One of the biggest challenges facing experimental studies of granular rheology is the opacity of the constitutive particles, which prevents direct observations of their behaviour and interactions. This thesis describes a series of original experiments where instantaneous forces between individual particles within the bulk of 2D flows are quantified. The specific type of granular flow we study is gravity-driven, dry, and in the slow to intermediate regime. Here we describe a novel adaptation of the photoelastic technique and explain how we applied it in an original setup to offer unprecedented insight into the force distribution within granular flows, as this has never been achieved experimentally in dynamic systems before. Firstly, using particle tracking and photoelastic force measurements we report coarse-grained profiles for packing fraction, velocity, shear rate, inertial number, and stress tensor components, as well as statistical observations drawn from the measurable forces. Despite the highly fluctuating and seemingly random nature of the force network, we draw analogies between discrete and continuous flow models and characterise force chain preferential orientations. Secondly, we interpret current rheological models in the context of our experimental system, and hence propose that non-local effects may in fact be dependent on the local force network fluctuation rate. The results of this work further the community’s understanding of granular force networks and complement the physical concepts applied in current non-local rheological models. Finally, we model how differences in the force network between mono- and bi-disperse avalanching granular media lead to the mechanisms that drive granular size segregation. This work then also provides quantitative, tangible support to granular segregation models based on the physical mechanisms that drive it. As the first experimental observations of their kind, our experiments can be used to validate existing and even future theoretical and numerical research. Furthermore, the physical mechanisms proposed in this work can be used to construct future models of granular behaviour that lie beyond the scope of this particular thesis.