A decomposition of the learning gap between indigenous and non-indigenous 6th grade Mexican students is carried out, in order to identify which factors, explain it. Four groups of factors are contrasted: socioeconomic (student and his/her family), school resources, cultural relevance of the educational model and discrimination at the school level, and test language. Multi-level regression models and Oaxaca-Blinder decompositions are estimated for Spanish and Mathematics results. Hypotheses are developed in order to allow indirect estimation of the incidence of the last two. The results show greater effects of socioeconomic factors than school factors; a significant reduction of the gaps in both knowledge areas (55-70%) as a consequence of these factors; A greater proportion of the gap explained in mathematics than in Spanish; practically nonexistent interaction effects between school type and the student’s ethnic condition. Explanations that emphasize the relevance of the socioeconomic marginalization of indigenous students are strengthened, against the role of school resources, the irrelevance of the educational model or discrimination mechanisms. A lower proportion of unexplained gaps in mathematics strengthens an alternative, more parsimonious hypothesis: a significant part of the gap derives from the fact that indigenous students are not evaluated in their first language.