In this study, it was aimed to develop a data collection tool to determine the verbal lecture self-efficacy perceptions of middle school students. Accordingly, an item pool has been created by examining firstly learning outcomes in the Turkish Language Teaching Program in 2018, Gee’s discourse analysis steps and related literature in terms of verbal lecture self-efficacy. The suitability of 70-item experimental form created was tested with 112 middle school students as pre-application by using “The think aloud method” in respect to lucidness and level. Afterwards, mentioned empirical form was presented to expert opinion. The draft of the scale created in accordance with expert opinions was applied to 334 middle school students. As a result of the exploratory factor analysis, it was specified that the experimental form is suitable for the two-factor model. The percentage of variance explained by all of the factors is 69%; the reliability co-efficient for all items of the scale was calculated as Cronbach alpha 86. The analysis results of the data obtained from the scale applied to 832 middle school students for confirmatory factor analysis showed that there is a consistence between the obtained fit indices, model and data, and the proposed model has a consistence at the medium or acceptable level. In the final form of scale, Cronbach alpha internal consistency reliability co-efficient was determined as .89. In consequence of the analysis made through the Mplus program, a verbal expression self-efficacy perception scale with 27 items in 4-point Likert type was developed. The findings obtained reveal that the scale aforesaid is a valid and reliable measurement tool.