The Relevance of Language as a Predictor of the Will for Independence in Catalonia in 1996 and 2020

The Catalan secessionist parties, if added together, have won all the elections to the Parliament of Catalonia from 2010 to 2021. Their voters have been increasingly mobilized since the start of the controversial reform process of the Statute of Autonomy (2004–2010). The aim of this article is twofold. First, it intends to test whether language is the strongest pre‐ dictor in preferring independence in two separate and distinct moments, 1996 and 2020. And second, to assess whether its strength has changed—and how—between both years. Only the most exogenous variables to the dependent variable are used in each of two logistic regressions to avoid problems of endogeneity: sex, age, size of town of residence, place of birth of the individual and of their parents, first language (L1), and educational level. Among them, L1 was—and still is—the most powerful predictor, although it is not entirely determinative. The secessionist movement not only gathers a plurality of Catalan native speakers, but it receives a not insignificant level of support among those who have Spanish as their L1. Conversely, the unionist group, despite being composed primarily by people who have Spanish as their L1 and have their family origins outside Catalonia, has a native Catalan‐speakingminority inside. This imperfect division,which is basedon eth‐ nolinguistic alignments—and whose relevance cannot be neglected—alleviates the likelihood of an ethnic‐based conflict.


Introduction
Secessionism, since the first successful declaration of independence was signed in 1776, is present in the political reality worldwide. Indeed, between 1945 and 2011, an average of 52 active secessionist movements per year has been identified (Griffiths, 2015). In parallel, there are two major approaches to analyse them: comparative studies and case studies. The latter is the option taken here, with the intention of bringing some clues to understand the growth of secessionism in Catalonia in recent years.
The Catalan case has received some attention because it is a secessionist process within a consolidated democracy, and within a relatively heterogeneous society, in linguistic and cultural terms. This heterogeneity is the result of several waves of immigration over the last decades. As a result, 36.4% of the population living in Catalonia were born outside of Catalonia (Institut d'Estadística de Catalunya, 2020). In this context, the Catalan language, a traditional marker of "Catalanness," went from being the first language (L1) of the vast majority of the Catalan population in 1900 (de Rosselló Peralta et al., 2020), to becoming L1 of only 31.5% of the population. Nonetheless, it is the reported language of identification of 43.2% of the population, while 94.4% say they understand Catalan and 81.2% can speak it (Institut d'Estadística de Catalunya, 2018).
At the beginning of the democratic period in Spain, the percentage of secessionists was relatively low (in 1979 it was 8.6%, according to Centro de Investigaciones Sociológicas [CIS], 1979). Between 1990 and 2010, it fluctuated around one-third of the population (see CIS, 1996CIS, , 2001; Institut de Ciències Polítiques i Socials [ICPS], 2020). From 2011 to the present, the percentage has been around 44% (see Centre d'Estudis d'Opinió [CEO], 2021a[CEO], , 2021bICPS, 2020). Therefore, in a few years, the growth of independence has been quite important, and it must be explained in the context of the political events that have taken place in Catalonia (Argelaguet, 2014;Casas et al., 2019;Colomines i Companys, 2020;Orriols & Rodon, 2016;Rico & Liñeira, 2014).
In this sense, the fact that the percentage of secessionists in Catalonia, either in the polls (about 44%) or in the electoral results (in the 2021 elections, the combined pro-independence parties obtained 51% of the votes), is striking because it is higher than the percentage of people who have Catalan as their L1 (about 32%) or even the percentage of those with Catalan as "language of identification" (43%). This difference could be considered an outcome of the decision made by the main Catalan nationalist parties to promote a political discourse whose aim is to go beyond the perimeter of its core ethnic group, Catalan native speakers. The main leaders of the Catalan secessionist movement-Artur Mas, president of the Catalan government and Oriol Junqueras, president of the Republican Left of Catalonia (ERC; Argelaguet, 2011)-have underlined, for instance, that in an eventual independent Catalonia, the Catalan and the Spanish languages will be official languages (Barbeta, 2012;Junqueras, 2012). This compromise, however, generated a great debate within the ranks of the secessionist movement with hard-fought positions (Sendra, 2016), and links with the more academic debate on the scope of the civic component that has been historically detected in Catalan nationalism (Conversi, 1997;Keating, 1996;McRoberts, 2001). However, other authors have provided data that would seriously question this thesis (Álvarez-Gálvez et al., 2018;Miley, 2007).
