Abstract
This study focuses on error analysis of first year medical students from the four medical schools at Mahidol University. A total of 44% of enrolled students participated, about 237. They were asked to translate sentences from Thai into English, translate a paragraph from Thai into English, and write an opinion paragraph in English on medical ethics. Data collected from the sentence-level translation, paragraph level translation, and opinion paragraph writing were analyzed to find the most frequent errors of these medical students by using the distribution of frequency. The top-ten errors of each medical school and of all medical schools were identified and the chi-square was used to find dependency among the three types of writing. The findings show that there is dependency between sentence level translation and paragraph level translation, and between paragraph level translation and opinion paragraph writing. The results will be of benefit for material development, especially in the provision of materials in the Language Learning Centre.

Keywords: error analysis, medical student, sentence level, paragraph level, translation.

Introduction
This study is one of four research projects derived from the need to collect data for preparing materials in the Language Learning Center and for developing a more student-centered teaching and learning process.

The four projects include:
1.  Should paragraph writing be taught to first year medical students at Mahidol University? (Phase I was disseminated at the Thai TESOL International Conference in Bangkok in January 2003 and both phases were published in Thai TESOL Bulletin in 2004. Honsa, Jr. and Clark (2004)
2.  Be more student-centered with the implementation of portfolio assessment and the Language Clinic (Disseminated at the CULI 5th International Conference in Bangkok in December 2003. Honsa, Jr. and Ratanapinyowong (2003)
3.  Medical students’ most frequent errors at Mahidol University. (Disseminated at the Thai TESOL International Conference in Khon Kaen in January 2004). Sattayatham and Honsa. (2004)
4.  Boosting students’ autonomy as well as writing accuracy and fluency through journal writing (Disseminated at the Asia TEFL International Conference in Seoul in November 2004 and published in Asia TEFL Journal in June 2005). Honsa, Jr. and Rtanapinyowong (2005)

   This research was conducted with first year medical students from the four medical schools at Mahidol University: namely, Siriraj Medical School, Ramathibodi Medical School, Bangkok Metropolitan Medical School, and Praboromchanok Medical School. These students are required to take two compulsory Introductory University English courses, the emphasis of which is on reading skills. As far as writing skills are concerned, these students are required to write an opinion paragraph on medical ethics based on a reading passage chosen from 30 passages from the Internet. Once they write a paragraph, 30 marks are given without grading their paragraphs. The fact that there is not enough writing practice, no grading, and 30 free marks given to all students who complete the paragraph, has prompted the researchers to make systemic amendments to the syllabus.

This study attempts to answer the following questions.

1. What are the most frequent errors of all medical students’ writing?
2. Is there any dependency among sentence level translation, paragraph-level translation and opinion-paragraph writing?

Literature Review
For this review of the literature, it is necessary to study the characteristics of the language errors, the pattern of the common errors, and the situation involved in the errors to help analyse errors systematically, as well as to understand, and find the causes of such errors.  In addition, the study of error analysis will improve the process of second language learning and develop more information for developing the second language acquisition theory.  In the past, it is believed that most language errors were caused by the transformation from one language to another language. Later, it was found that the influence of native language on the second language is quite minimal; that is, it affects only 3-25% of such errors. 
   Kitao and Kitao (2000) suggest that error analysis in language use can predict the learners’ type of errors which will be useful for developing teaching materials and selecting teaching methods. Moreover, the result of such analysis can be used as an indicator of the learners’ achievement. It can also be used for researching language acquisition and learners’ strategy in language learning. Dulay, Burt & Krashen (1982) say the result of error analysis can be used as an indicator of learning achievement and guidance for teaching.
   One of the difficulties of error analysis is how to define and scope the ‘errors’ as there are many definitions of errors. Corder (1981) differentiates errors from mistakes in the way that errors are systematic in nature being “errors of competence” which occur in the continuum of the learning process. They are the result of the learners’ transitional competence. On the other hand, mistakes are “errors of performance” which are not systematic.  
   Edge (1989) defines errors as forms that language users cannot correct by themselves even though they have been taught. James (1998) adds that language learners cannot correct their errors until they have additional knowledge on the topic. These errors occur in the course of the learner’s study because they haven’t acquired enough knowledge. Once they acquire additional knowledge, they will be able to correct their errors and the more errors the learners correct the more conscious of language they will become. James (1998) studied Error Taxonomies and classified errors into two types:

