This study aims to determining the self-efficacy perception of teacher candidates. It is a survey model. Population of the study consists of teacher candidates in 2010-2011 academic years at Ahmet Keleşoğlu Education Faculty of Education Formation of Selçuk University. A simple random sample was selected as sampling method and the study was conducted on 312 teacher candidates. The data of this study was obtained by “teacher self-efficacy scale” which was developed by Tschannen-Moran and Hoy (2001) and adapted into Turkish by Çapa, Çakıroğlu and Sarıkaya (2005). This scale consisted of three sub-dimensions which were “student participation”, “learning strategies”, and classroom management” and 24 items. Reliability value of the dimensions were 82; ,86; ,84 respectively. Cronbach Alfa reliability value of the dimensions become 82; ,86; ,86 as the result of the data analysis of the study. SPSS package program was used to analyse the data; and the data analysis, frequency, percentage, arithmetic mean, standard deviation, independent samples t-test, one-way analysis of variance (ANOVA) and LSD test were used. When the findings of the research were considered in general, the results has shown that the teacher self-efficacy perception level of the candidates was at desired level. However, teachers' views about the teacher self-efficacy perception level showed that most of them preferred lower level option "adequate", instead of "totally adequate" option that was the highest degree of participation of five likert-type scales. This illustrated that their perception about this issue was not at a very high level. When independent variables were taken into account, among the teacher self- efficacy perception levels of teacher candidates, a significant difference was detected in terms of gender, working as a teacher or not and type of education program variables but it revealed no significant difference in terms of academic achievement variable. Based on obtained results, several suggestions are made.