Description of corpora

To explore the rhetorical move structure of academic lectures and investigate whether there are disciplinary variations of academic lectures in terms of rhetorical move structure, this study has drawn on the academic lectures from the MIT

OpenCourseWare and Open Yale Courses websites to construct corpora for the current investigation. On the course homepages of MIT OpenCourseWare and Open Yale Courses, one can have easy access to rich learning resource of a great number of

69

courses from a wide range of academic disciplines for free, such as class notes, class handouts, and syllabi. More importantly, many of the courses on the two platforms offer complete lecture recordings, which were recorded from authentic classrooms, as well as the corresponding transcripts of lectures for users to browse online or

download for later use.

The resource provided by these platforms could be quite beneficial to those who wish to carry out self-learning considering the rich course materials and the wide variety of course topics offered on the two platforms. Also, according to the site statistics of MIT OpenCourseWare, more than 90% of all visitors are self-learners, students and educators from all around the world who claimed benefits of knowledge enhancement and/or learning new teaching methods from the courses. This means that the online open courses could actually be helpful to both teachers and learners, and also have certain degree of influence among users around the world as well.

Meanwhile, the two platforms have also attracted researchers’ attention

especially those who are interested in exploring lecture discourse. Thus, with the help of the MIT OpenCourseWare and Open Yale Courses websites, there have been several studies conducted to investigate lecture discourse from various aspects, such as exploring frequent linguistic features applying a corpus-based approach

(intensifiers, lexical bundles), or non-verbal features (gesture, gaze, prosodic stress)

70

from a multimodal perspective, (e.g., Crawford Camiciottoli, 2015; Liu, 2019; Liu &

Chen, 2020, forthcoming).

In addition to the fact that the lectures are from two distinguished universities in the US, the crucial reason to choose academic lectures from these two platforms is because of the easy access to the lecture recordings and their corresponding transcripts. From a multimodal perspective, Norris (2004) suggested that the combination of digital video recordings of a discourse and its corresponding transcription would allow researchers to better capture the multiplicity of different communicative modes as to how they interact to create meaning in a particular context. Moreover, in terms of analyzing rhetorical move structures, Lee (2016) emphasized the importance of watching lecture video recordings before conducting the coding process of moves. In this way, the analyst may have a clearer idea

regarding how a lecture is conducted and would have the opportunities to observe and notice non-verbal clues, such as lecturer’s movements and gestures. Such step could help the analyst to identify the discourse function of certain segments and better determine their move categories.

Unlike the move analysis on written discourse where interpreting authors’

intention and determining move categories are rather subjective and challenging, considering the clues are solely from the text and discrepancies between analysts’ and

71

authors’ understanding of a segment may also exist, with the help of lecture

recordings and the corresponding transcripts, it is more likely for analysts to better see speakers’ intention and determine move categories more objectively based on the clues observed from lecture recordings. Therefore, the researcher decided to collect and utilize the lecture recordings and their transcripts from the MIT OpenCourseWare and Open Yale Courses websites for the present analysis.

In terms of researching disciplinary variations, the researcher has drawn on the approaches taken by previous studies researching this topic, and evaluated the accessibility of lecture data to facilitate the decision of the target disciplines for analysis. It has been long established that variation exists across academic disciplines, and the different nature of different academic disciplines has also been widely

discussed in terms of both written and spoken discourses (Belcher, 1994; Biglan, 1973; Hyland, 2006; Hyland, 2008). Generally, in researching the differences among academic disciplines, disciplines are usually divided into two main fields: hard sciences and soft sciences (e.g., Cortes, 2004; Kashiha & Chan, 2013), or classified into four academic tribes: hard-pure, hard-applied, soft-pure, and soft applied (e.g., Hativa, 1997; Smeby, 1996).

It is quite clear that these classifications demonstrate the very different nature between disciplines of hard sciences and soft sciences. With the taxonomy in mind, a

72

cross-comparison study of disciplines would usually start with comparing certain features between disciplines of hard sciences and soft sciences, in order to determine whether the contrasting nature of disciplines would contribute to the usage of certain features. For example, Cortes (2004) compared the use of lexical bundles between history (soft sciences) and biology (hard sciences); Cotos and Chung (2019) compared the use of functional language across physics and chemistry (hard

sciences), and English (soft sciences). It would be more meaningful to conduct cross-comparisons across different sub-disciplines if discipline has been identified as a factor contributing to the different use/occurrence of certain features. Therefore, many cross-comparison studies on discipline variations usually start with comparing the use of one feature in one or more disciplines from both hard sciences and soft sciences before exploring differences among the sub-disciplines under the same discipline.

