The Gravettian in Belgium is documented in a limited number of sites, mostly cave or rock shelter sites located in the Meuse basin, and whose excavations often began in the 19th century. Our knowledge of the Belgian Gravettian owes a great deal to the work of M. Otte, who compiled, synthesised, and structured a scattered and often imprecise documentation, enabling him to identify eight proven Gravettian sites, as well as numerous potential sites (Otte 1979; Fig. 1; Table 1). This list was subsequently extended to two additional sites, bringing the number of proven Gravettian sites in Belgium to ten (Otte & Noiret 2007). These are classified into three typological “facies” or “groups”, which together form the Belgian Gravettian sequence and whose “kinship” is attested by the recurring presence of certain tool types (tanged points, points with flat retouch/Maisières points). The knowledge accumulated over the last few decades has made it possible to approach certain aspects of this model from a different angle. The aim of this article is therefore to draw up a critical assessment of the Belgian Gravettian sequence, based on data from four major sites: Maisières-Canal, the caves and the Upper shelter of Goyet, and Station de l’Hermitage. Maisières-Canal yielded an occupation layer dated to around 32 000 cal. BP (de Heinzelin 1973; Haesaerts & Damblon 2004; Jacobi et al. 2010; Table 2), contemporary with the end of the GI-5.2 (Rasmussen et al. 2014; Fig. 2). The retouched tools include numerous burins and points. The other types represented are quite diverse and include several dozen tanged tools such as burins (Fig. 4D) or scrapers, but the backed pieces are rare and do not include any classic Gravettian type (Gravette points, microgravettes, etc.). Points include three morphological groups – Maisières points (Fig. 4E, F), tanged points (Fig. 4A-C) and shouldered points –, which are produced following a specific method that involves the detachment of flat and direct retouch flakes with an organic hammer, sometimes including an apical tranchet blow (Touzé 2018; Fig. 4A, B, E, F). The tool blanks are mainly derived from bidirectional laminar production on a wide surface, carried out with a stone hammer, the main objective of which is to obtain wide, moderately thick, and sometimes distally tapered products. Bladelet production is also attested, but it occupies a marginal place in the lithic technical system. The originality of this industry in comparison with later Early Gravettian assemblages (e.g. Station de l’Hermitage), as well as the identification of tanged points and Maisières points similar to those of Maisières-Canal in a dozen sites in Belgium (Table 3; Figs 8-9), France, the United Kingdom and Luxembourg, allows us to envisage the existence of a pre-Gravettian lithic tradition in Northwestern Europe around 32 000 cal. BP (Touzé 2019). This tradition – the “Maisierian”, as proposed by J. B. Campbell (1980) – should be dissociated from the Gravettian itself, because: 1) its lithic industry does not present the usual Gravettian features (e.g. Gravette points, microgravettes); and 2) it predates the oldest Gravettian sites recorded in Belgium and North-western Europe (Touzé 2019). Of the 14C dates available for Station de l’Hermitage at Huccorgne (Straus 2000), the most reliable result places the Gravettian occupation around 31 100 – 30 000 cal. BP: it could therefore be contemporary with GI-5.1 (Touzé et al. 2016; Fig. 2). The retouched tools mainly include burins and backed pieces, mostly microgravettes. Some tanged pieces are also recorded, but none of them display the use of the retouch method observed at Maisières-Canal. The tools are made from straight, thin and usually narrow blades obtained by direct percussion with a soft stone hammer, and from bladelets. Blade production is carried out from two opposite striking platforms on reduction surfaces with limited longitudinal convexity. Bladelets are extracted at the end of blade production, or on the edge of flakes or blades (burin-cores). The industry of Station de l’Hermitage share many features with two sites located in the Paris Basin: Ormesson – Les Bossats, dated between 31 800 and 30 000 cal. BP, and Flagy – Belle Fontaine (Touzé 2019). These features include blade and bladelet productions as well as the retouched tools, the latter including microgravettes and tanged points (Fig. 10I-L). Furthermore, a recent study has shown that siliceous raw materials collected in the Paris Basin were imported to Huccorgne (Delvigne et al. 2021). These three sites therefore document the same Early Gravettian lithic tradition, which we can assume is also