence in presentation order in the text. Sequential argument chain A list of claims and supports in order in which they are presented in the text, not necessarily spanning the whole section of a research paper. Sequential argument pattern A frequently occurring argument step, or sequential argument chain, possibly allowing some claims in the chain to be skipped. Logical argument step A support=>claim or claim=>claim relation independent of presentation order, but deemed conceptually linked. Logical argument pattern A frequently occurring logical argument step. Argument structure The set of all the argument elements (i.e., claims and supports) found in a section of a research paper, linked into a network structure by logical argument steps. The argument structure supports the coherence of the section of the paper. Argumentation A way or method to develop the argument structure in a research paper. These approaches are unsatisfactory as they are not general methods that can be applied to any academic paper. They are akin to case studies of individual papers to understand the author’s arguments, but do not try to identify patterns that can be generalised to other papers. Erduran, Simon and Osborne (2004) and Simon (2008) used Toulmin’s framework to develop indicators of the quantity and quality of argumentation in science classroom discussion, as well as to measure improvement in argumentation after changes in instructional support. Several authors (e.g., Wingate, 2012) have noted that the framework can model only micro-level argument instances, and not macro-level argument structures. Most of the studies of academic arguments are for designing scientific research, and hardly any for scientific research papers and none that we know of for social science research. Specifically, we lack studies investigating types of arguments relating to research objectives, results and contributions. a commonly used argument framework in the field of academic writing and genre studies is Toulmin’s (2003) model of argument, which indicates that the argument claim may have qualifiers and potential rebuttals, and the argument supports include the data or evidence (forming the basis for the claim), the warrant (that authorises the step from the data to making the claim), and backing (that provides support for the warrant). However, in a research paper, some of these elements are not explicitly stated or are implied (van Eemeren et al., 2014). for example, the warrant is often implied prior knowledge in the research community, and is provided only when the research method is new or when the warrant is controversial. We include data, warrant, and backing under support. 5. Types of sociology research Social science research can be divided into different types, the simplest division being between quantitative and qualitative studies. Our pilot study (cheng, 2020) identified five types of sociology research: Investigative research, Development & Evaluation research, Descriptive research, Historical analysis, and Identification research. They are defined in Table 4. These types of research have quite different end goals and we expected the authors to use quite different information and argument structures. Because of the small number of papers belonging to the last three research types (descriptive research, Historical analysis, and Identification research), these were combined into INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 137 one descriptive research category for the analysis, as most of them used qualitative research methods. The different research types reflect different epistemic paradigms, seeking different kinds of knowledge and presenting different kinds of arguments. Investigative research, which investigates a research relation (usually a cause- effect relation) between two concepts or entities, typically employ quantitative research methods. They can be characterised as adopting a positivist paradigm of research, rather than a postmodernist, constructivist, interpretivist or critical theory paradigm (Lincoln et al., 2011; Schwandt, 1998), which is more likely to characterise descriptive research. The research type can be identified solely from the research objective and research method statements, which contain different indicative keywords. for example, as Investigative research seeks to investigate a causal or association relation between two concepts, its indicative keywords include affect, influence, and improve. using these definitions and indicative keywords, even undergraduate coders were able to identify the types of research quite accurately. Intercoder reliability results are provided in the next section. Table 4. Definition of each type of sociology research. 