Abstract
In this thesis, we are interested in the neurophysiological signature of the variation of the working memory load. This manuscript is organized along two main axes. The first axis corresponds to a state of the art of the study of working memory load. Working memory is a structure or a set of processes dedicated to the control and regulation of processing for the realization of a task. Given that the literature is rich in results obtained at multiple scales (neuron, local field, brain area, whole scalp, behavior), it is tempting to propose an integrative point of view, which we wished to address here. During the last decades, the communication between different scientific disciplines has evolved: from neuroscience to ergonomics, through mathematics and computer science, with "touches" of physics. As a consequence, the integrated analysis of the distributed and multi-scale system that is working memory is now possible. It is in this frame of reference that the present work is situated. The second axis of this manuscript is therefore naturally the description of the empirical work done during the thesis. The first step was a field investigation. The framework of application of our work being crisis management, we surveyed the participants of crisis cell simulations at the Laboratoire des Sciences du Risque (LSR, Institut Lines Télécom Mines Alès). This device is a semi-virtual training environment for strategic crisis management, based on multi-agent simulation. Our hypothesis that the crisis cell environment induces a significant increase in the perceived cognitive load was tested using this questionnaire. The next step was the design and implementation of the experimental protocol. Indeed, there are different possible approaches for the study of working memory load. We recorded the whole scalp EEG signal of twelve healthy participants while they performed a visual-spatial n-back task in three conditions of increasing difficulty, after an initial learning phase. We then analyzed the evoked spectral perturbations, and used time course decoding of cortical wave synchronizations at the individual level. We hypothesized that the synchronizations underlying the various cognitive functions embedded in the working memory load are potentially sequenced in time to avoid interference. We also hypothesized that the transient modulation of decoding accuracy of task difficulty would vary with increasing difficulty. Surprisingly, our results detected persistent decoding above chance level for concatenated frequency bands. This surprising result is in fact an encouraging clue for the implementation of a monitoring in operational conditions. Monitoring is the tracking of a measurable biosignal relevant for the live detection of an abnormal variation. In this context, the evolution over time of the signal must be indicative of the variable of interest. In our case, the EEG in the frequency domain is our measurable signal, it is indicative of the variation of the working memory load. It would be interesting to know to what extent the results obtained under experimental conditions are applicable to an ecological situation and can be combined with other neurophysiological measures. The challenge is to move towards a more flexible work environment that is more adaptable to humans in operational conditions.