population  which  is  intended  to  be  investigated.  In  this  research  the  subjects  being investigated are 30 students as the samples from the population.
The sampling  technique  used in  this research  is  random  sampling technique. Random  sampling  includes  choosing  subjects  from  a  population  through
unpredictable means. In its simplest form, all subjects have an equal chance of being selected  out  of  the  population  being  researched.  According  to  Sumanto  1990:  23
random  sampling  is  a  process  of  choosing  sample  so  that  all  individuals  in  a population  have  the  same  chance  and  the  same  opportunity  to  be  chosen  for  the
sample.  Therefore,  in  this  research,  the  researcher  chose  30  samples  from  120 population  randomly  without  looking  any  specific  criteria  for  every  samples  which
are in the form of text. Random sampling was done by these following steps:
1. List all the names of the students in population. It consists of 120 students.
2. Write the names in a little piece of paper and rolled the paper
3. Pick 30 roll of papers which contain students’ names randomly as the data
of the research.
E. Data Collection
A composition test is used as the instrument of data collection. It is the most direct and suitable way  of measuring participants’ writing ability. The composition
test requires the students to organize their own essays and to express their own ideas
and words so that there would be samples which is intended to measure.
The  data  were collected  from  the  students’  argumentative  essays  as  a  final
project in the fourth semester second year in August 2012. They were asked to write argumentative  essays  of  some  topics  they  can  choose.  Some  topics they  can  choose
are  Abortion,  Public  Transportation,  Entrepreneurship  and  Smoking.  The  test  was used as an instrument of the research. The test was designed to give their opinion and
argument about the topic they choose.
F. Data Analysis.
According to Sudaryanto 1993, there are two kinds of analyzing methods in research;  they  are  the distributional method and the referential  identity method. The
distributional method was used to analyze the Error Analysis on the use of cohesive devices the students’ text. Elements of the related language themselves are the main
determination  tool  in  analyzing  the  language  such  as  words,  syntactic  functions, clauses,  sentences,  and  so  on  Sudaryanto,  1993:16.  In  this  study,  the  element  that
the researcher worked with is the sentences. The data in this study were analyzed through two procedures: identifying and
counting the number of errors in using cohesive devices and evaluating the reason of the errors. For the first phase of data analysis, Halliday and Hasans 1976 cohesion
taxonomy  was  used  to  examine  the  kinds  of  cohesive  devices  used  in students’
argumentative texts. Then, it will be analyzed the error of cohesive  devices using in the argumentative text.
G. Procedure of Analysis
The  analyses  of  the  data  in  this  study  were  conducted  by  the  following procedures.  First,  the researcher read  each text  several  times  carefully paragraph by
paragraph,  and  numbered  the  paragraphs  and  each  sentence  to  ease  the  way  of analysis. Second, she underlined all related grammatical and lexical item of each text
that meets the criteria of errors in using cohesive devices on the text. They are errors in reference, substitution, ellipsis, conjunction, and lexical cohesion.
The researcher counted the frequency of error in using cohesive devices. The errors  are  analyzed based on the source of errors, whether  the inter-lingual  or intra-
lingual source of errors. There are three steps are used in analyzing the errors:
1. Data collection
2. Classification of errors into errors types
3. A statement of error frequency
After analyzing the text by using the three stages of analysis it will be known the  errors  types  and  source  of  errors  that  mo
stly  appear  in  the  students’  text.  The second step i.e. identification of error is not conducted in this research. It is based on
the  view  of  Dulay  et  al.  1982:139  that  an  error  is  any  deviation  from  a  selected
norm  of  language  performance,  no  matter  the  characteristics  or  causes  of  deviation might be. Therefore for practical reasons, this study does not distinguish errors from
mistake.  The  distinction  between  errors  and  mistake  is  not  crucial  here  because  the written  language  is  more  deliberate,  in  the  sense  that  students  spend  a  fair  time  in
choosing  or  producing  a  particular  construction,  this  would  allow  for  a  more restricted definition of the concept of error and would exclude lapses and mistakes as
described in most of the literature of EA.
CHAPTER IV
FINDINGS AND DISCUSSIONS
This chapter presents the description of error analysis on the use of cohesive devices which occur in students’ argumentative text. First of all, the analyses focus
on the error analysis of each cohesive device. Then, it is continued by presenting the analysis the sources of error in  using cohesive devices done by the students  in  their
texts whether they are caused by inter-lingual or intra-lingual sources and how related cohesive devices errors can influence the coherence of the argumentative texts.
A. Findings