The aim of this article is to check whether there has been a change over time in the role of the language in explaining the secessionist preferences that exist in Catalonia. The methodological option followed is to make a comparison of the data of two surveys conducted at two very different moments in time and in very different political contexts: 1996 and 2020. 1996 was the highest point of collaboration in the governance of Spain by the moderate Catalan nationalist force Convergència i Unió (CiU; Barrio & Barberà, 2011). On the other hand, in 2020, secessionism had most seats in the Parliament of Catalonia.
For each year, I have used strictly the same model of a logistic regression, with the same variables, to see the elements of change and continuity, and whether the exacerbation of the conflict has produced changes in the internal composition based on linguistic groups of the two opposing poles: the secessionist and the non-secessionist. The variables used are the maximum possible exogenous to the dependent variable (the percentage of supporters of independence) to avoid problems of endogeneity or inverted causality: sex, age, size of town of residence, place of birth of the individual and of their parents, L1, and educational level.
Therefore, the proposed analysis allows increasing knowledge about a specific case that has aroused some interest in recent years, while providing data that will contrast the Catalan case with other cases and, perhaps, improve the knowledge that has been generated with their comparative analysis. The article will proceed as follows: In the next sections, I will present some theoretical considerations about the study of nationalism and secessionism. Then, I will present the data and methodology, show the results and address their discussion. I conclude by pointing out some clues to understand the framework through which the political debate in Catalonia can take place in the coming years.

Theoretical Background
The study of secessionism is linked to the study of nationalism. Inside the latter, there is a very relevant debate, propelled from the seminal work of Hans Kohn, about the existence of two types of nationalism, civic and ethnic (Kohn, 1944). The first one is based on the idea of inclusiveness, so that an individual's belonging to the nation is linked to voluntary elements, including adherence to legal norms. In contrast, the latter assumes that being a member of a nation is due to ascriptive elements, including ancestry, blood inheritance, or customs. This dichotomy, despite having been widely used, is criticized for being more normative than descriptive (Tamir, 2019) because its two components-"civic nationalism" and "ethnic nationalism"-are loaded terms and, moreover, "ethnic nationalism" is used pejoratively (Yacobson, 2013); others consider that is perhaps not a dichotomy but a continuum (Smith, 1991) because these are concepts with blurred borders that, in addition, are overburdened by usage, by the actors in the political conflict (Brubaker, 2004); in this case, this distinction-described as "Manichean" because one is good and the other bad-is "both normatively and analytically problematic" (Brubaker, 1998, p. 274).
Despite all this debate, this typology remains widely used in quantitative analysis, as is the case of the studies based on questions on national identity that have been included in the International Social Survey Program over several years. Using factor analysis, two dimensions have been identified from the answers given to the question of how important some elements are to define one's own national identity. The question is posed as follows in the questionnaire: "Some people say that the following things are important for being truly [nationality]. Others say they are not important. How important do you think each of the following is?", and the respondent was then asked about the importance of: (a) having been born in country X; (b) having country X's citizenship; (c) having lived in country X for most of one's life; (d) being able to speak country X's language; (e) being of a certain dominant religion; (f) respecting country X's nationality, political institutions, and laws; and (g) feeling a certain nationality (International Social Survey Program, 2015).
This empirical approach raises some problematic issues: There is a great diversity of interpretations of data (Jayet, 2012); there are discrepancies about how to label the dimensions that were found; and there is no same constant classification for all countries of the seven items in their respective dimensions. Moreover, there is a quite important level of association among all the items that can blur the dimensions. Nonetheless, items that include "to be born," "to have country's citizenship," "to have lived," and "religion" are usually linked to an ethnic dimension; and items "language," "to respect political institutions," and "to feel" link back to a civic dimension (Larsen, 2017). In addition, however, some items are not easily attributable to the "civic" or "ethnic" dimension, as is the case of "language," "an element of ethnic nationalism (a heritage of a culture) and, at the same time, it belongs to civic nationalism as an instrument of communication and of the participation in political institutions" (Jayet, 2012, p. 72). This debate is also present in the Catalan case (Álvarez-Gálvez et al., 2018). On the one hand, the Catalan language is the most obvious ethnic cultural marker of Catalan identity but, on the other hand, the language can be learned and become a source of social integration so that language barriers become permeable, which means that the language can be clearly distinguished from other ascriptive attributes, such as familial descent (Miley, 2007).
