The model of Brophy (1983) makes it clear how teacher judgment may influence students’ motivation and emotion. It assumes that teachers have different expectations about students: some students are underestimated, at the same time, some may be overestimated . These judgments are reflected in teachers’ differential behavior by providing disparate emotional and learning support (Babad, 2009; Rosenthal, 1973). Students can decipher teachers’ differential behavior and change their motivation and emotions to react. Differential teacher behavior may change classroom enjoyment and can affect students’ learning goal orientation. If students feel more inspired and appreciated by teacher, they would be much more willing to improve themselves, face difficult challenges and work harder to learn.

2.1.2 The accuracy of teacher expectations and teacher judgments

To make sure how well teachers can evaluate student achievement, two coefficients can be used. First, a simple Pearson correlation between teacher ratings and student achievement can be calculated. In this circumstance, students are not treated as members of different classes. Second, a alleged rank element can be decided. Correlation coefficients are evaluated class-wise and averaged across classes by the help of Fisher’s z-transformation(Cronbach, 1955). The rank element better fits teachers’ evaluation perspective. Teachers initially view the performance competence of their class and regard it as a form of reference for making performance evaluation(Schrader & Helmke, 2001 ). Generally speaking, measures of accuracy decided by the rank element are little higher than total correlations across all classes (Hoge & Coladarci, 1989).

Teacher expectations reflect student achievement mainly because they are accurate and correlation  coefficients found in self-fulfilling prophecy studies usually range between .40 and .80 (Jussim, Robustelli, & Cain, 2009). Research on teacher judgments comes out with the similar results. Hoge and Coladarci found in a meta-analysis of 16 studies a median correlation of 66 between teacher judgment and student performance in a standardized achievement test. 

Other investigations focused on accuracy of teacher judgments for students’ motivation and emotion. Teachers typically show a high accuracy when they are asked about students’ expectancy of success, or how students expect to perform in the next exam. The class-wise evaluated correlations between teacher judgments and students self-reports are often higher than .60 (Urhahne et al., 2010, 2011 ). Teachers can predict students' academic self-concept with medium accuracy and correlation coefficients usually range between .30 and .60(Marsh & Craven, 1991; Praetorius, Berner, Zeinz, Scheunpflug, & Dresel, 2013). In general, correlations tend to be higher when predicted student characteristics closely correspond to student performance.

2.1.3 Teachers’ differential behavior towards different students

Enormous studies have shown that teachers behave differently towards different students based on their performance expectations (Babad,1993). Meta-analytically derived effect sizes by Harris and Rosenthal support the view that teachers favor high-expectancy students in comparison to low-expectancy students with respect to four factors: climate, feedback, input, and output. Teachers provide high-expectancy students with a warmer socio-emotional climate (r = .23), give them more differentiated feedback (r =.13), hand them more challenging learning materials (r = .26), and open up more opportunities to respond to teacher questions (r = .18) (Babad,1993). Furthermore, Harris and Rosenthal (1985) have shown that this teacher behavior is associated with better performance by high-expectancy students, and thus provided indirect evidence for the existence of self-fulfilling prophecies.

At most of time, teacher expectations reflect students’ actual performance(Babad, 2009). A high-expectancy student shows high performance and a low-expectancy student low performance. When comparing misjudged students this equalization has to be resolved because underestimated students perform at least as well as overestimated students (Urhahne et al., 2010, 2011; Urhahne,