Granular matter is a broad term that describes materials comprised of macroscopic grains. Granular material has unique properties that can mimic either a solid- or fluid-like system and has macroscopic behaviors such as segregation, shear- jamming, reversibility, and compaction. The finite size of the grains suggests that thermal fluctuations are neglibible compared to the macroscopic interactions such as gravitational potential. This implies that tools developed in thermodynamics cannot be readily applied. Instead, research into granular material uses a combination of bulk measurements (packing density, pressure) with grain-scale tracking of position, orientation, and forces. This thesis presents four main studies utilizing three-dimensional experiments and simulations to probe the dynamics of individual grain subject to cyclic compression.The first study uses numerical simulations to connect granular rotations to translations in sheared granular packings. It is proposed that rotations play an extensive role in the formation of shear zones in granular packing, in which the rotations allow for ball-bearing like motion that could reduce the stress from external pressures. In this study, we quantify the effect of friction on the shear-zone rotations and translations. We find a direct connection between average rotations and the vorticity of translations independent of the friction. The second study explores reversibility of grain translations and rotations in the context of memory formation. In granular matter, memory is formed via a response to an external perturbation, ranging from compression and shear to thermal cycling. In this project, we encode and read out memory of compression amplitudes for a cyclically driven granular system. We find that memory is significantly affected by the interparticle friction of each grain and is most readily extracted by quantifying reversible displacements within the sample. The third study experimentally measures three-dimensional orientations of granular spheres using our refractive index matched scanning setup. We apply a combination of deep learning and image processing to extract the position and orientation of individual grains subject to cyclic compression. Using this, we can quantify the spatial distribution of sliding and rolling motion of contacts. We find that sliding occurs deep within the sample where the grains are mostly fixed in place. This occurs with an increase in internal stress within the material. Finally, we explore collectively rotating states in three-dimensions. We introduce a new measure in which we can identify affine (collective) and non-affine rotations. We find that grain rotations are generated by minimizing sliding mo- tion between all contacts independent of the forces between each contact. Further, we find that the collective rotation component is more directly correlated to irreversible translations than the residual rotations. This result identifies that collective rotations are important to reversible states in sheared granular systems.