Coherence and Cohesion in English Discourse

3 Corpus and methodology
The present investigation, which uses both qualitative and quantitative 
methods, is based 
on the analysis of five texts taken from the 
Michigan Corpus 
of Academic Spoken English
(MICASE). As the name itself suggests, this corpus 

---

65
represents spoken discourse used in academic settings recorded at the University of 
Michigan, USA. The texts are taken from three different 
speech event categories, 
namely office hours, discussion sections, and study groups. Each category is 
represented in the corpus by approximately 15,000 words: office hours (two 
texts amounting to 16,176 words together) concern instructions given mostly 
by a graduate student on some specific topic or project; discussion sections (two 
texts of 15,542 words together) are additional sections of a lecture designed for 
maximum student participation; study groups (one longer text of 15,483 words) 
are informal student-led study groups. The total extent of text under examination 
is exactly 47,201 words. In order to obtain comparable results, figures are given 
both in actual numbers and in normalized frequency rates.
All the interlocutors in the texts are native speakers of American English, for 
the most part university teachers and graduate or undergraduate students. Since 
the main objective of the study is not a detailed comparison of the three different 
speech event categories, but in particular a consideration of possible pragmatic 
functions of the marker 
I mean
in academic 
spoken discourse, differences 
between the texts, such as those in the tenor of discourse, are not considered 
relevant here. 
As regards the methods used in the analysis, the texts were examined 
manually in order to obtain both qualitative and quantitative results although it 
should be noted that certain expressions, such as the clausal form 
I mean
, could 
be searched for using a customized search engine developed by the University of 
Michigan Digital Library. The manual examination was indispensable, however, 
since it was necessary to distinguish the marker 
I mean
from the matrix clause 
I 
mean
introducing a nominal object clause; this was especially difficult when the 
clausal form 
I mean
occurred at the beginning of the sentence structure.
With regard to the size of the corpus (approximately 47,000 words only) 
and its representativeness, it must be stated that, similarly to the corpus and 
results discussed in Chapter Two of this volume, this corpus represents a small 
specialized corpus. In agreement with Flowerdew (2004: 18), this is considered 
here more appropriate than large corpora for an analysis of one particular 
language feature such as the marker 
I mean
and its possible pragmatic functions 
in academic spoken discourse.