The data derives from a survey collected from 543 school-level teachers of English in Chile. The survey was originally distributed to 5435 teachers. The survey was developed with an aim of exploring teachers' beliefs about how second language grammar should be taught. The survey consisted of 50 items in total, in four sections: (a) background information (10 items), (b) beliefs regarding L2 learning and teaching (9 items), (c) beliefs regarding grammar instruction (23 items), and (d) classroom realities (8 items). Except for the background information section, the items took the form of a 6-point Likert scale. The entire dataset is included in an Excel file (.xlsx). The entire questionnaire is included as a supplementary file. The data is connected to the theoretical models proposed in [1]. In [1], those models were proposed based on descriptive statistics (e.g., agreement/disagreement rates) and focus-group interview data. Subsequently, in the current paper, the data was submitted to structural equation modelling to explain the theoretical models. Then, the data is visually depicted with figures created via AMOS.