1.  Linguistic category classification
This type of taxonomy specifies errors in terms of linguistic categories and in terms of where the error is located in the overall system of the TL. First, it indicates at what level of language the error is located: in phonology, grammar, lexis, text or discourse and if it is at grammar level, what particular grammatical construction does it involve? Some possibilities they list are: the auxiliary system and passive sentence complements. Having established the level of the error, one next asks about its class. Given that it is a grammar error, does it involve the class of a noun, verb, adjective, adverb, preposition, conjunction, or determiner? which leads to the assignment of a rank to the error, in terms of where it lies on the hierarchy of units that constitute its level. Finally, we need to specify the grammatical system that the error affects such as tense, number, voice, countability transitivity.

2.   The surface structure taxonomy
This is the second type of descriptive taxonomy proposed by Dulay, Burt and Krashen (1982) who describe this taxonomy as being based on “the way surface structures are altered” (1982, p. 150). Errors can occur because of change in surface structure in specific and systematic ways (1982, p. 150). There are four ways in which learners “modify” target forms in specific and systematic ways.

      2.1    Omission
Learners in the early stages of learning tend to omit function words rather than content words. More advanced learners tend to be aware of their ignorance of content words and rather than omit one, they resort to compensatory strategies to express their idea (Kasper and Kellerman, 1997).
      2.2    Addition
This manifestation of error, according to Dulay, Burt and Krashen, is the ‘result of all-too-faithful use of certain rules’ (1982, p.156) and they suggest there are subtypes. First is regularization, which involves overlooking exceptions and spreading rules to domains where they do not apply, for example producing the incorrect “buyed” for “bought.” Second, is double marking, defined as ‘failure to delete certain items which are required in some linguistic constructions but not in others’. Here is an example: He doesn’t knows me.
      2.3    Misformation
Dulay, Burt and Krashen define misinformation as use of the wrong form of a structure or morpheme, and give examples like:
               I* seen her yesterday.
               He hurt* himself.
    2.4      Misordering
This category is relatively uncontroversial. The learners can select the right forms to use in the right context, but they arrange them in the wrong order, for instance, adverbials, interrogatives and adjectives, yielding errors as in:
*He every time come late home.
               *Tell me where did you go.
               *The words little
         As Dulay, Burt and Krashen observe, misordering is often the result of learners relying on carrying out ‘word for word translations of native language surface structure’ (1982, p.162) when producing written or spoken utterances in the TL.

      James (1998) classified errors into five patterns as follows:
1.   omission
2.   over inclusion
3.   misselection (use wrong words not wrong forms)
4.   misordering
5.   blends: (Blending arises when two alternative grammatical forms are combined to produce an ungrammatical blend.)

            Richards (1974) classifies errors, according to their causes, into two categories:
1.   Interlingual Errors: These are errors that are caused by mother tongue interference.
2.   Intralingual & Developmental Errors: These kinds of error occur during the learning process of the second language at a stage when they haven’t really acquired the knowledge.  In addition, the errors are also caused by the difficulty or the problem of language itself. For as Dulay and Burt (1974) say almost 90% of errors are intralingual errors.

James (1998) states that there are four causes of errors:
1.     Interlingual errors (Mother-tongue influence). These kinds of errors are influenced by the native languages which interfere with target language learning. Learners translate word by word idiomatic expressions, vocabulary and even the grammatical rules of the learners’ first language into the second language. In contrastive analysis, it is believed that the type of errors made by the learners of the target language can be predicted and their causes can be determined. In order to prevent and eliminate these errors, Richards (1974) has given the following figures: Between 3-25 per cent of all errors are errors of mother tongue influence and 75 per cent of errors are ‘non-contrastive’ errors.
2.     Intralingual errors: These types of error are caused by the target language (TL) itself. Apart from recourse to L1 transfer, the learners in ignorance of a TL form on any level and any class can do either of two things: either they can set about learning the needed item, engaging their learning strategies, or they can try to fill the gap by resorting to communication strategies. Learning strategies are used for code breaking while communication strategies are encoding and decoding strategies. Both types of strategy can be the source of error.

