The goal of this study is to determine the reliability of the performance points of dentistry students regarding communication skills and to examine the scoring reliability by generalizability theory in balanced random and fixed facet (mixed design) data, considering also the interactions of student, rater and duty. The study group of the research consists of the 16 dentistry students who took communication skills elective course in Hacettepe University in 2014-2015 academic year. In the study, the variability sources are taken as student (s), rater (r), occupation (o) and task (t). 16 students are assigned with 16 different duties and whether the students fulfill the duties is rated by 8 raters from two different occupation groups, 4 standard patients and 4 academicians. On these facets using different designs G study has been performed. Within the scope of this study the sxrxt design where all variability sources are crossed; sx(o:r)xt design where the occupation nested on rater as the result of scoring of the students by two different occupation groups as standard patient and academicians; sxaxt and sxspxt facets where rater facet is taken as fixed facet according to occupation groups (sp: Standard Patient, a: Academician) were examined. For each design estimated variance components are discussed separately.