6. Method 6.1. Scope of the study Seventy papers were annotated with the coding scheme: 30 reporting Investigative research, and 20 each for development and Evaluation research and descriptive research. The information-argument profiles for the three types of sociology research were identified, characterised by frequency distributions and percentages for the different types of claims, supports, and information. Significant differences across the three types of sociology WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154138 Research Type Definition Investigative research The research seeks to investigate a causal or associative relation between concepts or entities, often by using quantitative research methods (e.g., questionnaire survey). Development & Evaluation research The research seeks to develop a complex concept (i.e., theory, method or system), or to evaluate such a concept, usually using a quantitative research method. Descriptive research The research seeks to explicate a phenomenon or an event, often using qualitative research methods (e.g., ethnography). Historical analysis The research seeks to explicate a change in a particular social, economic, cultural and institutional phenomenon (including the development of an entity) over a period of time, often based on the analysis of historical data. Identification research The research seeks to identify a subtype or instance of a concept (e.g., pattern), usually using a qualitative research method. research were identified. Sequential association rule mining (a data mining method) was used to identify frequently occurring patterns of claims and supports, considering the presentation order in the text (referred to as sequential argument patterns). frequently occurring support => claim patterns occurring in adjacent sentences, but irrespective of presentation order, were also identified (referred to as logical argument patterns). 6.2. Corpus The corpus for this study comprises research papers from ten journals with the highest impact factor in Incites Journal citation reports for sociology, listed in Table 5. The articles were published in the late 2015 or early 2016 volumes of the journals. Only articles reporting research that involved data analysis were included. Journal articles that report literature surveys or philosophical/theoretical discussions were excluded. Table 5. Number of research papers selected from sociology journals. 6.3. Coding of information-argument types We carried out manual annotation of argument elements in the text using oXygen XML Editor version 18.1 (a software for annotating and editing XML files). The coding was usually at the sentence level. Only one code was assigned to each text unit. If multiple argument types were possible (depending on the reader’s interpretation), the coder had to select the best code. figure 1 gives an example of annotated text. argument claims are INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 139 Code Journal title Sample S01 American Journal of Sociology 6 S02 Annals of Tourism Research 9 S03 Cornell Hospitality Quarterly 7 S04 European Sociological Review 6 S05 Gender Society 6 S06 Information Communication Society 8 S07 Journal of Marriage and Family 7 S08 Social Networks 10 S09 Qualitative research 5 S10 American sociological review 6 Total 70 Note: The different sample sizes is because Development and Evaluation research and Descriptive research papers are less common, and additional papers in these two categories were sampled from journals that published more of such research papers. T G edCo 1S0 G eltitalnruJo mSa ygoloicSofolanruoJnacireAm 6 elpm 2S0 3S0 4S0 5S0 6S0 7S0 8S0 9S0 hcraesRemsiruTofoslanAn 9 ylretrauQytilatipsHollenrCo 7 weivRelacigoloicSonaeporEu 6 yteicoSrednGe 6 ntioaicnummo CntioarmfoIn yteicSo 8 yliamFand age irarMofnalJour 7 skrwotNelaicSo 10 hcraeserevitatilaQu 5 10 6 0S1 alTot acebsiseziselpmastnereffidehT:etNo paperonalitaddiand on,mcomess le ar s.paperch esearrsuch weiverlacigoloicosnacireAm 6 70 tnempolevDeesua and ahcraesernoitaulaEv eplsame erwes iegorcato wthese tn is d mfro 70 hcraeserevitpircsDedna s paper f oreo mdehlisbut pa thlsarnu jo L v f n n s o n u y d h , shown with red background colour; argument supports with yellow background. 6.4. Sequential association rule mining To analyse sequential argument patterns, we carried out sequential association rule mining using IBM/SpSS Modeler version 18.1 to identify common sequences of argument elements that occurred in at least 20% of the papers. We analysed sequences of two to five adjacent elements, as well as sequences of up to five elements that are not necessarily adjacent (i.e., with any number of intervening elements), thereby allowing some claims in an argument chain to be skipped. Figure 1. Example argument element coding. In addition to mining sequential argument patterns, we also analysed logical argument patterns. It is difficult to analyse logical argument patterns because the argument supports may be presented a few sentences before or after the claim. furthermore, the argument support may not be explicitly expressed, but implied by adding citations in parenthesis, or by tacit appeal to general knowledge. We focused our analysis on explicit support => claim and claim <= support sentence pairs that have a clear relationship. We view this explicit support => claim relation as a kind of coherence relation, where the reader readily makes out a logical or conceptual relation between the sentences. 