Errors caused by learning strategies include:
2.1     False analogy: Learners assume that the new item B behaves like A: they know that “boy” (A) has its plural “boys” and assume that “child” (B) behaves likewise, so pluralizes to “*childs.”
2.2     Misanalysis: Learners form a wrong hypothesis. An example of this strategy occurs in: they are carnivorous plants and *its (ü their) name comes from. The false concept in operation here is that its is the s pluralized form of it. A false concept is the result of the learners misanalysing the TL. 
2.3     Incomplete rule application: This is the converse of overgeneralization or one might call it undergeneralization as the learners do not use all the rules. They change or decrease the complicated rules to simpler rules as they aim at simplification rather than attempt to get the whole complex structure. An example is seen in the deviant order of subject and verb ‘be’ in: Nobody knew where* was Barbie (ü Barbie was).  The learners have applied only two components of the interrogative formation rule: they have selected and fronted a wh-element (rule components 1 and 2), but have omitted to invert the subject and verb.
2.4 Exploiting redundancy: This error occurs by carrying considerable redundancy.  This is shown throughout the system in the form of unnecessary morphology and double signalling.
2.5       Overlooking co-occurrence restrictions: This error is caused by overlooking the exceptional rules. An example of this is I would enjoy *to learn (ü learning) about America caused by ignorance of the fact that the verb enjoy should be followed by a gerund complement.
2.6 Hypercorrection (monitor overuse): This results from the learners’ over cautious and strict observance of the rules. One might say that the learners’ deliberate suppression of a potential L1 transfer, for fear of being wrong, is another form of hypercorrection: an example of this is the seventeen year*s old girl.
2.7 Overgeneralization or system-simplification: This error is caused by the misuse of words or grammatical rules. An example is the generalization of the relative pronoun that as in:
     Bill, *that had a great sense of unconventional morality…
     The learners use that to the exclusion of who which can not be used here.

3.  Communication strategy-based errors
3.1 Holistic strategies or approximation: The term ‘holistic’ refers to the learners’ assumption that if you can say X in the L2, then you must be able to say Y. Lacking the required form, it must be all right to use another near-equivalent L2 item which they have learnt. It takes on a number of forms, the first of which is to use a synonym; The second is to use an antonym or opposite: not happy for ü sad. The third is to coin a word. Until you be unconscious to lose your *sensities. ( ü senses)
3.2 Analytic strategies or circumlocution: Analytic strategies express the concept indirectly, by allusion rather than by direct reference. This kind of error comes from the students’ experience. James (1994) finds that the learners in the classroom used the L1 transfer strategy much more than the acquirers. (Acquirers are people who are self-directed learning, such as a taxi driver, a foreigner’s house keeper.) 

4.  Induced Errors: These errors are the result of being misled by the way in which the teachers give definitions, examples, explanations and arrange practice opportunities. In other words, the errors are caused mostly by the teaching and learning process as follows:
4.1       Materials-induced errors: Teaching materials with errors will make the learners confused, and they will make similar errors again and again.
4.2       Teacher-talk induced errors: This kind of error might be caused by both native or non-native teachers, if they do not provide models of the standard TL in class.

1. Exercise-based induced errors: The learners make errors while doing exercises on sentence combining, for example, the teacher feeds to the learners the raw ingredients: simple sentences that the learners must combine. Conditionals linked by if or unless are examples: 

