Natural language requirements (NLRs) modeling and verification allow us to detect errors in requirements as early as possible in an effective way. The NLRs of system development usually involve states and modes either explicitly or implicitly. Different stakeholders have different understandings of these two terms. The misunderstanding of states and modes in NLRs severely impedes requirements modeling and validation. For example, conflicts in system validation may occur when users and developers do not share a consensus on the terms. Modeling and verification of requirements from the viewpoint of states and modes have never been investigated. Thus, in this paper, an innovative approach to analyzing requirements is proposed. To this end, an overview of states and modes in literature is performed to help us understand the relationship between them. The MoSt language (a Domain Specific Language, DSL) is then designed for requirements modeling and implemented by Xtext framework. Meanwhile, a model validator is realized to define user-defined rules, in order to statically check requirements. In the following, a code generator is also accomplished to realize the automatic model transformation from the MoSt model to the NuSMV model, which aims at the dynamic checks of requirements. The grammar, the model validator, and the code generator are integrated into an Eclipse-based tool which is available on GitHub. A case study on requirements for designing cars has been conducted to illustrate the feasibility of our approach. In this case study, we injected 11 errors. The results show that 6, 4, and 1 were detected in the static analysis, traceability analysis, and completeness analysis of the model respectively. Finally, the feasibility Preprint submitted to Elsevier 2 juin 2021 of our approach to reducing conflicts between user and developers has been illustrated in validation analysis.