Various neuropathological findings have been reported in bipolar disorder (BD). However, it is unclear which findings are well established. To address this gap, we carried out a systematic review of the literature. We searched over 5000 publications, identifying 103 data papers, of which 81 were eligible for inclusion. Our main findings can be summarised as follows. First, most studies have relied on a limited number of brain collections, and have used relatively small sample sizes (averaging 12 BD cases and 15 controls). Second, surprisingly few studies have attempted to replicate closely a previous one, precluding substantial meta-analyses, such that the latter were all limited to two studies each, and comprising 16-36 BD cases and 16-74 controls. As such, no neuropathological findings can be considered to have been established beyond reasonable doubt. Nevertheless, there are several replicated positive findings in BD, including decreased cortical thickness and glial density in subgenual anterior cingulate cortex, reduced neuronal density in some amygdalar nuclei, and decreased calbindin-positive neuron density in prefrontal cortex. Many other positive findings have also been reported, but with limited or contradictory evidence. As an important negative result, it can be concluded that gliosis is not a feature of BD; neither is there neuropathological evidence for an inflammatory process.