In this thesis colloidal dispersions were quantitatively studied by means of both light and electron microscopy. These colloidal particles, typically with a size somewhere between 1 and 1000 nm, have proven to be a good model system for physical phenomena in condensed matter physics, such as crystallization and the glass transition. As the colloidal particles are much larger than single atoms or molecules, they are also much slower, making it possible to study the phenomena in detail. Moreover, the larger size of the particles allows to characterize them with microscopy. We developed a method to effectively arrest colloidal dispersions for 3D study using light microscopy. We show that this method is applicable to a wide variety of colloidal dispersions. In addition, this method can also be used to preserve structures induced by an external field, such as an electric field, which would otherwise collapse when the external field is removed. This enables the development of new exciting materials. We also demonstrate that with the arrest method the nucleation of crystals of colloids can be studied in great detail. While previous microscopy studies were limited in their sample size, we show that we can now study volumes containing up to one hundred times more particles. In addition, we studied spherical assemblies of mixtures of binary colloids which form one of the Laves phases in bulk. We found that depending on the number ratio between the large and small colloids, the spherical geometry of the assembly plays no role, or surprisingly, leads to the formation of an icosahedral quasicrystal.