When the object of study becomes more focused on the analysis of secessionism, the initial point of reference is the seminal contribution of Donald L. Horowitz, who argues that the secessionism is explained in the framework of the intersection of ethnic identity and the socio-economic development of the groups involved. In poorer regions, through elites who exploit the resentment of the masses, secessionist ideologies are more likely to develop (Horowitz, 1985). At the same time, he shows that the relationship between ethnic diversity and severe ethnic conflict is nonmonotonic, with less violence for highly homogeneous and highly heterogeneous countries.
From here, the analysis of secessionism is devoted to answering several questions: Which are the factors that activate secessionism-once it is assessed that the cultural identity is very important but it is not determinant (Sorens, 2005)-given that there is a widespread range of competing hypotheses revolving around economic, cultural, and political factors? Does decentralization-and if so, under what circumstances-calm or ignite secessionism (Brancati, 2006)? Is secessionism (or, more broadly, ethnolinguistic mobilization) associated with violence (Brubaker & Laitin, 1998)? Is language conflict a pre-scription for violence, or rather can language conflict, "under certain potentially incendiary conditions…help to contain violence" (Laitin, 2000, p. 98)? Finally, does the probability of conflict depend on ethnic diversity, as measured by the index of ethnic fractionalization (ELF). Despite it being widely used, the ELF receives some criticism: It is difficult to count all the groups, and it does not take into account the cultural distance among groups (Fearon, 2003), its historical evolution (Drazanova, 2020), or the polarization between them (Esteban et al., 2012).
However, with the intention of overcoming the shortcomings of the analyses compared to a very large N, it may be convenient to resort to case studies to test general propositions in more detailed contexts and explore what are the elements that explain the growth of secessionism as, for instance, the present case of Catalonia.
In the Catalan case, like in other cases, secessionism can be analysed with electoral and other sociodemographic aggregated data or using opinion polls. With survey data, one can identify the strong predictors for independence, among them there is the national identification of the individual, also commonly referred to as subjective national identity (SNI). Because this variable is widely present in most of the analysis, it is helpful to clarify the problems associated with its use.
The SNI is operationalized through the so-called "Linz-Moreno question" (Moreno, 2006), which, despite its great academic use, has some problems: It does not adequately capture the identity intensity or linearity with national feelings and collects the existence of a large, too heterogeneous intermediate group (Guinjoan & Rodon, 2016). In addition, the Catalan case has the problem of multiple meanings of the terms "Spanish" and "Catalan" (both can have a national or administrative meaning); and that of not being formulated in an excluding way (Cussó et al., 2018). However, the use of this scale is very common and has generated a large amount of data that allows comparative analysis in space and time.
The SNI in Catalonia has changed over the years. These changes have been more important between 1991 and 2006 than between 2010 and 2020, the years of the acceleration of the bid for independence (Table 1).
Moreover, Serrano (2013) has proposed going beyond the sociodemographic factors that are given by birth, emphasizing the intervention of other relevant variables linked to nation-building policies, associated with changes in the institutional context that resulted from devolution arrangements, as the independent effect of media consumption in Catalan or, even, the support for fiscal autonomy. Despite all these elements, L1 is the key predictor for explaining SNI and, by Notes: The usual five categories have been collapsed into three; 1 In 2020, it is the merger of the CEO's three surveys: No. 962, no. 974, and no. 985. Sources: CEO (2006, 2010, 2011, 2020a, 2020b, 2020c, CIS (1996), ICPS (1991).
extension, the preference for independence (Chernyha & Burg, 2012). Therefore, even though it is a verified fact that the older generations have been replaced by more proindependence generations (Bartomeus, 2018), the significant increase in secessionism from 2010 onward would not be explained by the small changes in the SNI. For this reason, other complementary hypotheses have been raised, such as risk aversion by gender (Verge et al., 2015), or, above all, that the growth of secessionism would have been driven by political elites in an outbidding process (Barrio & Rodríguez-Teruel, 2017). As an alternative to this top-down hypothesis, others raise the bottom-up one, that is, there would have been a very intense pro-secession popular mobilization that, through large demonstrations or holding popular consultations (Muñoz & Guinjoan, 2013), would have pushed the parties, from 2008, to act in favour of independence (Dowling, 2014).
Without being able to ignore the weight of the SNI, other analyses have focused on fiscal policy preferences (Boylan, 2015), on individual exposure to the effects of the global economy (Hierro & Queralt, 2020), and on the concurrence of instrumental and welfare maximizing reasons . However, with aggregate or individual level data, others state that there would be no relationship between crisis and the rise of secessionism (Bel et al., 2019;, or find that the pro-independence vote in 2017 is more tied to individuals' place of birth and that, in contrast, typically economic predictors (GDP per capita, poverty level, or unemployment level) are not relevant (Maza et al., 2019).
