Total recall: Detecting autobiographical memory retrieval in the absence of behaviour.

Functional magnetic resonance imaging (fMRI) was used to investigate autobiographical memory in a single patient with a disorder of consciousness (DoC). The patient viewed video clips from their own recent mall visit, a control's visit to the same mall, and a novel bookstore. A machine-learning classifier trained on twelve healthy participants decoded the patient's brain activity, distinguishing the three conditions with balanced accuracy of 0.448 (p = .032) and distinguishing the patient's own clips from the control's clips with balanced accuracy of 0.609 (p = .032), both within the control group's performance range. These results suggest that autobiographical memory processes may remain intact in some DoC patients despite their inability to report them.

Study at a glance

Abstract

Functional neuroimaging has fundamentally changed our understanding of disorders of consciousness (DoC). While many DoC patients exhibit minimal to no behavioural responsiveness, a significant minority show neural evidence of awareness and preserved cognitive functioning. Although several cognitive functions have been explored in DoC patients, autobiographical memory -- the ability to form and retrieve personal memories -- has yet to be investigated. To address this gap, we used functional magnetic resonance imaging (fMRI) to investigate autobiographical memory in one DoC patient. The patient viewed video clips across three conditions: (1) Own - clips recorded from their perspective during a recent mall visit; (2) Other - clips from a healthy control's visit to the same mall; and (3) Bookstore - novel clips from an entirely different store that had not been visited. We trained a linear support vector classifier to associate fMRI activity in canonical autobiographical memory regions with each condition using data from twelve healthy participants. We then applied the trained model to the patient's data to 'decode' which condition their fMRI activity predicted. The model accurately distinguished between Own, Other, and Bookstore conditions in the patient (Balanced Accuracy = 0.448, p = .032), with performance within the control group range (p = .068). Similarly, the model distinguished between the Own and Other conditions above chance (Balanced Accuracy = 0.609, p = .032) and within the control group's distribution (p = .620), suggesting that the patient was still able to differentiate personal experiences from visually similar scenes, despite being behaviourally unable to report that this was the case. These findings provide preliminary evidence that autobiographical memory processes, critical to conscious awareness and identity, remain intact in some DoC patients, shedding further light on their covert capabilities and inner experiences.