I can’t afford a new car combined with I shall win the lottery. 
should yield
I can’t afford a new car unless I win the lottery.
but will also yield at times from at least one learner forms like
*Unless I can afford a new car I shall win the lottery. 
The likelihood is especially great when the students have been told
that unless is equivalent to if…not, which will suggest to them the possibility of replacing the negative element in can’t with unless.
4.4       Errors induced by pedagogical priorities: Learners’ achievement tends to match other teacher expectations of what they will achieve. Some teachers choose to prioritize one of the following: accuracy, fluency or the idiomatic in teaching communication, thus if fluency is considered as superior, accuracy would have lower priority or vice versa.
4.5       Look-up errors: There have been many learners’ dictionaries and grammar books in recent years, and these publications usually come with useful guidelines on how to look up aspects of the L2 about which one is in doubt. But, strangely, learners do not like to read such user-instruction, and as a result they frequently misuse these reference aids. In addition, the learners sometimes use the new words from the dictionary inaccurately or get incorrect references from the grammar books.
     There are many studies on error analysis because error analysis helps to improve the teaching and learning process. If learners’ errors and the causes of those errors are identified, errors can be corrected, though not all. Moreover, error analysis helps direct the focus of the teaching and learning process. For example, Angwatanakul (1980) finds that the most frequent errors of Thai learners are verb forms, articles and prepositions. Paster (1986) finds that most common interlingual errors are using the present simple tense in the place of the past simple tense, using the wrong verb form after modal, no inversion of auxiliary verb in questions, and using the wrong subject verb agreement. Michaeldes (1990) analyzes and puts errors into eight domains according to importance and frequency, such as wrong order of words, wrong tense, wrong use of articles and prepositions.    Cumming and Mellow (1990) study errors at the grammatical morpheme level and find that they can indicate second language learning ability. Polio (1997) studies second language writing, and error free writing, the use of the holistic scale, T-units and numbers of errors as criteria and finds that counting error numbers may be better for homogeneous population. Newmark and Reibel (1968) propose another approach to ignorance hypothesis which emphasizes avoidance strategy. Duskova (1969) also supports this same idea that learners who avoid using certain structures and have no errors in those errors may not know how to use those structures. Rujikiatkumjorn and Chiewkul (1989) analyzed errors of students at Khon Kaen University to find frequent errors made by students from each faculty and discovered that there is a dependency between errors and each faculty.

Methodology
This study consists of two parts. The first aim is to find the most frequent errors of medical students from the four medical schools at Mahidol University. The second aim is to identify the student’s dependency among three types of writing: including sentence level translation, paragraph level translation, and opinion paragraph writing.

Part I: Finding the most frequent errors of medical students

44% of all first year medical students in the year 2001 participated in the study (about 237 students). Three pieces of writing by each subject were collected at the end of the second semester. The details of the three pieces of writing are as follows:

1. Sentence level translation.
Each subject had to translate 32 sentences from Thai into English. In each sentence, we analyzed the errors in terms of grammar points, in other words, grammar points to be tested are determined in each sentence. There were 48 grammar points to be corrected in the sentence level translation. These grammar points are based on the previous study of error analysis of students at Khon Kaen University which won an award from National Research Institute of Thailand and on the research done by Chulalongkorn University researchers. Also, these grammar points cover most essential grammar points that might affect Thai students in reading and writing. The 48 grammar points are presented below.

Errors from sentence level (48 test points)

1. Order of adj.
2. There (is, are)
3. Subject-verb agreement
4. Direct-indirect object
5. Verbs of feeling
6. Past tense
7. Present perfect
8. Reported speech
9. Passive voice
10. Question tag
11. Question word (whom)
12. Relative clause (who)
13. Imperative
14. Comparative (the same as)
15. Comparative (adj.)
16. Comparative (adv.)
17. Comparative (the….er, the….er)
18. Superlative
19. Connector (as)
20. How much + n.
21. Conditional sentence (Unreal present)
22. Conditional sentence (Unreal past)
23. Suggest / recommend + V1
24. Relative pronoun (possessive / whose)
25. Possessive (of, s’)
26. Subordinate clause
27. Present participle
28. Past simple + past continuous
29. Expression
30. Infinitive (purpose)
31. Have something-done
32. Articles
33. Modal (ability-can)
34. Modal (probability = may, might)
35. Wh-clause as subject (what she wants)
36. Reflexive pronoun

       (themselves/yourself)

1. Possessive adj.
2. Connector (when)
3. Wrong choice of vocabulary
4. Punctuation
5. Capitalization
6. Spelling mistake
7. Preposition
8. Wrong use of pronoun
9. Omission of preposition
11. verb
12. noun

  Wrong order of part of speech

 
• Omission of
• No subject

2. Paragraph level translation 
Each subject had to translate a paragraph of about 80 words from Thai into English. There were 24 grammar points to be tested which covered the most essential grammar points that might affect Thai students in paragraph translation. The 24 grammar points are presented below.

