English-language abstracts written by Czech linguists: how are we doing?

English-language abstracts written by Czech linguists: how are we doing? Tomáš Gráf (Praha) ABSTRACT Article abstracts may be among the most efficient means of disseminating research results if certain basic principles regarding their content and structure are followed. No research outcomes have yet been made available on the structure of abstracts published in Czech linguistic periodicals. The current study presents asurvey of 120 such abstracts collected from four journals and, on the basis of rhetorical moves analysis, describes their shared features, strengths and shortcomings. The results show that many of these abstracts fail to include moves which are generally considered obligatory (results, data and method description) and that their informative value is consequently lower than sufficient. The study concludes by recommending that awareness of rhetorical principles amongst Czech academics be raised e.g. through instructional texts or courses. KEYWORDS research article abstracts, abstracts, moves analysis, linguistic journals, English for academic purposes 1. INTRODUCTION “Publish or perish”— an all-too-familiar slogan which succinctly if rather menacingly describes the pressure contemporary researchers face in the increasingly competitive environment of the academic world. In order to sustain and develop their professional careers and credibility, academics annually publish 2.5 million academic papers (Jinha 2010) in approximately 30,000 peer-reviewed journals (Boon 2016). The pressure is on them, moreover, not only to publish but also to keep abreast of at least apart of the published output; this, of course, can only happen through reading. And yet, obviously, all is not well that is written, and even less is of relevance to each academic’s narrow specializations, so academics need to make choices as to what to read. And they need to make them fast. This is where abstracts play akey role; awell-written abstract of about 200 words in length can provide abrief summary of the article, thereby allowing the reader to assess its relevance, importance and context quickly and efficiently. As an applied linguist with adeep interest in my field and also as one who understands the need to keep up-to-date with arapidly developing field, Iregularly peruse atotal of around twenty peer-reviewed journals. Over the years Ihave come to appreciate how thorough an overview can be obtained from awell-written abstract (see also Huckin 2001).1 Not only do such abstracts enable me to decide quickly which 1 Huckin (2001) claims that within the context of medicine it has become common that researchers read just the abstracts and often derive vital information just from them (cf. also Borko and Chatman 1963; Friginal and Sabah 2017; Johnson 1995). 72 LINGUISTICA PRAGENSIA 1/2018 article to read (or not to read!) as awhole, but they also provide me with at least something of an overview of the latest trends in the various linguistic subfields which are not immediately relevant to my own research, but which it is nevertheless useful for me to be aware of. Such an overview makes it easier for me to connect alarge variety of ideas and approaches. And, last but not least, good abstracts may also serve as pointers to the relevant passages of articles which Iotherwise might not decide to read in full. Imight even go so far as to say that Ihave become dependent on abstracts. And yet too often do Icome across abstracts which fail to serve any of these purposes; abstracts which provide neither context nor adescription of the data and methods used, and which frequently do not mention any results, let alone implications. Abstracts that are consequently wholly unhelpful to me and, Ibelieve, also to the wider academic public. Ido not wish to speculate as to whether such deficiencies in the structure of an abstract actually symptomize inherent flaws in the study itself, and, actually, Icannot, as bad abstracts rarely tempt me to read the ensuing articles. But Icannot stop wondering if this is the effect unsatisfactory abstracts have on other readers. It is the purpose of this study to analyse aselection of abstracts (written in English) appearing in anumber of linguistic journals published at the Faculty of Arts, Charles University, and to determine whether they provide the detail and structure that readers on the whole expect. Istrongly believe that it is precisely in journals of this kind, which are either less known to the international academic community or publish articles in alanguage other than English (here Czech), that abstracts must be of top quality, for if they are not they do not attract the readership its authors desire and need. One should also be aware of the fact that there exist certain cultural and publishing traditions which might affect the way abstracts in various journals are usually written. Hence, in my analysis, Isimply aim to provide adescription without attempting to be overly prescriptive. 