作为过程写作法的重要组成部分，同伴反馈早已获得了人们的注意。有了这种反馈方式，教学会变得更加以学生为中心，学习者的学习主动性会提高，学习者会得到重视。然而在传统的英语课堂中老师只是采纳教师反馈这种单一的反馈方式，而且以往的同伴反馈研究多在英语水平较高的学习者中进行,并且每个学生都参与评价。笔者带着探索的精神，在佛山南海中学分校高三文科平行班的英语写作教学中首次尝试了只有部分优秀学生充当评价者的“同伴中合作评价作文”的方式，具体研究了： 1）在英语写作教学中，学生评价者和被评价者对使用同伴中合作反馈的态度； 2）学生评分和老师评分的一致程度； 3）同伴中合作评价作文这种方式对学生的写作带来的影响。 参加本研究的56名学生在经历这种评价方式前后分别填写了一份调查问卷，8名同学在暑假接受了访谈。学生和老师的评分被收集起来并做了对比,作文中的内容结构错误也被分类并统计个数和百分比。分析的结果显示： 一、从问卷来看，学生批改人和被批改人对在写作教学中使用同伴互评作文方式的态度总体上是肯定的，批改人则明显更喜欢教师评价，并对自己单独看出错误的能力不是特别自信。在经历了一学期的同伴互评后，这些态度没有显著变化。但个别访谈显示，大多数人喜欢参与同伴互评。 二、在首次和末次评分中，学生和老师的单项评分和总评分的一致性不是很高，只有第二次评分的一致性较好。 三、 同伴互评对学生写作产生的影响主要有：（1）学生能找出大约一半的内容结构错误(46.03%)，其中比较容易找出的是要点问题(76.29%)、要点分配问题(62.96%)和主题问题(50%)。学生能自行更正其中约一半的错误（54.3%）；（2）学生在写作要点上犯错最多，学生反馈中，被正确采纳比例最小的、以及被错误采纳比例最高的是对连贯性问题的反馈，不被采纳的比例最高的是对要点错误的反馈；（3）除了写作清晰性方面有所提高外，学生的内容结构错误整体上并未减少，但是学生的一二稿分数仍有显著差异。（4）学生评价者在连贯性写作上提高明显，老师评价的一稿二稿平均分也在上升。 研究结果表明，同伴反馈模式不是特别适合英语水平比较低的学生群体。笔者建议教师多注意培养优生写好连贯性，这样容易见到成效；而对于大部分学生，写好要点、突出要点、增强连贯性是他们的难点，可以尝试作为重点击破。笔者在最后还提出了一些改进建议，以便在今后的英语写作教学中更好地运用此评价模式，从而收到更好的教学效果。

As an essential component in the process-oriented teaching of writing, peer feedback has long gained people’s attention. With the use of this approach, teaching becomes more learner-centered, learning autonomy will be developed, and the role of the learner will be emphasized. However, in the traditional English writing classrooms, only teacher assessment is employed, and most peer assessment is conducted among relatively high-proficient learners with every participant playing the role of assessor. To testify its feasibility among low-proficient learners, the present research experiments the adjusted model of peer assessment, or the model called collaborative assessment among peers (CA) in this dissertation, in English writing class for senior three students in the city of Foshan, with a purpose of addressing the following research questions: (1) What are the students’ attitudes towards the use of CA in EFL writing instruction? (2) To what degree does students’ marking agree with that of the teacher’s? (3) What are the possible effects of CA on students’ writing? Questionnaires are distributed to 56 participants before and after the experiment. Student marks and teacher marks are collected and different types of mistakes are classified and calculated during the research and analyzed later. Interviews are conducted to 8 students who are randomly selected during the summer holiday. The results show that: (1) According to the statistics, students hold an overall positive attitude towards the use of CA in their writing course both prior to and after the practice of CA. The interviews also reveal that the majority are willing to participate in the assessing activities. However, the student assessors are not very confident that they can locate the problems independently and they prefer TA to CA. The attitudes of both student assessors and student writers towards CA do not have significant change prior to and after the experiment. (2) There is poor agreement between specific and total marks awarded by the teacher and students to the same compositions in the first and third assignment. Only in the second assignment they show high agreement in their marks for almost all items. (3) Effects of CA on students’ writing have been found in the following aspects: a) student assessors can locate about half of the mistakes concerning content and organization (46.03%), and students are able to correct 54.3% of them. It is easier for them to find out problems in the aspects of Main points（76.29%）, Distribution of main points (62.96%)and Theme(50%). b) the writing of Main Point is the area where students have made most mistakes; most incorrect changes and least correct changes are made in the aspect of Coherence; the aspect that enjoys the highest percentage of not being implemented is Main Points. c) as a whole, the students do not make fewer mistakes concerning content and organization. But the number of intelligibility mistakes has reduced a bit, and significant difference is found in the marks for the first and final drafts. d) the student assessors have improved their ability of locating Coherence problems. They have also made fewer mistakes in the same aspect, and the mean values of total marks awarded by the teacher for the three assignments keep going up. The research findings show that CA is not very suitable for low-proficient students. If the teacher wants their advanced learners to make more rapid and obvious progress, train them to achieve better coherence. As for the ordinary students, the writing of main points and achieving coherence are difficult for them, so more emphasis could be laid here. Some suggestions are put forward by the investigator in the last chapter to help the model to be better applied into future English Writing courses.