 Errors from paragraph level: (24 test points)

1. Connector showing contrast (but, even though)
2. Describing people (height, face, age, hair)
3. Time sequence words (First, next…)
4. How to do something
5. Punctuation (…but She…)
6. Capitalization (.first…)
7. Verb + ing
8. Subj – verb agreement
9. Article
10. Spelling mistake
11. Be + adj.
12. Modal + V1
13. Wrong choice of words
Relative pronoun (possessive: whose)
• Subordinate clause (that…)
• Uncountable noun + plural ending
• There be
• Preposition
• Omission of main verb
• Past tense
• Infinitive
• Mother tongue interference (mistakes influenced by Thai sentence structure)
• Reported speech
• No subject

3. Opinion paragraph writing
Each subject had to write an opinion paragraph based on one reading passage from the Internet which was an external reading assignment in class. The medical students were asked to read three medical ethics passages among thirty selected passages from the Internet. Each student had to choose one of these passages and express their opinion about medical ethics in one paragraph. There are 28 test points that are usually found in paragraphs written by most Thai students.

Errors from opinion paragraph Writing: (28 test points)

*Tense (Non-parallel form of verb)
• Wrong use of verb to be
• Spelling mistake
• Wrong use of verb
• Article
• Omission of subject
• Tense (present continuous / present perfect)
• Subject – Verb agreement
• Direct translation
• Conditional sentence (unreal present, unreal past)
• Connector
• Wrong choice of vocabulary
• Wrong plural form
• Infinitive (purpose)
• Capitalization
• Punctuation
• Wrong use of pronoun
• Fragment of sentence (incomplete sentences punctuated as complete sentences)
• Wrong order of adverb
• Passive voice
• Possessive (of, is)
• Run-on sentence (two complete sentence joined by a comma)
• Omission of verb
• Relative pronoun (whose / who)
• Wrong form of noun
• Complex sentence without conjunction
• Comparative & superlative
• Question tag (wrong use of “tag”)

After analyzing the errors from sentence level (48 test points), paragraph level translation (24 test points), and opinion paragraph writing (28 test points), the researchers used the percentage and distribution of frequency to get the most frequent errors. The top ten errors made by each group of medical students were selected. Then the top ten errors made by the four medical schools were put in order from the most frequent errors to the least frequent errors. We got this data by multiplying the total number of errors of each item by 100 and dividing these by the total number of subjects:

= total number of errors of each item × 100
       
total number of subjects (237)

Part II: Identifying the dependency among sentence level translation, paragraph level translation, and opinion paragraph writing

The researchers applied Pearson’s Chi-square to the data collected to check the dependency among three types of errors at the significance level of 0.05. The three pairs of data to be applied with Chi-square formula are:
1.  Errors from sentence level translation and errors from paragraph level translation
2.  Errors from paragraph level translation and errors from opinion paragraph writing
3.  Errors from sentence level translation and errors from opinion paragraph writing

Findings
The researchers obtained the medical students’ frequency of errors and the dependency among three types of writing.

Part I: The medical students’ frequency of errors
There are three parts to the medical students’ frequency of errors compilations: sentence level translation, paragraph level translation and, opinion paragraph writing.

