Knowledge Representation and Reasoning (KRR) is an interdisciplinary research field dedicated to the formalization and conceptualization of knowledge in order to enable machines to solve tasks that require logical deduction/induction and to help people retrieve information from KR systems in more intuitive ways. Despite their success stories in semantic search, semantic parsing, question answering, recommender systems, etc., one commonly neglected aspect is that the world is ever-changing, and thus, statements about the it may only hold during a certain time period. Unfortunately, temporal information is often inaccurate, incomplete and of different types and forms (e.g., quantitative time versus temporal relations). This poses great challenges to conventional rule-based reasoning systems (i.e., symbolic reasoning), which are deterministic, and unable to address noise (errors or incompleteness). In order to address these challenges, this dissertation focuses on developing new methods to represent and reason about temporal information in a subsymbolic manner. Chapter 3 and Chapter 4 study how to subsymbolically represent various temporal primitives (i.e., time instants and intervals). Chapter 5 shifts focus to qualitative temporal relations and presents a subsymbolic approach to perform (spatio-) temporal reasoning. Chapter 6 investigates why subsymbolic approaches outperform conventional methods in terms of qualitative (spatio-) temporal reasoning, and finds conceptual neighborhood of qualitative relations can be discovered by subsymbolic approaches. Throughout the dissertation, I focus on how domain theory can be used in subsymbolic methods and how theories can be discovered by subsymbolic methods.