Primary progressive aphasia (PPA) is the term used to refer to a group of neurodegenerative conditions that affect predominantly the language network. Based upon the profile of language and speech involvement, three clinical variants are proposed by the International Consensus Criteria (ICC). Semantic dementia (SD) is the best-defined clinico-pathological entity characterised by anomia and impaired single-word comprehension. The second variant, progressive nonfluent aphasia (PNFA), is characterised by the presence of agrammatism, motor-speech disorders or both. Logopenic progressive aphasia (LPA), the newest variant, is defined by the concomitance of naming and sentence repetition deficits but spared motor aspects of speech. SD and PNFA are largely associated with frontotemporal lobar degeneration with TDP-43 and tau pathology, respectively; whereas LPA may represent an atypical presentation of Alzheimer s disease (AD). The applicability of ICC to an unselected clinical sample is unknown and there is, at present, no agreed clinical evaluation scale on which to derive the diagnosis. In addition, although each clinical variant has a distinctive distribution of brain atrophy, the neural correlates and stability over time of the core linguistic deficits in LPA have not been yet determined. To tackle these issues, the present thesis followed an integrative approach that assumes the interdependence between the clinical phenotype, distribution of cortical atrophy and pathology. The clinical phenotype was ascertained using a newly created language assessment scale that allowed quantification of core language deficits across PPA variants. Next, the cognitive stability over time was tracked using the Addenbrooke s Cognitive Examination and word-based language tasks. The neural correlates of LPA were determined correlating the core deficits with measures of cortical thickness. The estimation of underlying pathology for each PPA variant was achieved using amyloid imaging as a putative biomarker of AD. The results confirms that cases with PPA can be separate into three coherent clinical-pathological variants, supporting the hypothesis that segregated components of the language network are selectively damaged by specific pathologies. The integrative approach not only provides insights about how language is organised in the brain, but also it contributes to a better nosological characterisation, which is relevant to clinical practice.