A mathematics teacher who makes preparations to improve the mathematical competence of the students takes into account the importance of modeling competence (Han & Kim, 2020). Moreover, it is thought that teachers should be able to interpret a mathematical model as well as design a mathematical model. Bilgili, Özkaya, Çiltaş and Konyalıoğlu (2020) examined the middle school mathematics teachers' error approach in modeling in their study, and they found that most of the teachers had difficulty in distinguishing the correct or incorrect solution while evaluating the solution performed. They also found that teachers had difficulties in explaining the modeling of concepts on which they did not have sufficient knowledge. In the light of the information provided, it is important for teachers both to have mathematical models competencies (Akgün et al., 2013; Han & Kim, 2020) and to be able to interpret these models (Bilgili et al., 2020). It is thought that it is important for teachers to be canalized to different teaching methods to maximize their mathematical competencies and skills such as problem- solving and modeling before becoming a teacher. It is also thought that this teaching process being error-based may be effective in meeting expectations. Based on the above-mentioned issues, the answer to the following question was sought in this study: “Do error-based activities have an effect on teaching mathematical models?” Finally, as in most experimental studies, it was tried to obtain a model between the independent variable of error-based activity and the dependent variable of teaching mathematical models in this study. In this study, one of the 4th-grade branches in the faculty where the research was conducted was selected as the experimental group while the other branch was selected as the control group. Therefore, the experimental design of the research was determined as a “quasi-experimental design” with an unbalanced control group. An unbalanced control group design, which is among the pre-test and post- test control group designs, was used in this study, which was conducted to measure the effect of error- based teaching of mathematical models on prospective teachers' skills to identify errors related to models. The research group of the study consists of 62 prospective teachers studying in the 4th-grade of primary education mathematics teaching program at a state university. There are 32 prospective teachers in the experimental group and 30 prospective teachers in the control group. The data were collected using a knowledge test containing seven questions. This knowledge test has been included in the study of Bilgili et al (2020). The researchers prepared two solutions, one of which was correct and the other was incorrect, for each question in this test. These solutions prepared by using different models were presented to prospective teachers as two different solution sheets. The prospective teachers, who were not informed about whether the solutions were correct or incorrect, were asked to evaluate the solutions in both solution sheets. The knowledge test was applied as a pre-test to the prospective teachers in the experimental group and control group. Then, an error-based teaching method was used to teach mathematical models to the experimental group, while the instructor-centered teaching method was used to teach mathematical models in the control group. The study was conducted for three hours a week and lasted nine hours in total. Pre-tests were applied one week before the application, and post- tests were applied one week after the application. In the control group, the lessons were taught by giving correct models for the concepts. On the other hand, in the experimental group, the lessons were taught by giving examples of both incorrect and correct models. For each question in the knowledge test, solutions were created using a correct or incorrect model. Therefore, different evaluation criteria were determined and scored. Before proceeding with experimental procedures for the experimental group and control group, it was necessary to test whether there was a significant difference between the groups' mean scores obtained from the knowledge test. Therefore, the Shapiro-Wilk test was conducted, and it was observed that there was no significant difference between the knowledge test pre-test scores of the prospective teachers in the experimental group and control group; also, the data was found to have a normal distribution (p> .05). However, significant differences were observed between the groups in terms of their success grades, regardless of their pre-test scores. In other words, the method used had an effect on students' error approaches. Considering the eta-squared values, it is observed that being in different groups explains 44% of the difference in the knowledge test post-test scores, regardless of the pre-test scores. However, it was concluded that knowledge test mean pre-test scores were also an important predictor of post-test scores (F (1.59) = 11.933, p <.01). Considering eta-squared values, it is observed that the knowledge test mean pre-test scores alone explain 16.8% of the difference in the mean post-test scores.