As indicated previously, disciplinary variations among lectures of different disciplines have been recognized based on many previous explorations from various aspects (e.g., Ädel 2008; Brown & Bakhtar, 1988; Chang, 2012; Deroey & Taverniers, 2011; Dudley-Evans, 1994; Flowerdew & Miller, 1995; Neumann, 2001). Few if any studies have explored disciplinary variations across academic lectures from the aspect of rhetorical move structure. To approach this rather underexplored issue, it is crucial to conduct a preliminary cross-comparison between disciplines of hard and soft

73

sciences to determine whether discipline is a factor that could influence the rhetorical move structures in lectures, before comparisons across sub-disciplines could be conducted in future studies. Therefore, the researcher set to choose two disciplines, with one from hard sciences and the other from soft sciences, for the exploratory cross-comparison of rhetorical move structure in lectures.

To decide which two disciplines should be the targets for the current study, the researcher examined the lectures on the MIT OpenCourseWare and Open Yale Courses websites based on several criteria. First, the lectures should be from undergraduate-level courses, considering more undergraduate-level courses are offered on these two platforms, and many graduate-level courses may involve too much discussion among lecturers and students or presentations from students, which may not fit the usually teacher-led nature of the lecture genre. Second, the lectures for analysis should be at the same course level. Third, the selected courses need to have both the video recordings of entire lectures and the corresponding transcription, given many courses on the platforms do not actually provide such resource but class notes and syllabi. Fourth, to ensure the diversity of course topics and avoid potential

idiosyncrasies of lecturers, the researcher tried to find two disciplines that have a wide variety of course topics and different lecturers from the hard sciences and the soft sciences. Fifth, the courses were conducted by full-time professors, instead of

74

teaching assistants or international teaching assistants, to make sure the lectures for analysis were delivered by rather experienced teachers.

Based on the criteria above, the discipline of computer science was chosen as the target for the current analysis representing the hard sciences, because it had met all the criteria and contains the highest number of different lecturers as well as courses that provide both lecture transcripts and lecture recordings. After the target discipline for hard sciences was determined, the researcher then went on determining a discipline to represent soft sciences. Since the courses offered by the MIT OpenCourseWare are mostly from engineering and science fields, the researcher thus turned to examine the courses on the Open Yale Courses website. Although the number and the topics of courses offered on the Open Yale Courses website are rather limited compared to those on the MIT OpenCourseWare, most courses it offers are from the fields of humanities and social science. Among them, literature is the discipline that includes the widest range of courses and has the highest number of different lecturers as well as satisfied all the criteria. Thus, literature was chosen to be the other target discipline for the current analysis.

Finally, ten introductory-level courses conducted by ten different lecturers, with five from computer science and five from literature, met the aforementioned criteria and were considered suitable for data collection for the current analysis. Then,

75

following Lee’s (2016) procedure of data collection, multiple lecture recordings and the corresponding transcripts from the same course were collected from each lecturer.

The purpose was to better capture the general way each lecturer conducts a lecture.

Watching multiple lecture recordings of the same lecturer would help the researcher to better see how a lecturer conducts a lecture, interpret her/his intention, and identify the move structure. Instead of randomly choosing one lecture out of a lecturer’s lectures, one’s teaching style might emerge or be observed more easily if multiple lectures conducted by the same lecturer are viewed.

The researcher then selected three lectures out of each of the ten courses delivered by the ten lecturers by applying the following criteria. The researcher ensured that the three lectures delivered by each lecturer were selected from the earlier, the middle, and the later parts of the course, and each lecture should focus on different topics. It should also be noted that the lectures selected for analysis were not from a series of lectures of the same topic. Even if one has the same topic as other lectures, the researcher ensured that the selected one was the first one of the series.

Also, the lectures selected for analysis were not the first lectures of the courses, since teachers usually spend much time giving overall course introduction and requirements in the first lecture, and the move structure is likely to be very different from the one of a regular lecture. In addition, when selecting lectures for the analysis, the researcher

76

also tried to select the lectures that have similar length.

