According to the theory of the social mind and of sociocultural Second Language Acquisition (SLA) (Lantolf 2000, 2006, Léontiev 1981, Swain 2000, Wertsch 1985), every form of interaction is situated and strongly depends on the individualities involved, each one speaking from their own universe but also permanently adjusting to their partner's discourse. When it comes to the assessment issue, and more specifically spoken interaction assessment, both Cambridge English and the Centre International d'Etudes Pédagogique (CIEP) exams and certifications include the student-to-student and examiner-student conditions whereas the Goethe-Institut and the Cervantes Institute, which are also exam providers, prefer the teacher-to-student condition – the main argument for the latter being that the teacher's didactic/assessing positioning can better help the student to produce the best possible performance. The purpose of this research is to investigate which of the two conditions is preferable in terms of linguistic and pragmatic performance: teacher-to-learner or peer-to-peer. Our hypotheses are the following: 1. Evidence of didactic/assessing positioning is not limited to language professionals but can be found in student-to-student interactions as well. 2. The teacher-to-learner condition is not necessarily more favourable to the student than the peer-to-peer one. This research will deal with the results of a micro experiment completed in 2013 in connection with the Département de l'Évaluation, de la Performance et de la Prospective (DEPP) at the French Ministry of Education in an assessment situation involving A1, A2, B1 students and teachers/evaluators in order to compare the performances of the students doing the same task under two conditions: first with a peer and then a teacher.