The teaching context:
Experimental Animal Behaviour is a third-year undergraduate subject with an enrolment of 60 students. In this subject, students learn about the intellectual, organisational and logistical challenges associated with conducting research in animal behaviour. Students form teams of 4-5 students supervised by a graduate student or member of academic staff. After selecting a research topic, the team formulates a question, designs an experiment or sampling regime to test their idea, collects and analyses the necessary data, and finally reports on its findings in oral (group presentation) and written (individual scientific paper) formats. The projects are organised entirely in the students' own time, but they also attend several workshops/tutorials that address issues with collaboration (group dynamics) and offer advice on writing of scientific reports, oral presentations, and the peer review process.

The reasons for change in assessment method
The main piece of assessment in this subject was the scientific report that students submitted at the end of the semester. This meant that assessment was quite end-loaded, and furthermore that the 'feedback 'we provided (commentary and grades on the report) was not received until the end of the semester. Because there was no further opportunity for students to improve on the assignment, there was also little motivation for them to reflect on, and thus learn from this feedback. Not surprisingly, 'Quality of Teaching' survey scores identified feedback as an area of low student satisfaction.

We had two goals in changing our assessment practice. First, we wanted to provide students with feedback on their work that was useful in the sense that it promoted a genuinely reflective cycle of learning. Second, we wanted to expose students to the process of peer review, which lies at the heart of quality control of scientific publication. Although the majority of undergraduate students do not pursue scientific careers, the ability to critique constructively the work of others, and interpret and reflect on critical feedback are clearly generic skills that are valuable in many work environments.

The implementation of the change
We felt that implementing a process of peer review might allow us to achieve both of our goals simultaneously. However, when we first considered the implementation of a peer review process within the subject, it rapidly became clear that the task of administering this process would be extremely complex and time-consuming. Technological support of the process seemed to be the obvious solution. We found, however, that while web-based submission and review was increasingly employed by the editors of scientific journals, there was no readily available software to manage peer review within an educational environment.

We therefore developed our own software, which allows students to submit a draft version of their main piece of assessment (a scientific report) and receive reviews on this report from two other students and a supervisor, all within an anonymous double-blind framework. The review form includes structured yes/no questions that query important aspects of each section of the scientific report (e.g "Was enough detail presented to allow the methods to be repeated?") as well as open dialog boxes in which reviewers elaborate on their views. Students are encouraged to improve their report by incorporating this feedback into their final draft before they submit it for assessment. This process is formalised by requiring the students to submit a 'letter to the editor' with their final report, in which they explain how they have dealt with the reviewers' suggestions.
Our software was written in Java and is named APRES (Anonymous Peer Review and Evaluation System). Students access the system online via a simple interface and password-based login. The main menu of the programme offers options for uploading of reports for review, an online review form, and access to reports for reviewing and returned reviews. The report may be in one of three common formats: .doc, .rtf or .pdf. Reports and reviews are identified by student number, with dummy id numbers assigned to supervisors. APRES assigns each submitted project report to other students using a simple set of criteria. These criteria are that the reviewer may not be a student from the same group, and that all reviewers must receive projects for review from two different groups. Once submitted online, anonymous reports for review automatically become available to the assigned reviewers. Reviewers may download and print reports, and complete the online review, which is then automatically mailed to the report author.

Submission of the draft report is a hurdle requirement, but this draft is not graded. All students both review projects and receive reviews and thus learn the art of providing and responding to constructive critical feedback. To ensure that students take the reviewing process seriously, the quality of reviews is assessed and students are graded in the degree to which they provided 'useful' or reflective comments to their peers. The 'letter to the editor' is also graded, and provides an illuminating record of the degree to which the student has engaged with the reviews.
Although implementation of the peer review system has worked very smoothly overall, we identified a number of issues that will need to be resolved:

1. Some students fail to provide a timely review. The main problem with this non-compliance is that it disadvantages the author of the report. To date all students have complied because we have stipulated that failure to do otherwise leads to an automatic failure in the subject. Nevertheless, it is likely that in the future, some students may be unable to provide a review because of circumstances beyond their control. At present, the only solution to this problem is for the subject coordinator to complete the missing reviews. We are uncertain of whether there is a better solution.

2. Not surprisingly, there was variation in the quality of reviews provided by students, which ultimately reflects student aptitude and interest. We have attempted to limit this variation in several ways. First, the reviews themselves are assessable items worth 20% of the overall mark for the subject, thus providing some incentive for the students to provide a considered assessment. Second, the reviews are structured, forcing the students to address a range of issues. Third, each author receives three reviews, one of which is prepared by the supervisor (who will have had at least some prior experience with this activity). Nevertheless, we recognise that there is potential for dissatisfaction among high-achieving students that receive superficial or low-quality reviews. Although one might argue (perhaps facetiously) that this provides them with an introduction to the reality of publishing, it is reassuring that thus far, our survey results suggest that this problem has not arisen.

The effects of the change
Evaluations suggest that this initiative has been highly successful. First, there has been a dramatic and sustained rise in mean Quality of Teaching (QoT) scores evaluating feedback in this subject (before implementation: 2.8; after implementation: 4.2, 4.0, 4.3). Second, written evaluations from students indicated that they were enthusiastic about the approach. In the most recent survey (2005), 98% of students strongly agreed or agreed that "The reviews helped me improve my report" (n=49). Many nominated the peer review system as a highlight of the subject ("The review system is excellent; no other subject I have done gives this opportunity" ... "I learned a lot, reviewing was great!" ... "Web-based reviews were extremely helpful" ... "reviews were fantastic"). Finally, the mean grade for projects, in the year after the implementation of the program, showed a statistically significant improvement (p=0.012). Thus, peer feedback has helped students to effect genuine improvement in the quality of their project reports.

Many students comment that completing a structured, comprehensive review of other work is a revealing way of helping them to identify areas requiring attention in their own report. As Ramsden (1992) notes: "Giving feedback on another student's work, or being required to determine and defend one's own, not only increases a student's sense of responsibility and control over the subject matter, it often reveals the extent of one's misunderstandings more vividly than any other method".

Our procedure forces students to prepare successive drafts of their work and discover how they can improve their written work. Additionally, they gain an appreciation for the fact that reviewers can differ widely in opinion and competence (or can have surprisingly similar, independently derived opinions) and learn how to distinguish between helpful and unhelpful forms of feedback. The exercise is effective both as a taste of the scientific process, but also as the introduction of a key generic skill required in almost any career. Most importantly, it allows students every opportunity to submit a piece of work for assessment that genuinely reflects their capabilities and talents.

We believe that the separation of feedback from grading that is a consequence of our approach is beneficial for both teachers and students. It has improved our ability as teachers to focus effectively on these two qualitatively different tasks, and therefore we tend to perform them more fairly and consistently. It has also reduced the students' tendency to try to decode grades as feedback. Studies discussed by Black and William (1998) show that comments accompanied by grades often lead to reduced learning in comparison to comments without grades, and that indeed grades may act as a barrier to student understanding.