Abstract
Voice in academic writing refers to the author's sense of presence in a text which can be expressed through the use of lexical discourse markers. This study investigated meta-discourse markers of voice in humanities research articles based on Hyland’s classification. The research questions were: a) what lexical tools are used to express author's voice and how frequent are they? b) Are there significant differences between the use of hedges, boosters, attitude markers, self-mentions, pronouns, and directives?  Authors' uses of discourse markers of voice were coded and quantified. These expressions were compared across types of meta-discourse categories. The data included fifty randomly-selected research articles published in indexed Iranian journals in the last decade. The research method was a mix of qualitative (text analysis) and quantitative procedures (frequency counting and statistics). The results showed that 1) in Persian articles written in the humanities, hedges for the cautious expression of statements, boosters for definite expression of ideas, and attitude markers for the expression of author's feelings about his propositions were the most frequently used type of discourse markers, 2) the use of first person pronouns to express the author's voice clearly, pronouns referring to the reader of the article and directives were very infrequently observed, and 3) inferential tests showed that frequencies were significantly (p≤0.5) different across type of stance and engagement markers.
Keywords: Academic Writing, Research articles, Voice, Hyland, Stance markers
 
Introduction
According to the interactive meta-discourse model in the text, discourse markers are used to show the author's perspective on the propositional content of the text and on the reader (Hyland, 2005, p. 25). Interactive meta-discourse is related to the way the author of the text communicates linguistically with his/her writing and its readers. This includes various categories of lexical markers, examples of which are hedges, boosters or certainty markers, self-expression, and other markers. Based on Hyland (2005) and Taremi et al. (2018), different types of interactive discourse markers can be briefly defined as follows. Hedges express the author's doubt about the correctness of the stated proposition. Boosters depict the author's certainty about the correctness of the text. Attitude markers show the author's perspective on the text, the importance of the text, and his/her surprise. Self-expressions or self-references are phrases in a text through which the author announces his or her presence in the text. Subjunctives, imperatives and interrogatives are also used for communication between the author and the audience.
How authors express their active presence beyond the text and expresses their voice in the text has not been fully investigated, and research data on author’s voice is scant for Persian-speaking scholars. The author's voice and the way he deals with the ideas he writes can be heard through the words and sentences he writes. The question is, what linguistic markers does the Persian-speaking author of a scientific article use to determine his approach to the text? How does he interacts with his readers? How does he show the degree of his doubt and certainty? In other words, is there a way to determine the strength of the researcher's voice in his writing, considering linguistic models in the field?
The purpose of this study was to explore some components of author's voice in humanities research articles published in Persian. The aim was to see whether the author's voice in a scientific article can be heard through the words and sentences he writes. The purpose was also to see explore the frequency of words and phrases used to express the author's voice. Specifically, the aim was to quantitatively and qualitatively examine the use of lexical expressions/markers related to meta-discourse based on the Hyland (2005) interaction model. The following questions were examined: 1) What lexical tools are used to express the author's voice? 2) Does the frequency of use of various components of the author's voice (such as imperative phrases, precautionary phrases, first-person pronouns, etc.) differ in the articles?
 
Materials and methods
The research method used in this study include both quantitative methods based on detailed study and examination of lexical frequency and qualitative methods based on coding and text analysis. In the qualitative part, the selected corpus was examined in terms of the use of author's voice components using the linguistic model presented by Hyland (2005). The texts were coded by two experts to determine the use of author's voice components in these scientific articles.  In the quantitative phase, frequency and percentage were calculated for each of the voice marker types. First, fifty research articles in the humanities were randomly selected from among the articles published in journals indexed by the Iranian Ministry of Science, Research, and Technology. The unit of analysis was each sentence in the abstract, introduction, discussion, and conclusion of the articles.  In total, the coded texts included 73,676 words, and the longest section was the discussion and conclusion, and the shortest was the abstract. The analysis, coding, determination, and counting of the components of the author's voice were carried out based on the definitions specified in the theoretical and research background and based on the coding protocol agreed upon by two coders with an agreement coefficient of 97 percent.
 
 
Results and Discussion
The first aim of the study was to see what lexical tools are used to express the author's voice in scientific articles by Iranian scholars in the humanities. The analysis of the articles based on Hyland’s (2005) model showed that scholars used all categories of linguistic markers extracted from the model to express their presence in the text and to express voice. These markers included all categories of discourse markers reported in previous studies of research articles. These markers were scattered in the abstract, introduction and conclusion of the articles. The second aim was to explore the frequency of authorial voice markers. The results of Kolmogorov-Smirnov test showed significant differences among categories. Boosters or certainty indicators came with the highest frequency followed by hedges in the second row. References to the author himself, which indicate ownership of the idea, and imperative sentences had the lowest frequency.
 
Conclusion
The study showed infrequent and weak use of many discourse marker types that can strongly show authors voice. Inappropriate use of meta-discursive markers in research articles published by Iranian scholars can be partly attributed to writing culture in their academic community, insufficient training and workshops on academic writing, and lack of interaction between journal staff and authors. The explored articles can be marked weakly interactive where the author is afraid of taking responsibility for the claims and prefers to use markers less than expected, reflecting a weak author's voice.