6.5. Inter-coder reliability In this study, argument element coding was carried out by the two authors. The first author annotated the text first, which was reviewed and revised by the second author (deemed the more experienced researcher). This two-phase coding compensates for individual oversights and biases, as the second coder would be well aware of the first coder’s perspective and will take that into WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154140 T G G e 2 h 1 6 0 w 7 L v f n n s o n u y d h , consideration before changing a code. We found that coding of argument claims and supports has to be done by experienced researchers: it is difficult to train undergraduate and graduate students to do this analysis (cheng, 2020). In total, 21% of the codes were changed by the second coder. Intercoder analysis of the changes made by the second author reveals differences of interpretation, but also highlights the assumptions underlying the final coding. The error analysis suggests that the first author was looking at more superficial features (word markers that suggest the code), whereas the second author was looking at deeper intentions (that may need to be inferred). Some of the coding conflicts reflect difficulties with the following types of argument claims and supports: 1. Concept/theory/model-related claims and Literature review-related claims. This reflects a common confusion of whether a cited paper is related to a particular concept/theory/model; 2. Research result, General result and Research contribution/recommendation claims. This reflects the common difficulty of understanding general (broader) and detailed (narrower) research findings. a small intercoder reliability test was carried out on this coding. a sample of 10 articles from the corpus was coded by both authors. cohen’s kappa coefficient of 0.782 was obtained by analysing 261 argument and non- argument elements. cohen’s kappa measures the proportion of agreement after removing chance agreement between two coders, and our value of 0.782 can be considered substantial agreement. Some authors considered kappa values of 0.21 to 0.40 to represent fair agreement, 0.41 to 0.60 moderate agreement, and 0.61 to 0.80 substantial agreement (cohen, 1960; Landis & Koch, 1977; Viera & garrett, 2005). an analysis of the conflicts found that the following pairs of argument types tended to be confused: 1. Topic centrality and General statement claims. Both are broad overview claims, and whether there is an intention to highlight the importance of the research topic is a judgement call. 2. Literature-evaluate/critique, Concept/theory/model/-evaluate/critique, and Research gap claims. These three claims may overlap: a research gap may be highlighted by critiquing previous papers and theories. INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 141 3. Literature review-related claims versus Literature review-related supports. for example: • Literature-highlight claim is considered a claim, whereas Cited authors claim/opinion can be coded as supporting a claim, such as a General statement. • Literature-summarise/generalise is considered a claim, but it may be coded as Explanation/elaboration and Established knowledge to support a claim. This highlights the issue of whether to code a cited paper as a support or a claim. as for identifying the five types of research, six undergraduate students were recruited to do the analysis. The average inter-coder reliability scores (percentage agreement between coders, and cohen’s Kappa) were substantial at 0.72 and 0.60, showing that it not difficult to distinguish between the research types. 7. Findings 7.1. Information-argument profile across the three research types One-way analyses of Variance (aNOVa) were performed to identify significant differences in the information-argument type frequencies across the three types of research. Table 6 lists the information-argument types with significant differences (α=0.05). The information-argument profile for the three research types can be characterised as follows: 1. for Investigative research: Research gap, Research hypothesis, Research method, Research purpose (especially Research objective), and Research finding (including research result and research contribution) elements had the highest frequencies. 2. for development and Evaluation research: Concept/theory/model- related elements (especially Concept/theory/model-derive and Concept/theory/model-evaluate/critique) had the highest frequencies. The non-argumentative Descriptive information element occurred on average about six times, especially Concept/theory/model-details and Outline the structure. WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154142 3. for descriptive research: the Research result element had a slightly higher frequency. Descriptive information elements occurred about five times on average. This reflects different information and argument strategies needed for different research types: 1. Investigative research papers usually argue for a research gap to justify the research objective. research finding (particularly research contribution) is also claimed in the Introduction. 2. development and Evaluation research papers have to argue for the novelty (idea) for the theory or method proposed. as most of these studies in sociology sought to develop a new theory or improve an old theory, it is not surprising that there were more theory-related elements. an outline of the paper structure was often found at the end of the Introduction section, suggesting that there is no standard structure for the whole paper. 