Manual volumetric measurement of the brain's frontal lobe and its subregions from magnetic resonance images (MRIs) is an established method for researching neural correlates of clinical disorders or cognitive functions. However, there is no consensus between methods used to identify relevant boundaries of a given region of interest (ROI) on MRIs, and those used may bear little relation to each other or the underlying structural, functional and connective architecture. This presents challenges for the analysis and synthesis of such results. We therefore performed a systematic literature review to highlight variations in the anatomical boundaries used to measure frontal regions, contextualised by up-to-date evidence from histology, hodology and neuropsychology. We searched EMBASE and MEDLINE for studies in English reporting three-dimensional boundaries for manually delineating the brain's frontal lobe or sub-regional ROIs from MRIs. Exclusion criteria were: exclusive use of co-ordinate grid systems; insufficient detail to allow method replication; publication in grey literature only. Papers were assessed on quality criteria relating to bias, reproducibility and protocol rationale. There was a large degree of variability in the three-dimensional boundaries of all regions used by the 208 eligible papers. Half of the reports did not justify their rationale for boundary selection, and each paper met on average only three quarters of quality criteria. For the frontal lobe and each subregion (frontal pole, anterior cingulate, dorsolateral, inferior-lateral, and orbitofrontal) we identified reproducible methods for a biologically plausible target ROI. It is hoped that this synthesis will guide the design of future volumetric studies of cerebral structure.