Most proteins form specific, “native” three-dimensional structures that are required for them to function properly. This self-assembly process, called folding, is usually very reliable. When folding goes awry, however, non-native structures can result that lead to disease. I will discuss our work on the structural dynamics of the prion protein PrP, which misfolds through an unknown mechanism into an infectious form that causes “mad cow” disease. We use high-resolution optical tweezers to observe the structural dynamics of individual PrP molecules in real time as they either fold natively or misfold and aggregate, focusing on the microscopic mechanisms that determine the structural outcome (native or misfolded). Studying isolated PrP molecules, we measured the energy landscape that governs the native folding, showing that PrP folds in a single step without observable intermediates. From the shape of the landscape, we determined both the time required for the actual transition to take place, as well as the configurational diffusion coefficient that sets the fundamental timescale for folding. We also developed new analytical approaches to detect and characterize states that are rarely occupied, thereby discovering that a single PrP molecule can form several types of misfolded structures. Although these misfolded structures were not stable in a single PrP molecule and hence formed only fleetingly, when two PrP molecules were brought together to form a dimer, misfolding became instead dominant— the native structure was no longer observed to form at all. Deciphering the series of steps leading to the misfolded dimer, we reconstructed the energy landscape for the misfolding and uncovered a key intermediate driving the change in behavior. These results show how single-molecule probes can be used both to test physical theories of folding and to address practical questions in biology and medicine.