Iris Murdoch's profound and well-documented late life cognitive decline was caused by the accumulation of amyloid plaques and tau tangles in her brain: the molecular signature of Alzheimer's disease. Yet plaque and tangle deposition begins many years, if not decades, before cognitive changes first emerge, so the discovery of methods for defining and measuring this 'preclinical' phase is one of the central challenges in Alzheimer's disease research. Language change is regarded as a leading contender for the status of preclinical marker, with deep and detailed linguistic analysis now performed on speech and language samples from thousands of patients and controls. One of the earliest pieces of work relevant to this objective was the textual analysis that Professor Garrard and his colleagues in the University of Cambridge carried out on Murdoch's final novel, Jackson's Dilemma, which she wrote right at the end of the preclinical phase. Jackson's Dilemma showed changes similar to those found in the language of patients with established disease. After reviewing this work and the subsequent experiments it inspired, Professor Garrard will introduce the new generation of computationally based techniques in corpus linguistics by presenting some more recent work on longitudinal changes in Murdoch's informal writings (diaries and letters), which he has examined in a fine-grained manner using a machine learning approach.