In environments poor in vegetal resources, the acquisition of the hunter-gatherers’ alimentary and technical resources largely rests upon the success of the hunting activity; this activity thus possesses a structuring role in their techno-economic organisation. Identifying the hunting weaponry, understanding its design and its management is therefore particularly informative on the dynamics of the Palaeolithic societies, whose economies were likely mostly based on the exploitation of animal resources. This reflection is based on the detailed functional analysis of several large assemblages of microliths from the Middle Magdalenian (19.5-16 cal ka BP) of Tastet cave, Blanchard cave, and La Marche, and on a comparison with the available data on antler points from these sites. This comparative approach allows an in-depth reflection on the hunting weaponry for a period of the recent Palaeolithic marked by regional variations of certain productions and hunting purposes: the Early Middle Magdalenian (EMM; 19-17.5 cal ka BP) and the Late Middle Magdalenian (LMM; 18-16 cal ka BP) are characterised by the regionalisation of certain types of microliths and osseous projectile points, species of ungulates hunted, as well as art and ornament productions; these elements allow recognising several technical traditions in France and northern Spain. The archaeological assemblages considered here illustrate one part of the diversity of this period, as they are associated with distinct traditions and yield different morphologies of microliths and osseous points—two categories of artefacts often associated with the hunting activity, and which may have been used together as composite projectiles. La Marche is at the heart of the definition of the EMM tradition “with Lussac-Angles points” (notably characterised by the production of eponymous short, single-bevelled, slotted points and truncated backed bladelets), Blanchard cave is emblematic of the EMM traditions “with navettes” (notably characterised by long, double-bevelled, slotted points, and truncated backed bladelets), and Tastet cave is associated with the LMM tradition with scalene triangles of the northern slope of the Pyrenees, and yields non-slotted, single-bevelled points. Results allow discussing the economy of hunting weapons, offering hypothesis regarding the weapons’ design, and the potential reasons of design variations. The functional analysis shows that the use-wears on microliths and on certain osseous points are due to the impact and indicate the use of the microliths as projectile inserts: at La Marche, 34% of the 181 analysed microliths have a diagnostic impact fracture, 32% have lateral impact scars and 0.5% have linear impact traces. At Blanchard cave (layers B2-B6), 51% of the 173 analysed microliths have a diagnostic impact fracture, 35% have lateral impact scars, and 0.6% have a linear impact trace. At Tastet cave, where the sampling strategy is not based on the functional potential (as it is in the two previous sites) but includes most microliths excavated between 2013 and 2018, the frequency of damage is closer to the experimental patterns: 2% of the 126 microliths analysed in stratigraphic units 206a and 306 have an impact fracture, and 11% show lateral impact scars. Data from the microliths and projectile points analysis indicate that the projectiles were brought back to the sites after the hunting episodes, and that the damaged lithic or osseous inserts were then replaced if necessary. The accumulation of damaged lithic inserts and of osseous points indicate that these repairing sessions were recurrent. The high standardisation of osseous points (e.g., morphotype of Lussac-Angles, of la Garenne, of the north slope of the Pyrenees), and of microliths (e.g., truncated backed bladelet, scalene triangle), must have allowed the different components of composite projectiles to be interchangeable. The analysis of the orientation, location, and frequency of impact damages on lithic inserts and their comparison to experimental data suggest that microliths from the three archaeological assemblages were likely positioned laterally (i.e., insert distant from the penetrating point of the projectile) or disto-laterally (i.e., insert in direct proximity to the penetrating point) rather than axially (i.e., insert positioned as projectile point). The comparison of the specificities of the impact damage shows their high homogeneity among the different morphotechnic categories of microliths (i.e., longitudinal cutting edge, monopoint, double-point, non-slashing rectangle) at the intra-site scale. There are however substantial differences at the inter-sites level regarding the specificities of impact fractures: bending initiated fractures and burin-like fractures are the most numerous fracture types at La Marche, bending initiated fractures dominate the small set of impact fractures at Tastet cave, whereas spin-off fractures are by far the most numerous ones at Blanchard cave. These differences refer to distinct modalities of application of impact forces, and at this stage of the methodological developments, the hypothesis of variations of propulsion type cannot be verified. Without ruling out this possibility, we suggest that the differences of impact damage are rather correlated to a combination between the differences of projectile point morphologies and different hafting modalities of lithic inserts: the short single-bevelled points of Lussac-Angles have a groove shaped on one or both sides and the length of these grooves corresponds to the average length of the microliths of La Marche, suggesting that lithic inserts were generally isolated on a face of these points. The points from Blanchard cave are longer and the groove on their side are deeper and longer, and several juxtaposed lithic inserts can potentially be inserted in them. This hypothesis is consistent with the predominance of spin-off fractures on microliths, which can be due to the constraints of the axial hafting or, more likely here, to the juxtaposition of lateral or disto-lateral lithic inserts. The size of the single-bevelled points of Tastet cave is intermediary between the two other sites, and their calibre is closer to the Lussac-Angles points. The absence of lateral grooves may refer to a multitude of hafting modalities of lithic inserts, and at this point it is not possible to precisely define how inserts were hafted. These inter-sites differences in projectile designs can be due to multiple factors, including environmental ones (e.g., vegetal cover, climate, season, game type and availability), the efficiency sought from the projectiles (e.g., in terms of penetration depth or width of the slashed wound), or technical styles (e.g., related to cultural identities or technical transfers). Our data underline that the regional and chronological diversity of the Magdalenian is expressed in many fields, and among those fields, the variations of the hunting weaponry is particularly important as it has a key role in the economies of hunter-gatherer societies during the Palaeolithic. The joint approach of the lithic and osseous industries dedicated to the hunting activity thus participates in renewing the understanding of the complex rhythms of technical change among the Late Glacial societies of western France.