represented at other Belgian sites, even if this is difficult to demonstrate due to: 1) the ubiquitous character of the main markers of this tradition (microgravettes, tanged pieces that are insufficiently diagnostic); and 2) the mixtures, and the imprecision of the contextual data that affect the sites in question. The only occupation layer identified at the Upper shelter of Goyet is attributed to the Gravettian (Éloy & Otte 1995) and is dated between 29 200 and 27 900 cal. BP, indicating a possible correlation with the GI-4 (Fig. 2). However, the presence of nosed end-scrapers indicates the presence of an Aurignacian layer not detected during the excavations and therefore the existence of mixtures among the archaeological material (Fig. 3). The collections from Goyet caves are known to be affected by significant mixtures (e.g. Otte 1979). Several dates corresponding to the Gravettian are available: these are situated between 31 800 and 26 000 cal. BP (Germonpré & Hämäläinen 2007; Posth et al. 2016; Fig. 2). Due to the mixtures affecting the sites of Goyet, the study focused on the characteristic Gravettian tools. Among these, three tool types – the backed points with truncated basis (Figs 6A-C; 7A-D), the bi-truncated backed pieces (Figs 6D-J; 7E-I), and the “bi-pointed” backed pieces (Figs 6K-M; 7J-M) – probably belong to one (or several) occupation(s) more recent than those recorded at Maisières and Huccorgne: these types are therefore considered to belong to a “post-early phase” Gravettian. They are also known in other sites in Belgium (Table 3; Figs 11A-G; 12A-E; 13A-C), but there is no evidence that they are all contemporaries in the current state of our knowledge. Outside Belgium, backed points with truncated basis are documented in the Recent Gravettian of the Paris Basin and its margins (Nouel 1936; Klaric 2003, 2013; Kildéa & Lang 2011; Pesesse & Flas 2013; Fig. 11H-M), and in more southern sites where their precise dating is difficult to establish (Bouyssonie 1948; Pradel 1965, 1979). Bi-truncated backed pieces are known in Southwestern France, notably in the Middle Gravettian of the Pyrenees (Simonet 2009; Fig. 12F-H, K, L) and in the Recent Gravettian of the Abri Pataud (Bricker & David 1984; Fig. 12I, J), as well as in the Late and Final Gravettian of Central Europe (Polanská et al. 2021; Fig. 12M, N), although their characteristics may be somewhat different. The bi-pointed backed pieces are atypical in Western Europe. Similar pieces exist at Pavlov I (Czech Republic; Klíma 1997; Svoboda 1997), even though not all of them have a backed edge (Fig. 13D-U), but their dating is unclear. Other comparable tools are also known at Grub/Kranawetberg (Austria; Nigst & Antl-Weiser 2012) and Kostënki 8 (layer 2; Russia; Reynolds 2014). It is possible that some of these similarities with the Belgian sites can be explained by morphological convergences. The caves of Goyet also yielded several tanged points (Fig. 8A-E) and Maisières points (Fig. 8G, H), while a tanged piece was found at the Upper shelter (Fig. 6O). Several of these tools have obvious technical affinities with the Maisières-Canal industry, while some of the less diagnostic tanged pieces can be compared with the specimens from Station de l’Hermitage, as well as with some of those from Maisières-Canal. This suggests that the sites of Goyet were probably frequented at different times during the Gravettian, as was the case with Spy Cave (Pesesse & Flas 2013). The Mid-Upper Palaeolithic of Belgium can therefore be subdivided into three entities, as identified by M. Otte. The Maisierian and Early Gravettian correspond to relatively well-documented lithic traditions whose ranges extend over part of North-western Europe. A “post-early phase” Gravettian is also detected, but only within assemblages whose integrity is questionable, or which are quantitatively limited; only the discovery of new sites will allow us to acquire more precise data on this third phase. The suspicions that weigh on the integrity of several assemblages lead us to update the census of these entities with regard to the initial model (Otte 1979; Tables 3; 4), but also to abandon the principle of “kinship” that would link them within a “Gravettian phylum”, since the coexistence of “ancient” and “recent” tool types in some assemblages – all of which come from cave or rock shelter sites – is likely to be the result of the mixing of several Gravettian and/or Maisierian components. Finally, it should be remembered that the supposed evidence of a Middle Gravettian with Noailles burins and Raysse burin-cores presence in Belgium (Otte 1976, 1979) do not seem reliable (Touzé 2011).