In the end, a total of thirty lectures were selected and collected to form the corpora for the current analysis, with fifteen of them delivered by five different lecturers of computer science focusing on fifteen different topics, and fifteen lectures from five literature courses conducted by five lecturers on fifteen different topics. To facilitate later data analysis, both lecture transcripts and lecture video recordings of the 30 lectures were downloaded from the source websites. The lecture transcripts were saved in individual word document for later annotations. The videos were also downloaded in .mp4 format to facilitate further analysis.

After the data collection process was completed, the researcher constructed two lecture corpora, with one containing fifteen literature lectures (literature corpus) and the other including fifteen computer science lectures (CS corpus). The thirty lectures were delivered by ten different lecturers, with eight male and two female lecturers. All of the ten lecturers are full professors. These lectures were delivered between 2007-2016 and the length of each lecture is about 50 minutes. The lectures included in the two lecture corpora are regular class meetings. That is, they are not delivered

specifically for online courses.

The literature corpus contains 95,457 words and the total word count of the CS corpus is 104,687 words. The average word count of a literature lecture is 6,363.8

77

words and that of a computer science lecture is 6,979.1 words. The total number of hours of the lectures in the literature corpus and the CS corpus is about 25 hours.

Table 2 presents the details of each of the lectures included in the two lecture corpora in terms of the names of lecturers, the names of the courses, the total numbers of lectures in each course, lecture topics, and the numbers of words of each lecture.

Table 2. Description of the corpora Computer science

Lecturer

Course name/

Total number of lectures

Topic of Lecture Word count

Demaine Introduction to

Lecture 5: Binary Search Trees, BST Sort

Shortest Paths Problem

7,076

Lecture 15: Abstract Data Types, Classes and Methods

8,480

78

Lecture 4: Stochastic Thinking 5,988 Lecture 8: Sampling and

Standard Error

5,685 Lecture 13: Classification 6,043

Winston Artificial Intelligence (23)

Lecture 2: Reasoning: Goal Trees and Problem Solving

Topic of Lecture Word Count

Dimock Hemingway, Fitzgerald, Faulkner (25)

Lecture 4: Fitzgerald's The Great Gatsby

6,141 Lecture 10: Hemingway's To

Have and Have Not

6,600 Lecture 22: Faulkner's Light in

August

6,632

Fry Introduction to Theory of Literature (26)

Lecture 4: Configurative Reading 7,272 Lecture 9: Linguistics and

Literature

6,873 Lecture 17: The Frankfurt School

of Critical Theory

6,510

Hammer Modern Poetry (25)

Lecture 2: Robert Frost 5,447

Lecture 10: T.S. Eliot 5,144

Lecture 19: Wallace Stevens 5,145

Hungerford Lecture 5: Vladimir

Nabokov, Lolita

7,072

79

The American Novel Since 1945 (26)

Lecture 13: Toni Morrison, The Bluest Eye

6,296 Lecture 22: Edward P. Jones, The

Known World

6,860

Rogers Milton (24) Lecture 5: Poetry and Marriage 6,815 Lecture 11: The Miltonic Simile 6,732 Lecture 23: Samson Agonistes 5,927 It should be noted that the differences in terms of the numbers of words among lectures are marked in some cases, despite the similar length of class time. It is very likely that lecturers’ individual idiosyncrasies did play a part in dictating the number of words produced in each lecture. For example, there are 9,155 words in Grimson’s first lecture while Hammer’s second lecture only contains 5,144 words. As Connor and Mauranen (1999) indicated, a move/step may be realized ranging from a few sentences to several paragraphs. It is very likely that different lecturers could produce discourses of very different lengths to fulfill the same move/step in lectures. The primary focus of the current study is on the occurrences of different types of move and step in the literature and computer science lectures. Therefore, the different numbers of words across lectures might only have a rather little influence on the current analysis. In terms of the identification of the frequent lexical bundles in different lecture phases, it should be noted that a longer lecture does not necessarily guarantee longer beginning, main body, or ending phases. Therefore, the different lengths across lectures in the corpora would be rather unlikely to influence the

80

identification of frequent lexical bundles as well.

While the two lecture corpora were constructed to represent the lecture genre, it must be admitted that it was not possible for them to achieve the desired degree of representativeness, which is often the case for small-scaled rhetorical move analyses (Lee, 2009; 2016). As this study is rather exploratory in nature, this study would focus on describing and generating a preliminary rhetorical move structure of lectures in their entirety as well as exploring potential disciplinary variations, in order to serve as a reference for larger-scaled rhetorical move analyses in the future.