Indeed, in those municipalities with a more homogeneous linguistic composition, we find educational contexts in which the connection between language and social composition is accentuated. Where Spanish is the majority language, the minority of schools with a substantial number of Catalan-speaking pupils have a stronger family educational background than in other municipalities. However, in Catalan-speaking municipalities, the connection is accentuated in the schools with fewest Catalan-speaking pupils, where the family educational background is poorer than at schools with a similar linguistic composition in other municipalities.

Finally, we present a table to verify whether, irrespective of these differences in level of education, there are also differentiated patterns of distribution according to the parents' level of education among pupils who are Catalan-speaking, on the one hand, and those who are not, on the other. For this purpose we have separated the variable for level of education into figures for Catalan speakers and for non-Catalan speakers.

Table 5. Percentage of pupils in schools according to linguistic composition (by quartiles), family educational background and language used by pupils at home

As can be seen in Table 5, in Catalan-speaking municipalities, Catalan-speaking families are distributed similarly among the different quartiles for schools irrespective of their level of studies. Although those whose studies go beyond compulsory education tend to go more to schools with a majority of Catalan-speaking pupils, the differences are not significant. However, this is not the case in Spanish-speaking municipalities, where we do find marked differences: Catalan-speaking families with higher studies send proportionally more of their children to schools with Catalan-speaking majorities than do Catalan-speaking families with a more limited educational background. Once again, we observe the mechanism of "withdrawal" in these contexts, among the Catalan-speaking families with higher status.

In the case of non Catalan-speaking families, the logic tends to be reversed. In Catalan-speaking municipalities we find significant differences according to the level of education, while in Spanish-speaking municipalities the differences are smaller. In municipalities with a Catalan-speaking majority, pupils who do not use Catalan at home are distributed differentially according to their parents' level of education, which was not the case with Catalan-speaking families. Families who are not Catalan-speaking, with a low level of studies, have a stronger relative presence in less "Catalanised" schools than families with a higher level of education. However, in Spanish-speaking municipalities, although there are some differences according to level of studies, they are minor. The main difference is the greater presence of less educated families in the bottom quartile, notably greater than that of families with a better educational background.

Another feature of note in this table is that in bilingual municipalities we also find differences according to the level of education, both among Catalan-speaking pupils and those who are not Catalan-speaking. However, these differences are less marked than those occurring in more homogeneous contexts among the minority-language group.

We can thus see that the relationship previously analysed between the social composition of schools and their linguistic composition tends to be more pronounced among the families who are in a linguistic minority in each municipality. Thus, in Spanish-speaking municipalities, there is greater differentiation in the distribution of Catalan-speaking families among the quartiles, according to their level of education, than there is among families which are not Catalan-speaking (their presence being greater in more "Catalanised" schools, the higher the level of education). In Catalan-speaking municipalities, on the other hand, it is the non Catalan-speaking families whose distribution is differentiated according to the level of studies. Their presence is greater in schools which are less "Catalanised", the lower their level of education, while in schools with a Catalan-speaking majority we find a higher proportion of non Catalan-speaking pupils whose status is high than those of lower status. In brief: in linguistically homogeneous municipalities, the differences in distribution according to level of studies are greater among families in the linguistic minority group than in the majority group, which tends to have a distribution which is affected less by levels of education.

4. Conclusions

The data presented in this article has shown significant levels of linguistic segregation in the school system in the Catalan municipalities studied. However, levels of linguistic segregation are similar to those of social segregation, evidence that it is spontaneous, a consequence of the processes of residential segregation and the family's choice of school, which are widespread in the West, even in those education systems which, like ours, are not based on education policies which lead to segregation.(11)