In an economy based on hunting and gathering in temperate domains, collective mobility appears to be an essential facet of economic behaviour for humans during Prehistory. In order to contribute to the definition of the archaeological forms of collective mobility in the scope of a programme on the functioning of Mesolithic societies in Brittany, it seemed essential to record a rock shelter occupation, liable to offer a different spectrum of activities to the large, now well-known sites. The Pont-Glas rock shelter at Plouneour-Menez (Finistere) provides good conditions for assessing this question. Two granite blocks leaning against each other protect a surface of about twenty square metres, forming a cavity of about fifteen square metres with two (east and west) entrances. A recently fallen granite block initially closed the shelter and facilitated the installation of a cover. This shelter was discovered and surveyed by M. Le Goffic in 1987, who recognized a reworked level containing flint from the Final Mesolithic and Gallic pottery, above stones described as “paving”. In 2007 and 2008, the whole shelter was excavated. The very narrow western entrance does not contain any traces of human occupations or circulation, but the shelter itself contains a carefully paved hearth with eroded edges (structure 1), then about fifteen centimetres lower down, a spread of charcoal probably corresponding to a flat hearth (structure 2). In the same place, but in another Mesolithic SU lower down, the substratum is directly affected by trampling, indicating another zone of human occupation (structure 3). This sequence only extends over a thickness of half a metre and contains various disturbances. However, after the spatial analysis of the remains, it appears clear that the base of the arenaceous SU (lower SU 5.3, 5.6, 5.9) corresponds to the Mesolithic occupation, with hardly any traces of protohistoric disturbances. The spatial analysis of the remains reveals several interesting insights, in spite of the very marked disturbance of the site by humans and burrowing animals. We observe a concentration of lithic objects under block 2, with a partial time lag between the early and recent Mesolithic elements. Surprisingly, this distinction is also perceptible from a vertical perspective, with a lot of interference linked to reworking. In the same way, the La Tene pottery is confined to the upper SU, if we exclude the intrusions along blocks resulting from wall effects. The lithic material consists of 998 elements and includes an early component from the 8th millennium with narrow scalene triangles, narrow backed bladelets and points with retouched bases, and a recent component with symmetric trapezes. The latter dates from the second half of the sixth millennium before our era, as confirmed by a date on charcoal from the last level. A small pick in microquartzite from Forest-Landerneau evokes the Mesolithic groups from Lower Normandy and from the Paris Basin from the 8th millennium before our era; this is the first prismatic tool discovered in Brittany. The analysis of the chaines operatoires shows a wide diversity of rocks and the rarity of the initial debitage phases, whereas arrow head repair phases are very well represented. The anthracological analysis focused on four stratigraphic units: SU 5.6, 5.10/6 and 5.11, attributed to the Mesolithic, and the upper part of SU 5.3 dating from the Second Iron Age. The environment contemporaneous with the occupations of the rock shelter is an acidophile oak stand with holly. However, the observed taxonomic associations reveal incursions in other types of environments. It thus seems likely that the wood was selected for specific uses. In the Mesolithic stratigraphic units, the observation of small charcoals with a last growth ring made up solely of initial wood points to a springtime occupation of the shelter. Although the low sedimentary dynamics of a shelter beneath a granite block by no means ensure the integral conservation of occupations, the results obtained are particularly important for understanding the last hunter-gatherer societies. Indeed, it is the first time in the West that fractioned chaines operatoires in time and space have been observed for the Mesolithic. In addition, the number of armatures destroyed by use is far higher than for large regional Mesolithic settlements. On the basis of these two parameters, Pont-Glas is a unique example of a temporary hunting halt, where several individuals came to repair their arms, before leaving again with their cores and knapped tools.