3. descriptive research papers do not have a distinctive argument strategy. However, they have more descriptive elements and contextual information. a summary of the research results is often provided. an outline of the paper structure is often provided at the end of the Introduction section. INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 143 Table 6. Common information-argument types in the introduction sections (frequency distribution). 7.2. Opening move, middle game and concluding move We analysed the information-argument patterns in terms of which information-argument types tended to appear at the beginning, middle and end of the Introduction section - in other words, the opening move, middle game and concluding move (see Table 7). for all types of research, the Introduction section often opened with Topic centrality (20% to 40% of the sample) or General statement (20% to 30%). descriptive research sometimes opened with non-argumentative descriptive or context information (20% to 25%). WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154144 Information-argument type Type of research Investigative Development and Evaluation Descriptive (N=30) (N=20) (N=20) Main research claims - - - - Research gap Mean 2.03 1.00 0.80 SD 1.65 1.97 1.11 Research purpose claims Mean 4.57 2.05 2.45 SD 1.91 1.70 2.21 Research objective Mean 2.77 1.45 1.80 SD 1.63 1.19 1.36 Research hypothesis Mean 0.90 0.16 0.00 SD 1.09 0.50 0.00 Research method Mean 1.87 0.55 0.80 SD 2.16 0.89 1.06 Research finding claims Mean 3.00 2.05 1.80 SD 3.41 3.52 2.07 Research result Mean 1.07 1.05 1.30 SD 2.08 1.67 1.95 Literature review-related claims - - - - Literature-summarise/generalise Mean 2.53 1.50 0.80 SD 2.11 1.99 0.83 Concept/theory/model-related claims Mean 3.23 7.90 3.20 SD 3.48 7.76 3.76 Concept/theory/model-derive Mean 0.07 1.00 0.10 SD 0.25 1.34 0.31 Concept/theory/model- evaluate/critique Mean 0.17 1.45 0.05 SD 0.59 2.42 0.22 Descriptive information Mean 0.97 5.40 4.75 SD 1.61 5.13 5.16 Concept/theory/model-details Mean 0.10 1.45 0.15 SD 0.55 3.19 0.67 Outline the structure Mean 0.63 2.30 1.50 SD 1.40 2.64 1.91 Note: Significant differences across the research types (!=0.05). T ' at the end of the Introduction section, all research types had Outline the structure as the final element: more often for development and Evaluation research and descriptive research (40% to 45% of sample) than for Investigative research (23%). Investigative research and descriptive research concluded with Research result and Research contribution/recommendation 20% to 37% of the time. In contrast, development and Evaluation research made Concept/theory/model-related claims about 20% of the time. In the middle of the Introduction sections, Investigative research and descriptive research typically made use of this sequence of elements: Literature-summarise/generalise … Research gap … Research objective. descriptive research sometimes added a Concept/theory/model-apply claim. However, Development and Evaluation research can have any kind of elements (i.e., there is no dominant pattern). Table 7. Common argument elements at the beginning, middle and end of Introduction sections, across the research types. INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 145 Type of research Position Common argument elements Investigative research Begin Topic centrality or General statement Middle Literature-summarise/generalise … Research issue or Research gap or Research question … Research objective … End Research contribution/recommendation Inserted element - Research method: anywhere in the Introduction section - Concept/theory/model-define: anywhere in the Introduction section Development and Evaluation research Begin Topic centrality or General statement or (Topic centrality -> Concept/theory/model-define) Middle (any argument element) End Outline the structure Inserted element - Any argument element: in the middle of the pattern Descriptive research Begin Topic centrality or Descriptive information or Context information Middle Literature-summarise/generalise … Research gap (Concept/theory/model-apply)? … Research objective …(Concept/theory/model-apply)? … End Outline the structure Inserted element - Research method: before or after the Research objective - Concept/theory/model-define: anywhere before the Research objective 7.3. Sequential argument patterns We analysed the sequences of information-argument types starting with 2- step and 3-step adjacent patterns. We then analysed sequences of information-argument types with any number of intervening elements (i.e., allowing some claims/supports in an argument chain to be skipped). Sequential argument patterns in the Introduction sections are too numerous to list here. They are listed in the first author’s phd thesis (cheng, 2020). We attempted to derive one basic or overall sequential argument pattern for the Introduction section, by linking and combining the frequent sequences of adjacent and non-adjacent elements. The assumption is that researchers doing a particular type of research should be familiar with these basic argument patterns. In practice, researchers are expected to modify or extend the basic argument pattern to present a strong case for their research. The basic argument patterns that