As can be seen above, the speaking test for the A1+A2 level integrated test, as well as for the B1 and the B2 level tests consist of three activities. The first one involves questions relating to candidates’ immediate environment, work, hobbies, interest etc., while the C1 level speaking test consists of two activities the first of which requires candidates to respond to an opinion question. The A1+A2 level speaking test does not include mediation activities which are included in the B1, B2 and C1 level tests. That is, candidates at this level are required to perform orally in English, relaying information they have selected from a source text in Greek, so as to respond to a given communicative purpose. 

The speaking test procedure
The KPG speaking test involves two examiners and two candidates in the examination room. One of the two examiners is the ‘Interlocutor’, i.e., the one who conducts the exam (in other words, asks the questions, assigns the tasks and participates in the speech event). The other is the ‘Rater’, i.e., the one who sits aside silent and evaluates and marks the candidates’ performance. The Interlocutor also marks their performance, once candidates have left the room. Examiners alternate in their role as Interlocutor and Rater every three or four testing sessions. 
Examiners are trained for their roles as Interlocutor and Rater through seminars which take place systematically throughout the year. However, on the day of the oral test, examiners are given an examiner pack which contains guidelines regarding the exam procedure and oral examiner conduct, and the oral test material: questions, tasks and rating criteria. These are handed to examiners at least two hours before the exam begins along with the Candidate Booklet which contains the prompts for the exam (photos and/or Greek texts, since one of the activities involves mediation).

Assessment criteria for oral production
The assessment criteria for oral production reflect the functional approach to language use that the KPG exams adhere to. The structure of the assessment scales for the different level speaking tests are similar, while the nature of the criteria depends on the expectations for oral production, illustrative descriptors and can-do statements for each level and on the requirements of the speaking activities. Thus for the A1+A2, the B1 and the B2 level speaking tests, the assessment criteria are grouped under two main categories: criteria for assessing a) task completion and b) language performance. As a result, the Speaking Test Rating Scale, which provides a breakdown and description of each criterion on the oral assessment scale, is presented in two parts. Part 1 presents the Task completion criteria, which have to do with the degree to which the candidates achieved the communicative purpose of the task, while Part 2 presents the Language performance criteria, which focus on the quality of language output in terms of pronunciation, vocabulary, grammar and syntax, and coherence (for A1+A2), and in terms of the candidates’ phonological competence, linguistic competence, sociolinguistic competence and pragmatic competence for the B1 and B2 level speaking tests. 
Given the requirements of the C1 speaking test, the assessment criteria are grouped under two categories: a) overall performance for tasks 1 & 2, and b) assessment of task 2: the mediation activity. As a result, the Speaking Test Rating Scale, which provides a breakdown and description of each criterion on the oral assessment scale, is presented in two parts. Part 1 presents the Overall performance for tasks 1 & 2 criteria which focus on the quality of language output in terms of the candidates’ phonological competence, linguistic competence, appropriateness of language choices, and cohesion, coherence of speech and fluency; Part 2 presents the Assessment of task 2 criteria, which focus on the interaction and mediation skills of the candidates. 

Developments and research relating to the speaking test
Assessment of oral proficiency is a complex and largely subjective process in which many variables affect the quality and quantity of language output, while the rating of performance ultimately threatens the validity, reliability and fairness of the oral test procedure. Given that one of the most significant variables potentially affecting candidate output and examiner rating is the linguistic conduct of the examiner (Bachman et al. 1995; Bonk & Ockey 2003; Lazaraton 1996, McNamara 1996; O’Sullivan 2000) , the KPG exam system has established a systematic, intensive and on- going programme of oral examiner training. The goal of the training programme is to develop a database of approved oral examiners; that is examiners who have taken part in all training seminars and whose performance has been positively evaluated by specially trained observers. The Observation Project, which has yielded interesting data now systematically being analyzed, will be hosted in the KPG Corner of the next issue of ELT News. 
Another aspect of the speaking test being investigated systematically is the reliability of examiner marking. This is monitored and assessed through systematic inter-rater reliability checks during/after each exam administration. For the speaking test, inter-rater reliability is assessed through the data collected by observers who note the mark for each candidate awarded by each observed examiner. On the observation form, the observers also note their mark of candidates’ performance. This data is then collected by the RCeL, where inter-rater reliability estimates are calculated and results are interpreted. 
Finally, after each exam administration, oral examiners are requested to complete an oral test feedback form which aims to elicit feedback concerning the potential problems with test items, their usefulness, appropriateness and practicality. These results are sent to the project team directors, who share and discuss them with their test development teams. The project team undertakes a systematic post-examination review, weighs the results of the exams, and takes into consideration the responses on examiner feedback forms before deciding on potential revisions/changes deemed necessary for the improvement of the test papers.