Our everyday knowledge is rich and multifaceted, capturing general patterns (most dogs bark) alongside specific details (your friend’s dog bites). This dual capacity is a remarkable achievement of human memory, given the tension it involves: generalization requires integrating across memories to reflect commonalities, while specificity demands keeping memories separate to minimize interference. How do we build memory representations that capture both general and specific knowledge? In Chapter 1, I review prior research on how general and specific knowledge may coexist or compete in memory, drawing attention to gaps in current understanding. In the following chapters, I pair novel category learning tasks with behavioral experiments, computational modeling, and neuroimaging to tease apart how general and specific information are represented in memory. In Chapter 2, I show that object features are rapidly warped in memory based on their generalizability: shared features are misremembered as more similar (integration), while unique features resist this distortion (separation). These findings, replicated in a neural network model, reveal how general information can be integrated while specific details are kept separate—afforded by adaptive feature-level memory distortions. In Chapter 3, I use fMRI to explore how this manifests in the human brain immediately after learning and after a delay of approximately two weeks. Immediately after learning, general and specific information were flexibly accessed, with distinct brain regions amplifying shared or unique features depending on retrieval demands. Over time, neural representations transitioned from separated to integrated, reflecting a shift from memory specificity to generalization. In Chapter 4, I discuss how this collection of findings sheds light on how human memory reconciles the generalization-specificity trade-off—either by prioritizing one form over the other over time or by simultaneously representing general and specific information across distinct brain regions and feature dimensions. Implications for theories of hippocampal and cortical memory function are also discussed. Together, this thesis provides novel insights into the cognitive and neural mechanisms that allow us to build the richly layered knowledge structures we rely on to make flexible and adaptive decisions.