The current thesis explored diagnosis and rehabilitation of deficits in memory and language, using a multiple neuropsychological case study design. Broadly, the work evaluated the use of cognitive theory to diagnose patients’ clinical presentations and inform rehabilitation methods, and explored how outcomes from these interventions can be used to test cognitive theory in turn. This bi-directional link was explored in two ways: Firstly, theoretically-motivated groupings of word stimuli (e.g. ‘neighbourhoods’) were used to evaluate patterns of post-therapy generalisation, testing hypothesised associations between types of word stimuli. Secondly, the work identified proposed links between cognitive functions, using rehabilitation to test the validity, and nature, of these associations. The thesis is therefore comprised of two parts: Part 1 explored ‘neighbourhood’ effects in language and how they might be used to direct generalised improvement following rehabilitation (Chapters 2, 3, 4 and 5) and Part 2 evaluated associations and dissociations of functions in the cognitive architecture, across therapeutic and experimental contexts (Chapters 6 and 7). The work demonstrated that using theoretically-driven stimuli sets in rehabilitation can maximise generalised improvements following language treatment, and detailed how rehabilitation can be harnessed to test the integrity of associations between cognitive functions in the context of multiple deficits.