2. ON ABSTRACTS In principle, there is nothing mysterious about the abstract genre. On the contrary, what abstracts should be is clearly set out even in brief dictionary definitions. The Concise Oxford English Dictionary2 defines the abstract as “asummary or statement of the contents of abook etc.”. The Longman Dictionary of Contemporary English3 sees an abstract as “ashort written statement containing only the most important ideas in aspeech, article etc.”, and its older edition (1993) as “asummary of points e.g. of apiece of writing”. Similarly, the Merriam-Webster Dictionary4 gives the definition as “asummary of points (as of apiece of writing) usually presented in skeletal form; also: something that summarizes or concentrates the essentials of alarger thing or 2 Concise Oxford English Dictionary. Oxford University Press; 12th edition (2011) 3 The Longman Dictionary of Contemporary English. Pearson Education ESL; 6th edition (2015) 4 Accessed online on 3 October 2017: https://www.merriam-webster.com/dictionary/ abstract TOMአGRáF 73 several things”. The online Cambridge Dictionary5, going into rather more detail, describes it as “ashort form of aspeech, article, book, etc., giving only the most important facts or ideas”. The online Dictionary.com6 adds that an abstract is “something that concentrates in itself the essential qualities of anything more extensive or more general, or of several things; essence”. These definitions, then, in varying degrees of detail, clearly state that an abstract is ashort text which summarizes alarger text whilst mentioning its essential points. Amuch more detailed description of abstracts is presented by the American National Standards Institute, which on three dense pages not only provides adetailed definition (see below) of an abstract but also distinguishes it from other similar text types (such as annotations, extracts and summaries) and describes its use, its content in terms of components, and its presentation and style. As the definition comprises several details which are important in the present author’s understanding of what agood abstract is, alonger quotation from it is given here: “In this standard, the term abstract signifies an abbreviated, accurate representation of adocument, without added interpretation or criticism and without distinction as to who wrote the abstract. An abstract should be as informative as is permitted by the type and style of the document; that is, it should present as much as possible of the quantitative or qualitative information (or both) contained in the document” (‘American National Standard for Writing Abstracts’, 1977: 252). The key feature mentioned in this description is informativeness and the provision of as much relevant detail as possible. This is because “[a] well-prepared abstract enables readers to identify the basic content of adocument quickly and accurately, to determine its relevance to their interests, and thus to decide whether they need to read the document in its entirety” (ibid.). It also acknowledges the function of abstracts as sources of basic information (see above): “Readers for whom the document is of fringe interest often obtain enough information from the abstract to make their reading of the whole document unnecessary” (ibid.). Adistinction has to be made between so-called traditional abstracts and structured abstracts (Hartley 2008). Structured abstracts, typical in medical research, are organized into separate paragraphs with individual subheadings (e.g. background, aim, method, results, and conclusion) and follow the rhetorical structure of research articles that is commonly referred to as IMRD (Introduction, Method, Results, Discussion) (Swales 1990). Traditional abstracts are loosely organized into usually one paragraph of continuous text in which the sequence of rhetorical moves and, indeed, the inclusion of all of the components common (and compulsory) in structured abstracts is arbitrary. Hartley (2004) shows that, compared to traditional abstracts, structured abstracts are more informative, easier to read and search, and preferred by readers. Hartley (2008) recommends that authors writing for journals which require the tra5 Accessed online on 3 October 2017: http://dictionary.cambridge.org/dictionary/english/ abstract 6 Accessed online on 3 October 2017: http://www.dictionary.com/browse/abstract?s=t 74 LINGUISTICA PRAGENSIA 1/2018 ditional type ought to structure their abstracts as if they were writing the structured type but removing the subheadings and organizing the components into acontinuous text. Such abstracts should then, according to Hartley (2008), have aclearer structure enabling the reader to scan the content more efficiently while also making it harder for the writer to produce incomplete abstracts in which essential information might be missing. As abstracts often stand alone (e.g. when submitted as conference proposals), completeness is essential. Such adescription fits Borko and Chatman’s (1963: 150) concept of informative abstracts which, according to these authors, are supposed to obviate the necessity to read the article in full by enabling the reader to glean all the relevant information from the abstract itself. Borko and Chatman (ibid.) set this against the concept of the indicative/descriptive abstract, whose function