Verbal fluency is widely used as a clinical test, but its utility in differentiating between neurodegenerative dementias and progressive aphasias, and from healthy controls, remains unclear. We assessed whether various measures of fluency performance could differentiate between Alzheimer's disease, behavioural variant of frontotemporal dementia, non-fluent and semantic variants of primary progressive aphasia, progressive supranuclear palsy, corticobasal syndrome and healthy controls. Category and letter fluency tasks were administered to 33 controls and 139 patients at their baseline clinical visit. We assessed group differences for total number of words produced, psycholinguistic word properties and associations between production order and exemplar psycholinguistic properties. Receiver operating characteristic curves determined which measure could best discriminate patient groups and controls. The total word count distinguished controls from all patient groups, but neither this measure nor the word properties differentiated the patient groups. Receiver operating characteristic curves revealed that, when comparing controls to patients, the strongest discriminators were total word count followed by word frequency. Word frequency was the strongest discriminator for semantic variant of primary progressive aphasia versus other groups. Fluency word counts were associated with global severity as measured by Addenbrooke's Cognitive Examination Revised. Verbal fluency is an efficient test for assessing global brain-cognitive health but has limited utility in differentiating between cognitively and anatomically disparate patient groups. This outcome is consistent with the fact that verbal fluency requires many different aspects of higher cognition and language.