we derived for each research type can be considered a theory underlying the argument structure of each research type. We summarise the basic sequential argument patterns separately for each research type below. Examples of the argument patterns from our corpus are available from dr-ntu (data).5 7.3.1. Investigative research The sequential pattern mining did not detect any pattern with a high frequency (over 50%). The most common pattern was Research objective -> Research method -> Research result (40%). Moreover, General statement was the first argument claim in about 30% of the Introduction sections; and Topic centrality in about 20%. The results indicate the following basic argument pattern in the Introduction sections: (General statement or Topic centrality) … Literature-summarise/generalise … (Research issue or Research gap or Research question) … Research objective … Research contribution/recommendation. In addition, we found four other more specialised patterns: 1. The basic argument pattern, with additional literature review to support Research gap. a Literature-summarise/generalise claim may be WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154146 inserted before or after a Research gap claim, or between two Research gap claims. 2. Multiple Research gap claims, supported by literature review. 3. Extended literature review, leading to the Research objective. 4. additional Research method element, with the Research method appearing after the Research objective or between two Research objective claims. 7.3.2. Development and Evaluation research The sequential argument patterns were more elaborate than those for Investigative research, as development and Evaluation research papers had more Concept/theory/model-related claims (20%). a Concept/theory/model- evaluate/critique claim (20%) was often the final element. Thus, the following basic sequential argument pattern was derived: (Research objective or Concept/theory/model- define/derive/apply)… Research idea/approach … Research objective … (Research result or Concept/theory/model- evaluate/critique or Research contribution/recommendation). In the above pattern, Research idea/approach replaces the Research gap in the argument pattern derived for Investigative research, because development and Evaluation research generally proposes a new idea or approach, rather than address a Research gap in the literature. 7.3.3. Descriptive research Non-argumentative elements (i.e., Descriptive information and Context information) played an important role in descriptive research. The following elements each comprise 10% to 15% of the Introduction sections: 1. Descriptive information (especially Data context, and Outline the structure); 2. Context information (especially Practical background, and Historical background) INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 147 3. Literature review-related claims; 4. Concept/theory/model-related claims. This indicates that descriptive research papers have more descriptions of qualitative data, which is based on a theory/model/framework and previous research. The analysis of sequential argument patterns did not find any pattern with a high frequency (over 50%). The results indicated the following overall/basic argument pattern in the Introduction sections: (Topic centrality or Descriptive information or Context information) … Literature-summarise/generalise … Research gap … (Concept/theory/model-apply)? … Research objective … (Concept/theory/model-apply)? … Outline the structure. Concept/theory/model-define claim can be inserted anywhere before the Research objective claim; the Research method claim may be inserted before or after the Research objective claim. 7.4. Logical argument steps in the Introduction sections We also analysed logical argument steps irrespective of sequential order in the text. However, this analysis was limited to argument elements occurring in adjacent sentences, so that the relation between argument claim and argument support is direct and explicit. Thus, only a relatively small number of explicit argument steps were found in each Introduction section. We found that authors exerted considerable effort to support the Research objective with Research gap (18 of 70 papers) or Research motivation/justification (13) or Topic centrality (7). Investigative research and descriptive research papers tended to support the Research objective with a Research gap claim. Investigative research also used Topic centrality, whereas descriptive research used Practical problem to support a Topic centrality. Two common logical argument steps were found in all research types: 1. Literature-summarise/generalise => Research gap (11 of 70 papers) WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154148 2. Explanation/elaboration => Topic centrality (10). This indicates that authors typically support a research gap by summarising previous studies, and support the importance (centrality) of the research topic with additional explanations. The common logical argument steps across the research types are shown in figure 2. Figure 2. Common logical argument steps across the types of research. 