is to list the subjects discussed and describe the structure of the article without aiming to be asubstitute for the article itself. They do, however, admit that some abstracts may be best described as acombination of informative and indicative approaches (see also Lorés 2004; Tankó 2017). Much of what we know about abstracts, how they should be structured and what they should contain derives from the seminal work by Swales (1990), who sees abstracts as aspecific genre of academic communication. He understands genre as “amore or less standardized communicative event with agoal or aset of goals mutually understood by the participants in that event” based on an interplay of “cognitive, rhetorical and linguistic features” (Swales 2009: 17–18). For the purpose of identification and illustration of the rhetorical organization of texts, Swales has developed the methodology of rhetorical move analysis, which sees atext as asequence of discourse units with clear communicative functions. These units are called moves, and they may be further divided into smaller sections called steps. Together, these functional units fulfil the communicative purpose of the genre (Connor et al. 1995). Lorés (2004), however, points out that as the identification of moves is determined both by lexicogrammatical features and by cognitive criteria it is asubjective process (cf. Crookes 1986; Dudley-Evans 1994; Paltridge 1994). To reduce its impact, Cortes (2013) suggests that criteria for identification ought to be based on awell-researched and illustrated, frequency-based description of specific linguistic components such as lexical bundles (see also Tseng 2011). Pho (2008) points out that the identification of moves is problematic as it is based both on abottom-up approach (identification is determined by certain linguistic signals) and on atop-down approach (identification is determined by the content). In the current author’s view, these two approaches go hand in hand and supplement each other. If asentence starts with “The aim of the study is to…”, we are clearly dealing with move 2 (describing the purpose) but we also have to inspect whether this is actually in agreement with the subsequent content. Elsewhere, it is possible that no linguistic signals are present (this is especially true for move 1) and we have only the content to rely on when identifying the move. Carrying out the analyses presented below the current author found he frequently had to rely not only on linguistic components but also on move shifts which stemmed from the actual meaning of the analysed texts. Whilst apurely lexically based analysis would undoubtedly speed up some parts of the analytical process it might not always be as accurate. TOMአGRáF 75 Following Swales, research-article abstracts in many scientific fields have been analysed and recommendations have been provided as to how they should be structured. Lorés (2004) argues that an informative abstract should follow astructure for scientific articles that is commonly known as IMRD (see above). The introduction ought to reveal the purpose of the study and the goals of the research. The method section should describe the data and methodology used. The results section is to provide asummary of the findings, and these are then considered in the discussion section, which should also contain an interpretation of the results and mention of possible implications (e.g applicability of the findings) (see also Cutting 2012). Diani (2014) suggests afive-move sequence for linguistics abstracts. This extends Lorés’s (2004) four-move model by including an introductory move called ‘situating the research’. Halleck and Connor (2006) analyze the genre of conference proposals, which share many features of research-article abstracts but frequently allow agreater word count. As aconsequence, more moves may be included. These include context setting (called ‘territory’ by Halleck and Connor), indication of aresearch gap, statement of aim, description of data and procedures, presentation of (anticipated) results or findings, explanation of the benefits, and an importance claim (stating topicality and applicability to the “real” or research worlds). Several studies illustrate that adegree of flexibility is present with regard to the order of moves. Pho (2008) observes that situating and presenting the research are frequently reversed. Tankó’s (2017) observations are similar (especially in that abstracts often start with the statement of the overall aim), but she also notes that cyclical configurations in which moves are interrupted by another move and then resumed are also common (e.g. move 1, move 2, move 1, move 3 etc.). Whilst this is also acknowledged and not disrecommended in the American National Standards for Writing Abstracts (1979: 253), abstracts with an arbitrary order of moves may be slower to process. One of the first studies which focussed specifically on abstracts in applied linguistics (AL) and inspired several similar analyses was that of Dos Santos (1996). He analysed 94 abstracts and observed afive-move pattern of text organisation: Move 1: situating the research Move 2: presenting the research Move 3: method description Move 4: results summary Move 5: discussion of the research. Amongst the key features were move optionality (whilst nearly all of the analyzed abstracts contained moves 2, 3 and 4— Dos Santos consequently calls these moves obligatory— moves 1 and 5 were realized only in about ahalf of the abstracts); move embedding (the inclusion of one move within another); and move reversal. His results suggest that applied linguists modify the structure of abstracts in order to produce cohesive texts and not just checklists of content. Lorés (2004), through an analysis of 36 AL abstracts, found that approximately athird of the abstracts did not follow 76 LINGUISTICA PRAGENSIA 1/2018 the IMRD rhetorical structure but adopted the CARS structure7, which Swales (1990) recommends for article introductions. She also identified several submoves within each move (such as claiming centrality, reviewing items of previous research etc.). Pho’s (2008) analysis of 20 AL abstracts provided similar results to those revealed by Dos Santos (1996): moves 1 and 5 were found to be optional, whilst moves which contribute to situating and presenting the research and its results were deemed compulsory. Pho also found several instances of embeddings. Tseng’s (2011) key observation (based on an examination of 90 AL abstracts) was that afour-move rather than afive-move structure was prevalent, with move 1 appearing only in 41% and move 5 in about 80% of the abstracts. This supports Weissberg and Buker’s (1990) model, in which move 1 is seen as optional. Tseng also reported considerable variation in move optionality and order in the various journals he analysed. In the literature presented above, there is astrong consensus that abstracts are ahighly functional text type whose purpose is to provide the reader with as good an overview as possible of the article in question. To this end, it is recommended that abstracts are composed of sections (moves), each with aclear purpose. While some of these sections appear to be present in most articles and are felt to be obligatory (the description of the aim, data, method and results), other sections are less so (context setting, applicability). However, many authors suggest that although the inclusion of all five moves is highly recommendable acertain degree of freedom in choosing their order is left to the author. From this point on, the current study describes the practice of abstract authors who contribute to selected Czech journals on linguistics. 3. DATA AND METHOD The study corpus consisted of asample of 120 abstracts (written in English) from four major linguistic journals published at the Faculty of Arts, Charles University. All were published in 2015–2017. None of these journals provides detailed guidelines for writing abstracts. Each of the abstracts was analysed with regard to the presence of key rhetorical moves and steps using Swales’s (1990) general methodology for genre analysis, and Dos Santos’s (1996) and Tseng’s (2011) studies of AL abstracts as more specific guides. It has been shown (see e.g. Berkenkotter and Huckin 1994) that abstracts complying with these recommendations are more highly rated by reviewers. The working procedure was as follows. Firstly, Iperused each abstract to be examined to familiarize myself with its basic pattern of organization. Secondly, Iused acombination of bottom-up and top-down approaches to identify the individual moves. Changes in topic and communicative purpose along with specific linguistic features signalled boundaries between moves. These occurred both between and within sentences. 7 CARS stands for “creating aresearch space” and is commonly used to describe astructure for research article introductions. It is athree-move sequence consisting of anumber of steps (Feak and Swales 2011). TOMአGRáF 77 Dos Santos’s (1996) and Swales and Feak’s (2009) five-move model was used as the analytical framework. Move 1 (M1, Background) sets the context of the study. This enables the reader to situate the research within aparticular research context. Example(1) clearly shows that the narrow context of the study is oral fluency and not, for example, psycholinguistic or phonetic aspects of speech rate. It is also recommended that M1 mentions aparticular research gap that the study is attempting to fill. (1) Speech rate is generally recognized as one of the most robust aspects of oral fluency. Little is known about the speech rates of L2 learners. Move 2 (M2, Aim) specifies the aim(s) of the study, and provides the justification for it. Dos Santos (1996) suggests that this may be achieved by describing its key features or by presenting its purpose (see example 2). (2) The current study aims to present the results of acomparison of the speech rate (SR) of advanced learners of English with that of native speakers. Move 3 (M3, Method) describes the data and the analytical procedures used. It may mention, where appropriate, specific techniques, variables, metadata and statistical tests, as in example (3). (3) Aparallel corpus of 50 advanced (C1) learners of English and 50 native English speakers of asimilar age and educational background was used. Speech