In conclusion, to explain secessionism in Catalonia, instead of re-including the SNI due to measurement and endogeneity problems with the dependent variable , I will use the variables that are as exogenous as possible to it: L1, individual's place of birth and their parents' place of birth, gender, age, size of town of residence, and level of education. It has been ruled out to include the ideological variable, measured through self-location on the left-right axis, because in Catalonia it carries an endogenous component linked to the national question (Dinas, 2012).
In consequence of the previous exposition, the hypotheses that are going to be tested are the following ones: H1: Language is one of the basic features defining an ethnic group (Fishman, 1999, p. 4); it is able to generate a nationalist movement and even propose a secessionist project; is the most relevant predictor of the socio-demographic variables considered. So, in the case of Catalonia, and in both years under analysis, those citizens with Catalan as their L1 will have more probabilities of being in favour of independence than those citizens whose L1 is Spanish.
H2: Given that it is well established that the escalation of a national conflict-like one related to independence-may lead to a growing social divide alongside language groups, it is expected that the weight of language as a predictor of secessionism in any person will increase its importance if the context becomes politically agitated. So, in the case of Catalonia, between 1996 and 2020, the likelihood of wanting independence if the L1 of this person is Catalan (in contrast to if their L1 is Spanish) will increase. Specifically, it should be noted that the growth of secessionism will be proportionally stronger among Catalans with Catalan as their L1 than among the rest of the people. If this happens, it will be evidence that the independence process would have reinforced the social division based on the linguistic alignment of individuals.
H3: The growing warming of the political debate in a context where the language factor is relevant must imply that each language group will be progressively aligned with each option (Catalan native speakers with independence and Spanish native speakers with non-independence). The outcome will be that in both blocs-for and against independencetheir language diversity will be reduced, measured by applying the formula for effective number of parties to calculate the effective number of language groups (ENLG). In this sense, ENLG = 1/ ∑ n i=1 p 2 i , within a society or a group of people containing several language groups, where p i is the language group proportion in the society (or other grouping), according to their L1.
However, because in the case of Catalonia there is enough evidence that the secessionist movement has intended to avoid the risk of social division by underlining that the current linguistic rights of all will not be changed in an eventual Catalan republic (where Catalan and Spanish will be official languages), it will be expected that this homogenization will not only not grow within the secessionist bloc, but, instead, its internal diversity in language groups will increase. This fact could be interpreted as a relatively successful result for the proindependence movement of having appealed in its political discourse to elements contrary to the division of society by language alignments.

Data and Methodology
The data for this analysis comes from two face-to-face surveys. One is the 1996 Study 2228 of the CIS. The second is the merger of the three waves of the 2020 Political Opinion Barometer of the CEO. Despite CIS depending on the Spanish government and CEO on the Catalan one, there is no evidence of substantial bias in their data. They both follow rigorous validation mechanisms of the methodology for obtaining the data, which are widely used by the academic community thanks to their policies of transparency. Moreover, at CEO, the samples are designed considering cross-quotas of gender, age, and place of birth.
The dependent variable is the opinion regarding the independence of Catalonia, and it has been dichotomized ( Table 2).
The independent variables included in the logistic regressions are as exogenous as possible to the dependent variable, so that the sense of causality was unidirectional and endogeneity problems were avoided. The vari-ables are sex, age, size of town, individuals' place of birth, parents' place of birth, L1, and level of education. L1 is a trait on which the individual has no capacity for choice, and it is prior to the self-conscious formulation of one's own identity. As for the level of education achieved, despite being the result of a set of diverse factors, I have incorporated it into the analysis because lacking an appropriate measure of social class or socioeconomic status, education could be used as a proxy predictor of social economic status. The frequencies of these variables and their cross tabulation with the dependent variable are shown in Table 1A in the Supplementary File. In this table, two variables have been added, although they are not used in the analysis.
In the next section, I will show the results of the two logistic regressions. The same model has been used to compare the data in 1996 with that of 2020. In the independent variables, the base group is always the first category listed.

Results and Discussion
The two regressions return similar results in terms of the direction of the association between the categories of the independent variables and the dependent variable, except for sex and age group. Statistical significance is higher in the 2020 data. Pseudo-R 2 (0.280 in 1996 and 0.321 in 2020) and correctly classified cases (74.8% in 1996 and 73.8% in 2020) are similar (Table 3).