1.  Sentence level translation.
The top ten errors made by the four medical schools are put in order from the most frequent errors to the least frequent errors as follows:

Order	Errors	%
1 2 3 4 5a 5b 6 7 8a 8b 9 10	Conditional Sentence (Unreal past) Articles Question tag Past tense Connector (as) Have something done Present Participle Wh – clause as subject Suggest/recommend+V1 Question word (whom) Comparative (adv) Past simple + past continuous	85.23% 78.48% 75.94% 66.66% 62.86% 62.02% 58.64% 56.96% 56.11% 54.85%

The findings show that the use of tense-sequence troubles the students most. This includes the overgeneralization of tense-sequence in the conditional sentence showing past unreal as in item 22 “* If John is alive, he is 80 next year.” and the wrong tense-sequence in coordinate structure as in item 28 “* When I arrived home, it has still raining.” These errors are mostly due to the overgeneralization of the target-language restrictions on tense-sequence as James (1998) mentioned. Other prominent errors are the articles and the question tag. We do not have articles and question tag in Thai system, so the students tend to omit them or use them wrongly.

2. Paragraph Level translation

In this part, the top ten errors made by the four medical schools were put in order of the most frequent errors to the least frequent errors as follows:

order	errors	Percentage
1 2a 2b 2c 3 4 5 6 7 8 9 10	Wrong choice of words Connectors showing contrast How to do something Describing people (height, face, age, hair) Reported speech Omission of main verb Articles Spelling Mistake Punctuation Subordinate clause There (is, are) Infinitive	57.8% 53.5%   51.6% 36.28% 34.59% 28.27% 17.29% 16.87% 15.18% 14.34%

The findings show that the area which causes difficulty is the vocabulary. The students cannot choose the right word to fit the content as in item 13 “*Although she’s aged, she still looks beautiful.” These types of error are interlingual errors which are caused by mother-tongue interference in the form of translating Thai into English, word by word. Another error is the students’ failure to use the connector showing contrast as in item 1 “*Although she is old, but she is beautiful.” The students have the tendency to keep rules of their native language, even though they had been taught the correct forms of the target language. They still make the same errors. In Thai, we use both “although” and “but” in the contrast statements. The failure to describe people in terms of height, face, age and hair is another error which is due to ignorance of rule restriction as in item 2”* Her face is egg. She tall 160 cm. She is long hair. My teacher is a 55 woman”. Reported speech is another prominent error as in item 23 “*You must know how taste each food.” The students do not follow all the rules. They select a wh-element, but omit to invert the subject and verb. They make the incomplete rule application.

3.  Opinion Paragraph Writing
In this part, top ten errors made by the four medical schools were put in order from the most frequent to the least frequent as follows:

order	errors	Percentage
1 2 3 4  5a  5b 6 7 8 9 10	Wrong choice of vocabulary Articles Wrong plural form Wrong use of verb “to be” Conditional sentence Punctuation Connector Fragment Subject-verb-agreement Spelling mistake Omission of subject	64.90% 62.40% 38.90% 36.29% 34.18% 33.33% 32.40% 30.38% 29.10% 21.10%

The findings show that the error that figured prominently is the wrong choice of vocabulary as in item 12 “*Their daughter recover from leukemia and this baby is harmless.” The students use the word inappropriately and their English vocabulary does not meet the standard. Also, mother-tongue interference is a likely cause of this error. The articles are still the prominent errors in the opinion paragraph writing.  This is also a case of omission because there are no “articles” in the Thai system. Another prominent error is the wrong use of verb to be as in item 2 “*Genetic screening doesn’t correct.” This error is not a case of mother-tongue interference, but it can be attributed to intralingual errors. The students misuse the verb to do instead of using the verb to be in front of adjectives.

Part II:   The finding of dependency among sentence level translation, paragraph level translation and opinion paragraph writing
The researchers applied Pearson’s Chi-square to the data collected to check the dependency among the three sets of errors at the significance level of 0.05. 

1.   Errors from sentence level translation and errors from paragraph level translation. (See Table 1)

	Cases
	Valid		Missing		Total
	N	Percent	N	Percent	N	Percent
Sentence & Paragraph	237	100.0%	0	0%	237	100.0%

Sentence & Paragraph Cross tabulation Count

	Paragraph			Total
	F	G	V
Sentence   F	-	1	-	1
G	2	25	124	151
V	1	-	84	85
Total	3	26	208	237

Chi-Square Tests

	Value	df	Asymp. Sig. (2-sided)
Pearson Chi-Square	24.477a	4	.000
Likelihood Ratio	28.466	4	.000
N of Valid Cases	237

a. 5 cells (55.6%) have expected count less than 5. The minimum expected count is .01
Note: F = fair, G = good, V = Very good

   By using the Chi-Square test, we conclude that there is a dependency between sentence level translation and paragraph level translation at 0.05 level of significance. This means that if the students can translate from Thai into English very well at sentence level, they can also translate into English very well at paragraph level as well.

