The primary focus of this thesis is on normative synaptic plasticity theories, which establish computational links between experimentally observed synaptic plasticity phenomena and the critical behavioral and developmental functions that they support. In Chapter 2, we will introduce and define this class of theory from the ground up. We will also critically review previous literature dedicated to developing and testing normative plasticity theories, and produce a set of guidelines that future modeling efforts should attempt to adhere to in order to facilitate the testing of these theories. In Chapter 3, we show how a reward-modulated normative plasticity rule can produce sensory representations that compensate for noise and are efficient, in that they selectively represent task-relevant information without wasting metabolic resources. In Chapter 4, we observe that our algorithm has many similarities to perceptual learning in the mouse auditory cortex: we adapt it to demonstrate how reward and context information delivered by acetylcholine signals from the nucleus basalis could underlie both context-specific adaptation in auditory cortex and reward-based perceptual learning in mice. In Chapter 5 we develop a theory called `impression learning', which proposes a mechanism for learning sensory representations by adapting synapses to minimize a prediction error between predictive signals arriving at apical dendrites of pyramidal neurons and incoming sensory information at basal dendrites. This theory generalizes the Wake-Sleep algorithm, and improves on previous prediction-error based theories of learning by demonstrating how learning can occur continuously with sensory perception, rather than requiring an offline learning phase. In Chapter 6, we close off the thesis with a theoretical examination of the difficulties associated with studying complex, adaptive systems experimentally. Our results across the chapters of this thesis collectively demonstrate the importance of normative theories of plasticity, both for conceptualizing learning in the brain and informing experiments that investigate adaptive neural circuits.