In the above table we may first observe that the indices of segregation differ markedly from one municipality to another, and among municipalities with the same linguistic profile. Although intermunicipal residential segregation affects the linguistic segregation of pupils, there are other phenomena which clearly affect the uneven distribution of pupils in each municipality, independently of the concentration of Catalan speakers. The indices would appear to reflect considerable levels of segregation in a high proportion of municipalities. The contrast between linguistic segregation and segregation based on levels of education is particularly significant, as it enables us to see the magnitude of the issue. The conclusion is very clear: in general, levels of linguistic segregation in these municipalities are no less marked than those of social segregation. Furthermore, the intensity of linguistic segregation and social segregation tend to go hand in hand, so that municipalities with a high level of one type of segregation tend to record high levels in the other.(6)

These trends advise us, albeit indirectly, of the high probability that linguistic segregation in schools in the municipalities will be heavily conditioned by social segregation. This hypothesis is examined in greater depth below.

In addition to this first reading on a general level, we can make some more specific observations. The first is the fact that the highest levels of linguistic segregation are related to families who do use neither Catalan nor Spanish at home. In these cases, if we exclude municipality B, which has a policy for the distribution of pupils of foreign origin which has been agreed by all the schools, and municipality D, where the immigrant population is low, all the indices are notably high and reflect a clear tendency for immigrant pupils to be concentrated in a small number of schools. Secondly, in most municipalities, we detect a clear tendency for the concentration of the Spanish-speaking and/or Catalan-speaking population. In addition, in two of the municipalities where levels of segregation are lowest (H and I), we find a phenomenon which artificially reduces levels of segregation: some families with a medium-high socioeconomic status send their children to grant-maintained schools in neighbouring municipalities (we will return to this question in due course).

To explain the intensity of intra-municipal linguistic segregation, we must turn again to residential segregation. Residential segregation partly explains both linguistic and social segregation in the schools of certain municipalities (especially municipality G, with one of the largest populations in Catalonia, where there are substantial sociodemographic differences between districts and residential areas), but this is only a partial explanation, and, in some municipalities with high levels of segregation, a secondary explanation. Social segregation in schools is often not only an effect but also a cause of residential segregation, since the family's decision about where to live is conditioned by the sociodemographic environment of districts or towns, which then tend to become progressively homogenised. Residential segregation, even when it correlates with social segregation, (7) may not be so much a cause as an effect, especially in a historical context of great residential mobility, as has occurred in recent years.

It is true, furthermore, that parents' preference for enrolling their children at a school near home does not carry much weight if the local school is not considered satisfactory, especially among families with a higher socioeconomic (or academic) status. The family's choice of a grant-maintained school reveals clear limitations to the explanation of segregation by residential factors, especially the segregation of sub-groups associated with high status. (Benito and Gonzàlez, 2008). To the extent that there is a correlation between the linguistic and social composition of schools, the causal link between residential segregation and linguistic segregation is weakened. Finally, many small and medium municipalities have a catchment map for schools which allows, or at least would allow, policies for assigning pupils to schools which are very little influenced by residential segregation.

As the foregoing paragraphs suggest, an understanding of the intensity and the morphology of linguistic segregation in schools calls for a study of the extent to which social and linguistic segregation in schools are linked. Before undertaking this, however, we present a contingency table which clearly illustrates the levels of linguistic segregation which we have been discussing. This table shows the distribution of Catalan-speaking pupils among schools, based on a grouping of the schools by quartiles. In the first quartile we find schools with less than 25% of Catalan-speaking pupils, in the second schools where there are between 25% and 50%, in the third schools which have between 50% and 75%, and in the fourth those schools where Catalan-speaking pupils account for over 75%. We shall also refer to the third quartile as the quartile of suitability, as it is the one reflecting what we would consider a priori to be the minimum proportion of Catalan-speaking pupils needed to generate everyday use of Catalan in the school's informal activities.(8) The quartile in which we find the mean percentage of Catalan-speaking pupils for the municipality will be referred to as the mean quartile.

Table 3. Percentage of pupils in schools according to linguistic composition (by quartiles)

Linguistic composition of schools (% of Catalan speakers)	Type of municipality
	Spanish-speaking municipalities	Bilingual municipalities	Catalan-speaking municipalities	Total
0-25%	44,3	14,1	5	25,2
25-50%	35,1	15,3	10,6	22,8
50-75%	13,2	42,7	46,6	31
75-100%	7,3	27,9	37,9	21
Total	100	100	100	100

The percentage of Catalan speakers in the different municipalities ranges from 34% to 72%. The fact that schools with less than 25% (first quartile) and over 75% (fourth quartile) of Catalan speakers together constitute 46% of all schools in itself indicates a significant imbalance in the distribution of pupils. However, to be analysed with precision, the data must be set against the linguistic composition of the municipalities. Collective data is provided for all types of municipality, as it corresponds to relatively homogeneous profiles of linguistic distribution, enabling us to give a more synthetic reading of the different scenarios we can find in Catalonia. If we consider the data for each group of municipalities, the imbalances are still marked. The factor which best reflects this is that in all three types of municipality less than 50% of pupils attend schools in the respective mean-quartiles (second quartile in Spanish-speaking municipalities, quartile of suitability in the others), these being the quartiles where we would find 100% of the pupils in the municipality if they were evenly distributed.

In municipalities classified as Catalan-speaking and bilingual, we find a similar situation. In both cases the mean quartile and the quartile of suitability coincide, meaning that all pupils could be attending school in language environments which clearly favour familiarity with Catalan as a language for informal communication. Even so, a significant minority of pupils attend schools where Catalan occupies a more dominant position (fourth quartile), (9) while another significant minority attend schools where Catalan speakers are in the minority: 15.6% in Catalan-speaking municipalities and 29.4% in bilingual municipalities.