2.   Errors from paragraph level translation and errors from opinion paragraph writing. (see Table 2)

Table 2: Case Processing Summary

	Cases
	Valid		Missing		Total
	N	Percent	N	Percent	N	Percent
Paragraph & Essay	237	100.0%	0	0%	237	100.0%

Paragraph & Essay Cross tabulation Count

	Essay			Total
	F	G	V
Paragraph    F	2	-	1	3
G	-	2	24	26
V	22	16	170	208
Total	24	18	195	237

Chi-Square Tests

	Value	df	Asymp. Sig. (2-sided)
Pearson Chi-Square	13.575a	4	.009
Likelihood Ratio	11.364	4	.023
N of Valid Cases	237

a. 5 cells (55.6%) have expected count less than 5. The minimum expected count is .23

   By using the Chi-Square test, we concluded that there is a dependency between paragraph level translation and opinion paragraph writing at 0.05 level of significance. This means that if the students can translate into English very well at paragraph level, they can write a good opinion paragraph as well.

3.   Errors from sentence level translation and errors from opinion paragraph writing (See Table 3)

Table 3:Case Processing Summary

	Cases
	Valid		Missing		Total
	N	Percent	N	Percent	N	Percent
Sentence & Essay	237	100.0%	0	0%	237	100.0%

Sentence & opinion paragraph writing cross tabulation count

	Paragraph			Total
	F	G	V
Sentence   F	-	-	1	1
G	13	14	124	151
V	11	4	70	85
Total	24	18	195	237

Chi-Square Tests

	Value	df	Asymp. Sig. (2-sided)
Pearson Chi-Square	2.717a	4	.606
Likelihood Ratio	2.967	4	.563
N of Valid Cases	237

a. 3 cells (33.3%) have expected count less than 5. The minimum expected count is .08

By using the Chi-Square test, we concluded that the performance in sentence level translation and that in opinion paragraph writing are independent at 0.05 significance level. This means that if the students can translate from Thai into English very well at sentence level, they still cannot write a good essay.

Conclusions
It was found that the most frequent errors from this data were on the syntactic and lexical levels with inadequate lexical and syntaxtic knowledge leading to the errors of overgeneralization, incomplete rule application, omission, and building of false concepts. Within these errors, we detected mother-tongue interference. However, some linguistic items, such as articles, tense, and verb forms appeared to be the source of frequent errors. The errors may result from inadequate learning as well as the complexity of English structure which our native language structure does not have.
We might conclude that the errors made by the students are both from the interlanguage and mother tongue interference. The errors caused by mother tongue interference are in a small proportion. For the dependency test, it was found as follows:
1. There is a dependency between sentence level translation and paragraph level translation at 0.05 level of significance.
2. There is a dependency between paragraph level translation and opinion paragraph writing at 0.05 level of significance.
3. The performance in sentence level translation and the performance in opinion paragraph writing are independent at 0.05 significance level. It may be that the student might not know how to use transitional words properly to convey a coherent idea when writing their own paragraph. In contrast to paragraph translation, the students do not have to think about coherence. The transitional words and the coherence already exist in the paragraph. The students just translate from one language to another.