rate was measured in words per minute (wpm), the temporal information being derived from awaveform analysis and the word count from aconcordancer software. Move 4 (M4, Results) presents asummary of the key findings. (4) The learners’ mean SR was 154 wpm and it ranged from 120 to 210 wpm. The natives’s mean SR was 185 wpm, ranging from 145 to 265 wpm. About 70% of the learners actually spoke as fast as the slower native speakers. Move 5 (M5, Conclusion) discusses the results and their value, and suggests what their implication(s) might be. (5) The overlap between the learner and native SRs shows that the majority of the learners at the C1 level of proficiency have attained aSR comparable to that of native speakers. Consequently, attempts to increase their SR through classroom intervention are not deemed necessary. Finally, Ievaluated each abstract using asimple points system in which one point was awarded to each move containing all of the required information. Quarter points were deducted for the omission of one or more parts, and azero score was given for moves which were missing altogether. Thus, ascore of 1 point signifies afully real- 78 LINGUISTICA PRAGENSIA 1/2018 ized move, and 0.25 points amove which is present but lacks more than two important components. For example, in Move 3 (data and method), 1 point is awarded for the inclusion of both data and method description, and quarter points are deducted for the missing or insufficient description of them. Here, ascore of 0.5 points might mean that either the data or the method are not described at all, or that both of these parts are present but fail to provide all that is required. Thus the maximum score for an abstract is 5 points (i.e. 1 full point for each fully realized move). 4. RESULTS The corpus of 120 abstracts contained 18,274 tokens. The text length ranged between 59 and 358 words, with amean of 152 words (SD=56). Table 1 shows basic descriptive statistics for each of the four journals and reveals that very short abstracts here are not uncommon. Eighteen abstracts (15% of the total) were up to 100 words in length, fifty-three abstracts (44%) were up to 150 words, twenty-four abstracts (20%) were up to 200 words, seventeen abstracts (14%) were up to 250 words, six abstracts (5%) were up to 300 words, and two abstracts (1.7%) were longer than 300 words. Although these journals specify amaximum number of words, many of the authors here fall well below it and also below the customary length recommended in international linguistics journals8. Amedium-strength correlation (r=0.35, p < 0.001) was found to exist between the length of abstracts and the score awarded in the moves analysis presented below. This shows that those writers who exploit the space available to them are more likely to include more rhetorical moves. Journal Abstracts Tokens Mean tokens SD Token range Časopis pro moderní filologii (ČMF) 30 4,114 137 46 62–256 Linguistica Pragensia (LP) 30 4,145 138 42 59–224 Studies in Applied Linguistics (SALi) 30 5,525 184 59 82–312 Acta Universitatis Carolinae— Philologica (AUCP) 30 4,490 150 62 72–358 Total 120 18,274 152 56 59–358 Table 1.Survey of the sources of the examined abstracts The analysis of the moves in the whole corpus revealed (Table 2) that the least frequently used move is Move 5 (discussion of implications), which was not present in 91 (76%) abstracts, and was fully realized only in 19 (16%) of them. The second least frequently used move is Move 4 (presentation of the results), which was not present in 48 (40%) abstracts. However, results were fully presented in 46 (38%) abstracts. 8 Abstracts in Studies in Applied Linguistics (SALi) are amongst the longest in the present corpus. ANOVA with post-hoc Tukey HSD test proves asignificant difference between the mean length of abstracts in SALi and in ČMF and Linguistica Pragensia. TOMአGRáF 79 The most frequently represented move is Move 2 (purpose of the study), which was fully realized in 50 (42%) abstracts and at least partly covered in another 58 (48%) abstracts. Move 2 in its full or at least apartial form is thus present in 108 (90%) abstracts. The second most frequently represented move is Move 3 (data and method), which is present in 90 (75%) abstracts; however, only in 29 abstracts (24%) is it realized fully. As for Move 1, it is fully realized in 56 (47%) abstracts and at least partly covered in another 27 (22.5%) abstracts. Move 1 in its full or at least apartial form is thus present in 83 (69%) abstracts. Scores M1 M2 M3 M4 M5 0 (move not included) 37 12 30 48 91 0.25 3 6 23 7 1 0.5 10 35 23 14 6 0.75 14 17 15 5 3 156 50 29 46 19 Mean score 0.60 0.68 0.48 0.49 0.20 Table 2.Scores for moves in the whole corpus (n = 120 abstracts) Scores M1 M2 M3 M4 M5 015 5 8 10 22 0.25 1 2 6 2 1 0.5 3 11 10 6 2 0.75 3 3 1 3 1 18 9 5 9 4 Mean 0.40 0.58 0.41 0.49 0.20 SD 0.44 0.35 0.34 0.41 0.36 Table 3.Scores for moves in Časopis pro moderní filologii (n = 30 abstracts) Scores M1 M2 M3 M4 M5 09 3 7 16 24 0.25 2 1 10 2 0 0.5 1 6 5 3 2 0.75 2 6 2 1 0 116 14 6 8 4 Mean 0.62 0.73 0.42 0.36 0.17 SD 0.45 0.33 0.36 0.43 0.35 Table 4.Scores for moves in Linguistica Pragensia (n = 30 abstracts)