The ability of cells to receive and process chemical information is fundamental to the function of biological systems. Cell responses to salient chemical cues are characteristically reliable and precise, a striking example being the guidance of axons to their targets during brain development. However, chemical signalling pathways are subject to multiple sources of unavoidable noise, due to the random nature of molecular motion and interactions. This suggests that cellular systems may be optimised for transduction of noisy signals, which provides a guiding principle for understanding cell function at a biophysical level. In pursuit of this idea, we study three model problems, motivated by sensing and signalling in axon guidance. Our approach is to construct mathematical models, and through analysis and simulation, extract general principles and hypotheses about the quantitative nature of biological noise and mechanisms for sensitive chemosensation.First, we quantify the physical limits of chemosensation imposed by the random thermal motion of the molecules that need to be counted. In particular, we show how this depends on the dimension and spatial extent of the domain of diffusion. Although recurrent diffusion in 1d and 2d is detrimental to measurement precision, we find these effects are suppressed when sensing is performed within a confined space. Second, we study the stochastic behaviour of the inositol 1,4,5-trisphosphate receptor ion channel, which couples receptor activity at the membrane to the downstream calcium response that regulates axon growth and turning. By modelling the basic biophysical events that control ion channel opening, we introduce a new principle for understanding the origin of the multiple gating modes observed in single channel recordings. Third, we examine the finely-tuned response of dorsal root ganglia neurons to very shallow neurotrophin gradients. We show how paracrine signalling within the ganglion could explain this extreme sensitivity, as well as the currently unexplained biphasic dose response of many axon growth and guidance cues.Overall, this thesis gives new quantitative insights into chemical signalling in biological systems, across multiple stages of processing and spatial scales. Our models make testable predictions to motivate new experiments, and provide a strong foundation for further theoretical and computational study in both axon guidance and broader domains of biophysics.iii