Suggestions
Upon reviewing students’ essay writing, the researchers discovered that most of the students still make other kinds of serious errors. For example, they do not use a topic sentence with clear transition words in writing an opinion paragraph when they are not told and guided to do so. There are a lot of fragmented sentences and run-on sentences in their paragraph writing. They do not know how to use punctuation correctly. Only a small percentage of students have organizational skills. Therefore, the students should be taught to write topic sentences, supporting details, transition words and concluding sentences when writing a paragraph.
   The afore-mentioned problems occur because writing is not incorporated into the two Introductory University English Courses for medical students at Mahidol University, these students do not have a chance to practice writing in class. At the end of the semester these students have to write an opinion paragraph on medical ethics without having been taught how to do it. As writing experts (Weigle, 2002; Hamp-Lyons and Cordon, 2000) say writing skills and assessment should be a continuous process. To equip these medical students with a more efficient approach to writing, a writing lab should be established and fully and efficiently utilized to facilitate and supplement the writing process and assessment. The research finding can be used as criteria for material development both in the writing lab and other corners in the Language Learning Centre. It would enable us to isolate the errors these students make, then we could provide the proper teaching materials and method to correct their errors. It would help to improve the English ability of future Thai doctors to communicate better with fewer errors. Most medical students’ English, which is at intermediate level, can be upgraded to advanced level if their errors are reduced. The findings will be useful for providing materials in the language learning center where these medical students can equip themselves with higher language proficiency even after they leave the classroom. We hope that these doctors will be well prepared to communicate with foreign patients who come to Thailand for medical services at a lower medical expenses compared to their home countries.
   In a broader context, the research findings, about dependency, at significant level, between paragraph level translation and paragraph writing, suggest that if students practice frequent translation at a paragraph level, they will eventually be able to write a good paragraph. Thus, more research is needed to find empirical evidence to support these findings. If these findings can be generalized, it will greatly benefit the EFL teaching and learning process in the near future.

References
Angwatanakul, S. (1980). Relationships between communicative competence and learning achievement in the English language. Chulalongkorn Educational Review, Chulalongkorn University.

Corder, S.P. (1981). Error analysis and interlanguage. Oxford: Oxford University Press.

Cumming, A., & Mellow, D. (1996). An investigation into the validity of written indicators of second language proficiency. In A. Cumming and R. Berwick (Eds.), Validation in language testing (pp. 72-93) Clevedon: Multilingual Matters.

Dulay, H.C. and Burt, M.K. (1974). Errors and strategies in child second language acquisition. TESOL Quarterly, 3(2), 129-136

Dulay, H., Burt, M. & Krashen, S.D. (1982). Language two. MA. Rowley: Newbury House.

Duskova, L. (1969). One sources of errors in foreign language learning. IRAL, 7(1), 11-36.

Edge, J. (1989). Mistakes and correction. London: Longman
Hamp-Lyons, L. & Condon, W. (2000). Assessing the portfolio. New Jersey: Hampton Press.

James, C. (1994). Don’t shoot my dodo: On the resilience of contrastive and error analysis.  International Review of Applied Linguistics, 32(3), 179-200.

James, C. (1998). Errors in language learning and use: Exploring error analysis. London: Longman.

Kasper, G. & Kellerman, E. (1997). Introduction, in G. Kasper and E. Kellerman (Eds.), Communication strategies: Psycholinguistic and sociolinguistic perspectives (pp. 1-13). London: Addison Wesley Longman.

Kitao, K. & Kitao, S. (2000). Selecting and developing teaching/learning materials. http://www.Atech.ac.jp/-iteslj/Articles/Kitao-Material.Html. April 20.

Michaeldes N. N. (1990). Error analysis: An aid to teaching. English Teaching Forum, 28(4), 28-30.

Newmark, L. & Reibel, D.A. (1968) Necessity and sufficiency in language learning. IRAL, 6(3), 145-164.

Pastor, A.C. (1986). The ILokano ESL learner: A study of interlanguage. M.A. Thesis: National University of Singapore.

Polio, C.G. (1997). Measures of linguistic accuracy in second language writing research.  Language Learning, 47(1), 101-143.

Richards, J.C (ed.) (1974). Error analysis: Perspectives on second language acquisition. London: Longman.

Rujikiatgumjorn, S. and Chiewkul, S. (1989). Error analysis of English language performance of KKU freshman students. PASAA, 19(1), 93-94.

Weigle, S.C. (2002). Assessing writing. Cambridge: Cambridge University Press.

Appendix (See PDF file)