This interdisciplinary project mixes research in the humanities and computational sciences to understand and explain the role of visual-motor exploratory strategies and lived experiences (e.g. individual concerns), in analysis of learning dynamics of fire-fighters. Traditionally, learning is studied by comparing behaviour (especially outcomes) before and after practice, which does not facilitate understanding of transitional phases during the process of learning, where new behaviours can temporarily alternate with previous behaviours, in an ‘intermittent regime’. First, this interdisciplinary project seeks to model transitions between the three phases of learning dynamics (initial, intermittent, and final) to understand how the frequency of novelty (non novelty vs. prescribed novelty vs. chosen novelty) of learning situations influences the intermittent regime. We based our statement on recent studies suggesting that self-controlled choice in practice might enhance learning. We will test the hypothesis that the duration of the intermittent regime is associated with the ability to pick up opportunities for actions (affordances) offered by the environment. Conducting this project in an ecological performance context raises a scientific challenge for computational sciences, going beyond the climbing task; distinguishing head movements from eye movements to locate visual fixations within the global scene (not on the local scene viewed at a given time by the individual). The first aim for the computational sciences, in terms of signal and image processing, will be to identify visual fixations in the global scene (using the techniques of SLAM: Simultaneous Localization And Mapping). Although studies on learning have provided evidence about behavioural changes, it was recently suggested that how these behavioural changes are experienced may be important to understand how an individual constructs meaning about an activity. Fire-fighters are suitable to investigate these aspects of learning for two reasons: (i) fire-fighters must face situations which are both unique and familiar (a fire in a home has features common to other home fires with a uniqueness due to the architectural features of the house). Therefore, fire-fighters must be adaptable to different emergency situations, while taking advantage of their past experiences. (ii) The nature of their practice activities can impact upon their ability to save lives. Training in a climbing task where they must experience and cope with novelty is a social issue for the effectiveness of interventions and the protection of fire-fighters. Based on these ideas, a second aim of this humanities project is to model the phenomenology (the experience) of the dynamics of learning as the novelty of the situations is imposed or chosen by each participant. In the computational sciences, a key scientific challenge is to achieve an integrative approach to record behaviour (vision and motor skills), allied to phenomenological data to determine for each individual the extent of an initial repertoire, which is explored during learning (by assessing the duration of the intermittent regime) and then stabilized at the end of learning. We will address the second aim by using unsupervised machine learning techniques such as clustering methods, designed to group individuals together regarding their degree of similarity, without having any a priori expectations on the number of groups and the features of these groups. Here the scientific challenge is to take into account the temporal dimensions of learning, behavioural and phenomenological characteristics and uniqueness of each individual. This modelling in the humanities and computational sciences will be used to design optimal practice environments to train fire-fighters by enhancing their adaptive behaviours, inviting them to safely explore novel, functional behaviours.