8. Conclusion We have analysed the types of information and argument claims and supports used in the Introduction sections of sociology research papers, as well as common patterns of claims and supports. We identified three types of sociology research: Investigative research, development and Evaluation research, and descriptive research. a coding scheme was developed for information-argument types, divided into those that carry or imply an argument claim (group 1), those that play the role of argument support (group 2), descriptive information types (group 3) and context information types (group 4). We identified significant differences in information-argument profile (i.e., distribution of information-argument types) among the three types of research papers, as well as differences in sequential argument patterns and logical argument patterns. The Introduction sections reflect different information and argument INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 149 ! ! strategies across the three research types: Investigative research papers usually argue for a research gap to support the research objective. research contribution is also often claimed in the Introduction. development and Evaluation research papers typically argue for the novelty of the proposed theory or method. There are also more theory-related elements to support the new theory or method, and more descriptive elements to give details of the theory. descriptive research papers do not have a distinctive argument strategy, but have more descriptive elements and context information, especially at the beginning of the Introduction. a summary of the research results is often provided. Both development and Evaluation research and descriptive research papers have Outline the structure at the end of the Introduction nearly half the time, suggesting that there is no standard paper structure, and a paper outline is helpful to the reader. The results of sequential pattern mining of Introduction sections indicate the following basic argument pattern in the Introduction sections: (General statement or Topic centrality) … Literature- summarise/generalise … (Research issue or Research gap or Research question) … Research objective … Research contribution/recommendation. development and Evaluation research and descriptive research papers may modify the basic argument pattern with Concept/theory/model-related claims. development and Evaluation research papers often replace Research gap with Research idea/approach, whereas descriptive research papers open with Descriptive information (especially Data context) or Context information (especially Practical background, and Historical background) about a quarter of the time. from an analysis of the logical argument steps irrespective of sequential order, but presented in adjacent sentences, we derived a typical argument structure for the Introduction section, represented in figure 2. Our method of argument element coding using XML tags has allowed us to carry out quantitative analysis and sequential pattern mining to gain insights into the information-argument structure of academic texts. We expect that applying this analysis method to other sections of the research paper (i.e., literature review, results, discussion/conclusion) will yield more interesting results. We believe our results for Introduction sections would generally hold for other social science fields, though this is to be confirmed in subsequent studies. Scientific and engineering disciplines may have other research types that have evolved other kinds of information-argument structure. WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154150 Our information-argument structure analyses were carried out mainly at the sentence level to find patterns in what freeman (2011) termed argument macrostructure. In complementary work, we are analysing the information microstructure of argument claims and supports at the level of concepts, entities and semantic roles, to identify information links between claims and their supports. an initial study of how comparison information structures and cause-effect information structures are linked together to support the validity of the cause-effect claim (in research result) has been reported in cheng and Khoo (2021). Acknowledgement This study was funded partly by the Singapore Ministry of Education research grant MOE2015-1-Tr05. Article history: Received 10 February 2022 Received in revised form 30 August 2022 Accepted 11 September 2022 References INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 151 Cheng, W. -N. (2020). Argument and information structures in sociology research papers: Analysis of the abstract and introduction sections. Doctoral thesis, Nanyang Technological University, Singapore. Cheng, W. -N., & Khoo, C. S. G. (2021). Information structures in sociology research papers: Modeling cause-effect and comparison relations in research objective and result statements. Journal of the American Society for Information Science and Technology, 72(11), 1367-1385. Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37-46. Folli, R., & Harley, H. (2013). The syntax of argument structure: Evidence from Italian complex predicates. Journal of Linguistics, 49(1), 93-125. Freeman, J. B. (2011). Argument structure: Representation and theory. Springer. Erduran, S., Simon, S., & Osborne, J. (2004). TAping into argumentation: Developments in the use of Toulmin’s argument pattern in studying science discourse. Science Education, 88(6), 915- 933. Hinton, M. (2019). Language and argument: A review of the field. Research in Language, 17(1), 93-103. Jalilifar, A., Keyvan, M., & Don, A. (2017). The effect of writing prompts on developing argumentative essays of Iranian university students of English. Journal of Modern Languages, 27, 48-76. Kanoksilapatham, B. (2005). Rhetorical structure of biochemistry research articles. English for Specific Purposes, 24(3), 269-292. Kathpalia, S. S., & Khoo, C. S. G. (2020). Generic structure and citation functions in introductions of biological science articles in English-medium international journals. Ibérica, Journal of European Association of Languages for Specific Purposes, 40, 123-148. Dr Wei-Ning Cheng received her phd in Information Studies from the Wee Kim Wee School of communication and Information, Nanyang Technological university, Singapore. Her research interests are in argument and information structures analysis, knowledge organisation and graph visualisation. Dr Christopher Khoo is an associate professor in the Wee Kim Wee School of communication and Information, Nanyang Technological WEI-NINg cHENg & cHrISTOpHEr S. g. KHOO Ibérica 44 (2022): 127-154152 Kohen, F., Milsark, G., & Martin, N. (2011). Effects of syntactic and semantic argument structure on sentence repetition in agrammatism: Things we can learn from particles and prepositions. Aphasiology, 25(6-7), 736-747. Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159-174. Lee, S. H. (2013). Argument structure as an interactive resource by undergraduate students. Linguistics and the Human Sciences, 9(3), 277- 306. Lincoln, Y. S., Lynham, S. A., & Guba, E. G. (2011). paradigmatic controversies, contradictions, and emerging confluences. In N. K. Denzin & Y. S. Lincoln (Eds.), The Sage handbook of qualitative research (pp. 97-128). Sage. Liu, X., & Furneaux, C. (2015). Argument structures in Chinese university students’ argumentative writing: A contrastive study. English Text Construction, 8(1), 65-87. Munyua, J. M. (2021). Basics of research ang academic writing: A text book for research students and lectures. Exceller Books. Moretta, F. D., Feltracco, A., Jezek, E., & Magnini, B. (2018). Designing a methodology for semantic type tagging of argument positions. IJCoL: Italian Journal of Computational Linguistics, 4(2), 57-72. peldszus, A., & Stede, M. (2013). From argument diagrams to argumentation mining in texts: A survey. International Journal of Cognitive Informatics and Natural Intelligence, 7(1), 1-31. Rusfandi. (2015). Argument-counterargument structure in Indonesian EFL learners’ English argumentative essays: A dialogic concept of writing. Regional Language Centre. RELC Journal, 46(2), 181-197. Schwandt, T. A. (1998). Constructivist, interpretivist approaches to human inquiry. In N. K. Denzin & Y. S. Lincoln (Eds.), The Landscape of Qualitative Research (pp. 221-259). Sage. Sheinfux, L. H., Melnik, N., & Wintner, S. (2017). Representing argument structure. Journal of Linguistics, 53(4), 701-750. Simon, S. (2008). Using Toulmin’s argument pattern in the evaluation of argumentation in school science. International Journal of Research & Method in Education, 31(3), 277-289. Swales, J. (1990). Genre analysis: English in academic and research settings. Cambridge University press. Tindale, C. W. (1999). Acts of arguing: A rhetorical model of argument. SUNY. Toulmin, S. (2003). The uses of argument (2nd ed.). Cambridge University press. van Eemeren, F. H., Garssen, B., Krabbe, E.C.W., Snoeck Henkemans, A.F., Verheij, B., & Wagemans, J.H.M. (2014). Handbook of argumentation theory. Springer. Viera, A. J., & Garrett, J. M. (2005). Understanding interobserver agreement: The kappa statistic. Family Medicine, 37(5), 360-363. Wingate, U. (2012). ‘Argument’ helping students understand what essay writing is about. Journal of English for Academic Purposes, 11(2), 145-154. Zhang, L., Kopak, R., Freund, L., & Rasmussen, E. (2011). Making functional units functional: The role of rhetorical structure in use of scholarly journal articles. International Journal of Information Management, 31(1), 21-39. university, Singapore. He teaches courses in knowledge organisation, and data analytics. His research interests are in knowledge graph, text mining, and multidisciplinary study of academic writing & thinking. NOTES 1 By “logical” we do not mean logical entailment as used in formal logic (propositional logic or first order logic), but that the support => claim links are independent of the surface (sequential) presentation in the text. The support and claim need not even appear close together in the text. Thus, a logical argument structure is a representation of the conceptual structure that is hypothesised to be constructed in the reader’s mind after reading the research paper. 2 https://researchdata.ntu.edu.sg/file.xhtml?fileId=96819&version=2.0 3 https://researchdata.ntu.edu.sg/file.xhtml?fileId=96821&version=2.0 4 https://doi.org/10.21979/N9/Ld3EBQ 5 https://researchdata.ntu.edu.sg/file.xhtml?fileId=96819&version=2.0 INfOrMaTION aNd arguMENT paTTErNS IN THE INTrOducTION SEcTIONS Of SOcIOLOgy rESEarcH papErS Ibérica 44 (2022): 127-154 153