engleski

Međujezična analiza odnosa između metadiskursa i koherencije u pisanom diskursu neizvornih govornika

One of the most important property of communicatively competent non-native language users is the ability to use their knowledge about how to arrange language elements (words, sentences) to form an articulate written text which will display not only accurate but also adequate use of language in the given linguistic and situational contexts. In communicative competence models, this ability has, generally, been subsumed under the notion of discourse competence (cf. Canale, 1983, Bachman & Palmer, 2010, CEFR, 2001) which includes knowledge about cohesion and coherence. Constructing a coherent text requires the ability to use metadiscourse (cf. Ädel 2006, Hyland 2005), i.e. cohesive devices as means of establishing syntactic and semantic relations between sentences. A number of studies comparing native and non-native use of metadiscourse (mainly focusing on connectors) in written texts as indicators of text coherence and cohesion have shown that non- native users tend to overuse or misuse certain metadiscourse elements (for example, they use connectors in inappropriate places or in ways leading to misunderstanding). However, with few exceptions, previous studies have only included non-native users of English. Furthermore, they have not given sufficient consideration to specifically which metadiscursive elements are overused or underused, which has raised the question as to what factors influence those results in specific second languages and learning contexts. The present study sets out to fill this void by systematically exploring cross-linguistic similarities and difference and the role of source and target languages and cultures in texts written by non-native users of three different foreign languages (German, English and Hungarian) who share a common first language (Croatian). The study addresses a) the frequency and variation of non-native writers' use of metadiscursive elements in sentence- initial position (SI-ME), b) cross-linguistic similarities and differences in non-native and native writers' choices of these metadiscursive elements, in terms of their frequency and variation, and c) the relationship between the use of SI-ME and ratings of text cohesion and coherence. By comparing texts in different foreign languages written by writers who share their first language, the study aims at exploring to what extent the observed differences may be attributed to the interference of culture- specific patterns in creating an argumentative essay. Namely, languages and cultures do not necessarily share their approaches to rhetorical organisation, interpretation of text coherence, and ways in which they mark relations between linguistic units in texts. It is this culture-specific variation that underlies the assumption that the concept of coherence acquired in the first language influences the creation of texts in a second language. For example, the Hungarian language requires global marking, which is why discourse markers play an important role in Hungarian, whereas in English, metadiscourse, especially in sentence-initial position, tends to have controversial status. Also, the use of metadiscourse may depend on writers' perception of concepts (such as the extent to which readers and writers share background knowledge), which may also emanate from culture-specific conceptualisations of coherence and cohesion. Qualitative and quantitative analyses of SI-ME were conducted on argumentative essays written by Croatian learners at B2 level in English, German and Hungarian as foreign languages. Participants were asked to write an essay of about 230 words presenting two points of view, expressing their opinion and giving supporting arguments. Thus, comparable argumentative essays were elicited but in order to avoid topic bias, topics were varied. In total, the English subcorpus consisted of 7, 257 tokens, the German subcorpus contained 8, 585 tokens, and the Hungarian subcorpus had 6, 212 tokens. The analysis of SI-ME was based on Hyland's (2005) Interpersonal model of metadiscourse, i.e. its interactive dimension. Writers select interactive metadiscourse elements to create relations between and elaborate on propositions, to sequence and refer to information in different text parts, and to, ultimately, mould their relationship with the reader. These elements are particularly prominent in argumentative essay in which writers seek to present and defend an argument, and to ensure that readers understand their intended line of reasoning (cf. Ädel 2006). For the quantitative analysis of SI-ME in the corpora, Sketch Engine (http://www.sketchengine.eu) was used. The essays were also rated for overall text quality, cohesion and coherence by trained raters following two sets of criteria. The first set included the text type (i.e. the degree to which a text complies with conventions of text structure in terms of introduction, body and conclusion), cohesion (appropriate and efficient use and range of cohesive devices), and coherence (both local and global). The second set included task achievement, vocabulary (range, appropriateness, orthography), and grammar (range and accuracy). Each component was evaluated on the assessment scale ranging from 0 to 5. The total number of points was taken as an indicator of overall text quality. Points for the two cohesion subscales and the two coherence subscales were taken as indicators of text cohesion and coherence. The results point to several noteworthy conclusions: second language writers' use of SI-ME is predominantly redundant, they vary the use of SI-ME to a limited degree (e.g. but in English, and its counterparts aber in German and de in Hungarian appeared almost as often as all other SI-ME put together), they use a restricted set of SI-ME, and generally show a lower level of knowledge of the appropriate use of SI-ME (e.g. disrupt text coherence by using a wrong metadiscourse element in terms of its function). The limited range and variation of SI-ME seems to be an effect of language proficiency: the fewest instances, the least variation, as well as the greatest number of formal errors was found in the Hungarian subcorpus, indicating an overall lower proficiency level. The English subcorpus showed the greatest variation and a general overuse of specific SI-ME (which was also observed in the German subcorpus), but, interestingly, non- native English writers tend to create non- existent expressions and make stylistically inappropriate choices (e.g. For starter). In general, deviations from native- speakers' use may be attributed, as expected, to the influence of the first language, but also other foreign languages participants are exposed to, particularly English. The correlation analysis generally indicated that there is a positive relationship between SI-ME and text coherence, cohesion and overall text quality. However, this finding might not necessarily suggest that the more SI-ME non- native writers use, the higher the levels of cohesion, coherence and text quality, but that text raters might be inclined to perceive texts abundant in metadiscursive elements as more coherent. This warrants further research and an alternative method of measuring text coherence. Learners' erroneous use of SI-ME can be attributed to their insufficient knowledge of stylistic constraints on the choice of metadiscursive elements, especially in formal styles (as indicated by use of informal SI-ME typical of oral interaction in the English subcorpus). These findings can be partially attributed to the effects of instruction. Namely, metadiscursive elements are very often taught explicitly by presenting simplistic lists of items with no elaboration on the constraints of their use. Moreover, teachers may inadvertently choose to place greater emphasis on particular features either as a way of complying with external evaluation criteria (such as the ones imposed by standardised school-leaving exams) or because of their own lack of awareness of rules and constraints. As a result, learners do not develop adequate knowledge of use of metadiscourse. Instead, in learners' minds they are accumulated in form of lists where metadiscourse items are assigned a particular function (e.g. items used to start a paragraph or to introduce a counter-argument) without realising that each metadiscursive element has semantic, syntactic and, especially, stylistic properties governing its use. These findings have important implications for practice and future research. More studies are required to investigate purposefulness of explicit versus implicit teaching and learning. Although SI-ME play a role in text coherence, in order to gain a better understanding of this issue other metadiscursive elements and means of establishing coherence must be considered.

discourse competence ; metadiscourse ; non-native text ; sentence-initial metadiscursive elements ; text cohesion and coherence ; English as a foreign language ; German as a foreign language ; Hungarian as a foreign language

nije evidentirano