Magnetohydrodynamic (MHD) instabilities can limit the performance and degrade the confinement of tokamak plasmas. The Tokamak à Configuration Variable (TCV), unique for its capability to produce a variety of poloidal plasma shapes, has been used to analyse various instabilities and compare their behaviour with theoretical predictions. These instabilities are perturbations of the magnetic field which usually extend to the plasma edge where they can be detected with magnetic pick-up coils as magnetic fluctuations. A spatially dense set of magnetic probes, installed inside the TCV vacuum vessel, allows for a fast observation of these fluctuations. The structure and temporal evolution of coherent modes is extracted using several numerical methods. In addition to the setup of the magnetic diagnostic and the implementation of analysis methods, the subject matter of this thesis focuses on four instabilities which impose local and global stability limits. Al1 of these instabilities are relevant for the operation of a fusion reactor and a profound understanding of their behaviour is required in order to optimise the performance of such a reactor. Sawteeth, which are central relaxation oscillations common to most standard tokamak scenarios, have a significant effect on central plasma parameters. In TCV, systematic scans of the plasma shape have revealed a strong dependence of their behaviour on elongation κ, and triangularity δ, with high κ and low δ leading to shorter sawteeth with smaller crashes. This shape dependence is increased by applying central electron cyclotron heating (ECH), which increases or decreases the sawtooth period, depending on the plasma shape. The response to additional heating power is determined by the role of ideal or resistive MHD in triggering the sawtooth crash. For plasma shapes where additional heating and consequently, a faster increase of the central pressure shortens the sawteeth, the low experimental limit of the pressure gradient within the q = 1 surface is consistent with ideal MHD predictions. The observed decrease of this limit with elongation is also in qualitative agreement with ideal MHD theory. Edge localised modes (ELMs), occurring in TCV Ohmic high-confinement mode discharges, were observed to be preceded by coherent magnetic oscillations. The precursors prior to small ELMs, believed to be of type III, and prior to larger ELMs, previously referred to as TCV large ELMs, show the same characteristics, which allows for an identification of both ELMs to be of type III according to the usual classification scheme. The detected poloidal and toroidal mode structures are consistent with a resonant flux surface close to the plasma edge. Unlike conventional MHD modes, these precursors start at a random toroidal location and then grow in amplitude and toroidal extent until they encompass the whole toroidal circumference. Thus, the asymmetry causing and maintaining the toroidal localisation of the ELM precursor, must be intrinsic to the plasma. Soft X-ray measurements show that the localised precursor always coincides with a central m = 1 mode, which can usually be associated with the sawtooth pre- or postcursor mode. A comparison of the phases indicates a correlation with the maximum of the central mode preceding the toroidal location of the ELM precursor and, therefore, a hitherto unobserved coupling between central modes and ELMs. Highly elongated plasmas promise several advantages, among them higher current and beta limits. During TCV experiments dedicated to an increasing of the plasma elongation, a new disruptive current limit, at values well below the conventional current limit corresponding to qa >2, was encountered for κ >2.3. This limit, which is preceded by a kink-type mode, is found to be consistent with ideal MHD stability calculations. The TCV observations, therefore, provide the first experimental confirmation of a deviation of the linear Troyon-scaling of the ideal beta limit with normalised current at high elongation, which was predicted over 10 years ago. In addition to the ideal beta limit, several other MHD events are observed in highly elongated plasmas. The axisymmetric mode causes vertical displacement events and thereby, imposes a lower current limit. This operational limit is in good agreement with theoretical predictions of the growth rate of the axisymmetric mode. Furthermore, minor disruption are occasionally observed. They are caused by bursts of tearing modes located close to the q = 1 surface, which are destabilised by flat central current profiles, typical for highly elongated plasmas. Neoclassical tearing modes (NTMs), which have been observed to limit the achievable beta in a number of tokamaks, arise from a helical perturbation of the bootstrap current caused by an existing seed island. Neoclassical m/n = 2/1 tearing modes have been identified in TCV discharges which are characterised by a low electron collisionality νe*, a medium ion collisionality ν*, a medium value of beta and strongly peaked pressure and current profiles. In contrast to other tokamak experiments, where sawteeth, fishbones or ELMs generate the seed island, the required seed island is provided by a conventional tearing mode. The island clearly shows a conventional and a neoclassical growth phase. The TCV results provide the first clear observations of such a trigger mechanism which could also explain the occurrence of "triggerless" NTMs observed in other experiments. The slowly growing seed island also allows a measurement of the critical seed island width ωcrit, which is observed to increase with increasing density. In conclusion, several local and a global stability limits are analysed. These instabilities can limit the pressure gradient and thereby, the performance of the plasma. The presented results reveal several previously unobserved features of commonly observed instabilities. Since the most of the new observations can be explained by theory, they improve the predictive capabilities with respect to new experiments. The experiments have also shown some new interactions among different instabilities, which add to the already crowded complexities of MHD phenomena in fusion plasmas.