To clearly show the main findings, in Table 3 there are the odds ratios (the Exp(B) of both regressions) and their equivalence in probabilities of wanting independence in relationship to the base group (the first category in each variable).
Gender is associated differently in both years. While in 1996, being male reduces the chances of agreeing with independence, in 2020 it is the other way around and, in addition, it is with minimal statistical significance. It could be explained by the combination of the possible likelihood of the proximity of independence with risk aversion, more present among women.  As for the age group, in 1996, the chances of agreeing with independence increase in some categories with respect to the base group, but in others they decrease, as in the case of those over 74, the only one with statistical significance. It could be a generational effect: The individuals in this group were of military age during the Spanish Civil War. By 2020, the chances of wanting independence increase in all categories compared to the base group. The fact that this group, of those who have lived through the independence process in their teens, has the lowest percentage of secessionists of all, and the fact that the group of 55 to 64 year olds has the highest percentage of secessionists, would nuance the alleged role that has been played in the process of socialization by the media and the education system under the responsibility of the Government of Catalonia-which is in the hands of Catalan nationalists (Miley & Garvía, 2019;Tobeña, 2017). In any case, an analysis of the effects of age, cohort, and period on the will to independence, and considering the main effect of primary socialization will be needed. Thus, in 1996, among young people aged 18-24 and with Spanish as L1, those who agreed with independence were part of the 20.3%. In this same subgroup, in 2020, the pro-independence individuals are 20.7%. But, among young people aged 18-24 and with Catalan as L1, those who agreed with independence are part of the 54.8%. In this same subgroup, in 2020, the proindependence individuals are 67.2%.
The size of the town of residence is also associated with the will for independence in both years, although with different statistical significance. The larger the town is, the lower the chances of wanting independence compared to the base group are (the smaller towns). This circumstance is linked to the presence of more inhabitants from Spanish immigration in the big cities, most of them surrounding Barcelona, and in Barcelona itself.
Being born in Catalonia significantly increases the likelihood of agreeing or wanting independence. However, probabilities are slightly reduced in 2020. There is also an incremental positive effect when the individual's parents are both born in Catalonia (or at least one parent), compared to those who do not have any parent born in Catalonia. Therefore, the idea that primary socialization becomes a relevant factor as a predictor of secessionism is reinforced. However, secessionism has not grown evenly in both groups. Between 1996 and 2020, in the group of both parents born outside Catalonia, secessionism has gone from 16.2% to 23.3%, an increase of 43.8%. And in the group of those who have both parents born in Catalonia, secessionism has gone from 55.8% to 68.6%, a 22.9% increase.
This data confirms confirm that the most important predictor is L1. If one has Catalan as L1 is much more likely to want independence than if L1 is Spanish. However, the growth of secessionism has been uneven depending on the L1. Among those who have Spanish as L1, percentage has gone from 16.3% in 1996 to 24.6% in 2020, an increase of 50.9%. And among those who have Catalan as L1, secessionism has gone from 56.4% to 73.4%, 17 points, and an increase of 30.1% (Table A1 in the Supplementary File).
Because the marginal effects provide information on which factors contribute most to explaining the dependent variable, the average marginal effects of the exploratory variables are visualized in Figure 1 for ease of interpretation. I show the likelihood of independence given values on the explanatory variables, while holding all other variables at their means.
In 1996, in terms of average marginal effects, the probability of being in favour of independence is the greatest when the L1 is Catalan (23.9%), followed by when the respondent is born in Catalonia (12%) and when they have a high level of formal education (10.6%). On the contrary, when the respondent is older than 74 years, the likelihood decreases (14%).
In 2020, the likelihood of wanting independence is greatest when, again, the respondent's L1 is Catalan (29.6%), when their level of formal education is the highest (14.9%), when they are 55-64 years old (13.2%), when their parents are both born in Catalonia (12.3%), and when they born in Catalonia (10.8%). On the contrary, the likelihood of preferring independence is lowest when the town they are living in has between 50.000 and 150.000 inhabitants (16.3%).
In summary, to have the Catalan language as L1 has the largest effect on preferring independence in both years. However, it must be underlined that in both models where it is included, they only correctly classify 73-75% of cases (see Table 3). Therefore, it means that the rest of the cases (about 25%) should be classified with the concurrence of other variables not included in the model.
A more descriptive approach is useful to complete this analysis. With reference to From the crosstabulation of the L1 by the will for independence in 1996 and 2020 (in columns), the ENLG for the whole of Catalonia and for the pro-independence and unionist subgroups are calculated (Table 4).
