Redox processes in transition metal oxides have been in the focus of solid-state research for decades, given their substantial potential for applications in the fields of sensorics, information technology, superconductivity, and energy conversion. In this context, strontium titanate has become one of the most intensively-investigated solid oxide with perovskite structure and mixed ionic-electronic conductivity. Despite of these considerable research activities, the role of extended defects such as dislocations and grain boundaries in the electronic and ionic transport remains under debate. It is well known that the oxygen nonstoichiometry determines the amount of electronic charge carriers allowing for a control of the electronic conductivity via self-doping by oxygen vacancies upon reduction and oxidation. Assuming a homogeneous bulk phase, this effect can be modelled using the concepts of point defect chemistry in good agreement with macroscopic experimental data [1]. When having a more detailed look on the crystal’s properties on the nanoscale, however, it appears that the redox phenomena are far more complex than previously thought. We present experimental investigations on single crystals, bicrystals and ceramics using high-resolution imaging techniques such as local-conductivity atomic force microscopy (LC-AFM), Kelvin probe force microscopy (KPFM), and scanning nearfield optical microscopy (SNOM), which reveal that dislocations are easy reduction sites where oxygen vacancies are preferentially generated. In turn, filaments with high electronic conductivity evolve around the dislocations in the originally insulating matrix and act as nanoscale short circuits [2]. As dislocations form a three-dimensional network in the surface layer and along of inner surfaces such as grain boundaries, the macroscopic properties of strontium titanate can be turned from insulating to metallic and even to superconducting behaviour. This transition takes place at a very low global oxygen vacancy concentration thus reaching a state that is far outside the prediction of homogeneous defect chemistry. Since there is an attractive interaction between dislocations and oxygen vacancies, the nanoscale conducting filaments can even be manipulated mechanically at room temperature [3]. As dislocations are intrinsically present in real crystals and ceramics, our findings not only can explain failure mechanisms in solid oxide electrolytes, but also challenge traditional models describing the mechanisms of electronic transport and superconductivity in self-doped transition metal oxides.