In 1996, the ENLG for Catalonia was 2.15. For the unionist subgroup, it was 1.83 and for the secessionist one, it was 1.88. Both groups were very similar in homogeneity, while they seemed to confront each other like a mirror: more than 2/3 in the dominant group and 1/3 in the minority group, respectively.
the ENLG for the pro-independence subgroup rose to 2.06. Therefore, while unionism has become more homogeneous with the increase in the percentage of those who have Spanish as their L1, secessionism has become more heterogeneous because it has also penetrated this part of the population.
This double dynamic may show the results of the political strategy followed by the leaders of both blocks. It has been tested that Catalan political parties generally do not use the main identity marker in Catalonia (language) for outbidding purposes and, if they do, it is the unionist parties that resort to it more than the pro-independence parties (Sanjaume-Calvet & Riera-Gil, 2020). In this sense, in recent years, the unionists have increasingly used Spanish in the Parliament of Catalonia; and most pro-independence parties have incorporated some leaders who do not have Catalan as L1.
The growth of secessionism cannot be explained without taking into consideration that the secessionist leadership has promoted a political discourse that it is trying to become rooted in a civic use of the question of language, i.e., the desirability of the Catalan language to become the shared language of the inhabitants of Catalonia, while respecting the language rights of the Spanish speakers, and assuming that in the future Catalan Republic, Catalan and Spanish will have the status of official languages. Without this strong commitment (although it is contested within its own ranks), the secessionist project could be at risk. Catalonia nowadays is a subordinate political community. The plurality of its citizens has family origins outside Catalonia (54.2% of voters have no grandparents born in Catalonia; see Table A1). The L1 of the plurality of the Catalan electorate is Spanish (56.4%; see Table A1). With these two sociodemographic traits, it is wise for the secessionist project to avoid an ethnic outbidding because it could alienate its supporters who are Spanish native speakers and, consequently, it could jeopardize the strength of the proindependence block.

Conclusion
In this article, I have used the same logistic regression model with some sociodemographic variables to explain probabilities in wanting independence in Catalonia in two quite different political contexts (1996 and 2020), with the aim to compare change and continuities between them, and to expose some political consequences.
In both years, L1 has been the strongest predictor of those considered. Second, the likelihood of wanting independence depending on whether one has Catalan as L1 compared to Spanish was 77% in 1996 and 80% in 2020, only a low increase of probability. Third, the growth of secessionism has been proportionally stronger among Catalans with Spanish as their L1 than among Catalans with Catalan as their L1. Given the impact of ethnolinguistic diversity on ethnic conflict, after calculating the ENLG for the whole of Catalonia and for the unionist and secessionist subgroups, I can assert that the unionist subgroup has become more homogeneous, while the pro-independence subgroup has become slightly more diverse. It shows that secessionism succeeds in attracting more different people, according to their language (or cultural) background.
These results are congruent with the proposition that despite language being a strong marker of ethnic identity, the language issue plays a lesser role within the whole conflict as could have been previously expected, as both blocks under political confrontation do not reflect strict language alignments.
Cross tabulating the variable L1 with the will of independence within the whole electorate, the weight of each resulting subgroup is calculated (Table 5). This procedure becomes relevant to realize what kind of political dynamics might occur in Catalonia.
The largest group within the Catalan electorate (37.9%) is composed by those who have Spanish as L1 and do not want independence. The second larger one is those who have Catalan as L1 and want independence (26.9%). The third group is those who have Spanish as L1 and want independence (13.9%). The fourth one, with 9.7%, is those who have Catalan as L1 and do not want independence. None of the remaining groups reached 4% of the sample. The third and fourth groups weaken the association between L1 and independence. Moreover, two more aspects have to be underlined relating to the third group: Its weight within Catalan society is not small (13.9%) and its weight within the pro-independence group is quite big (31.9%). In short, although Catalan independence has the support of a great majority of native Catalans, its relative success cannot be understood without the participation of Catalans whose family origins are outside Catalonia. The growth of secessionism this last decade must therefore be explained also by the ability of pro-independence leaders and organizations to make a discourse based on civicdemocratic arguments. These arguments have been quite effective in appealing to sectors with a Spanish cultural background and, consequently, to transcend the language borders that unionism could be tempted to consolidate.

Acknowledgments
The author would like to thank Raül Tormos for his support and helpful feedback on a previous version of the manuscript, and the anonymous reviewers for their insightful comments. The usual disclaimer applies.

Conflict of Interests
The author declares no conflict of interests.