2.1. Linguistic data: isoglosses and Asturian varieties

The territory included in the community of Asturias is part of the northern Iberian Peninsula occupied by the constituent Romance varieties, that is, historical Latin dialects that originate in this area and which, as a result of the Christian repopulations linked to the Reconquista, spread southward giving rise to consecutive dialects (Gimeno-Menéndez, Reference Gimeno-Menéndez1990; Penny, Reference Penny1991; Gargallo, Reference Gargallo1995, Reference Gargallo, Sousa, Romero and Álvarez2014). When analyzed from an integrative perspective, these northern varieties constitute a dialectal continuum with internal differences, but without abrupt and well-defined dialectal borders (Penny, Reference Penny and Fernández Catón2004). The differences that occur between the Romance varieties spoken from the Atlantic coast of Galicia to the Mediterranean coast of Catalonia are gradual, smooth, and marked by independent isoglosses that predominantly have a north to south orientation.Footnote ² Asturias is in an area where these isoglosses, essentially those corresponding to phonetic variables, are concentrated without overlapping.

The historical varieties spoken in the northern area between the Galician and Portuguese domains, and Spanish are traditionally identified with the cover labels of Leonese and Astur-Leonese (Andreose & Renzi, Reference Andreose, Renzi, Maiden, Smith and Ledgeway2013). The Spanish philologist, Ramón Menéndez Pidal, was the one to discover, for Romance linguistics, the singularity of this peninsular variety spoken in the territory occupied by the former Leonese kingdom (Andrés, Reference Andrés and Rodríguez2007). In his 1906 work, El dialecto leonés, he analyses a series of linguistic features that serve to characterize and identify a Romance variety used in areas of the provinces of Asturias, León, Zamora, Salamanca, in adjacent areas of Cantabria and Extemadura, and in the Portuguese municipality of Miranda do Douro (Menéndez Pidal, Reference Menéndez Pidal1906). The study is part of the observation of the rural varieties spoken in this area in his time—beginning of the 20th century—of the research into previous medieval texts and of the consideration of the historical limits of the ancient kingdom of León. Although some authors of the 19th century had already drawn attention to the similarities between Asturian and the languages used in different areas of the kingdom of León, it was Menéndez Pidal who established the leonés label to identify this Romance linguistic variety (Gómez, Reference Gómez Turiel2012). It is possible that, for the author, the denomination was basically justified for historical reasons, since the popular names for the varieties spoken in this area were very diverse and were determined by the local or regional nature of the language (Morala, Reference Morala Rodríguez and Rodríguez2007; Fernández-Ordóñez, Reference Fernández-Ordóñez and Juan Carlos2010).

The work of Menéndez Pidal on Leonese served as a stimulus for subsequent research that allowed us to know the linguistic characteristics of the domain in greater depth, to specify the choice of a less limited name (Astur-Leonese) and distinguish between Asturian, Leonese, and Mirandese (Cano, Reference Cano González, Holtus, Metzeltin and Schmitt1992; Ferreira, Reference Ferreira1995; Borrego, Reference Borrego Nieto and Alvar1996; Martínez, Reference Martínez Álvarez and Alvar1996; García Arias, Reference García Arias1997; Andrés, Reference Andrés and Rodríguez2007). The Asturian varieties—collectively known as Bable, Asturian or Asturian language—stand out in the Leonese domain because of their greater vitality and because they are popularly identified and denominated in a unitary way (Morala, Reference Morala Rodríguez and Rodríguez2007:99; see also Navarro, Reference Navarro Tomás1962, map 4 for the popular denominations of these varieties). Despite linguistically belonging to the Astur-Leonese domain, Asturian is distinguished from Leonese and Mirandese by “peculiar sociolinguistic attributes: great volume of speakers, high linguistic preservation, intense literary activity, urban presence” (Andrés, Reference Andrés and Rodríguez2007:245).Footnote ³ In addition, it is necessary to indicate that since 1981 there is a standard model for the Asturian written language used in education and written publications.

Although an Asturian linguistic atlas is not, as yet, available, there is a massive set of publications that allow us to know the dialectal variation of the Asturian rural languages in detail. Since the pioneering studies of Menéndez Pidal, four linguistic areas, one belonging to the Galician linguistic domain and three other properly Asturian ones, and therefore of the Astur-Leonese domain, have been distinguished within the administrative territory of Asturias. Studies carried out after Menéndez Pidal’s publications which use traditional dialectology methods confirm this division, especially in the most conservative rural language. Following the proposals of Cano (Reference Cano González, Holtus, Metzeltin and Schmitt1992) and García Arias (Reference García-Arias2003), the four identifiable linguistic areas within the territory of Asturias are: Galician-Asturian area; western Asturian area; central Asturian area; and eastern Asturian area (Map 2).

Map 2. Dialectal areas in Asturias.

The boundaries between these varieties are established based on the layout of three isoglosses that run from north to south and that correspond to phonetic variables related to modern results of vowel and consonantal units of vulgar Latin: vocalic diphthongization (terra ‘soil’, porta ‘door’ vs. tierra, puerta); decreasing diphthongs (queisu ‘cheese’, cousa ‘thing’ vs. quesu, cosa); aspiration of f-, [f]echa ‘date’ vs. [h]echa (Map 2). Starting from these phonetic features, and following the model proposed by García Arias (Reference García-Arias2003: 43), it is possible to suggest a dialectal classification of Asturias in four areas (Table 1).Footnote ⁴

Table 1. Classification of language varieties in Asturias

In previous studies similar in objectives and methods to ours, the dialectal data used to establish correlations with the geographical distribution of surnames are taken from linguistic atlas and are analyzed by quantitative methods (Goebl, Reference Goebl1996; Manni, Heeringa & Nerbonne, Reference Manni, Heeringa and Nerbonne2006; Rodríguez-Díaz, Manni & Blanco-Villegas, Reference Rodríguez-Díaz, Manni and Blanco-Villegas2015). The linguistic information used in these analyses enables more refined and solid comparisons. As for the linguistic domain analyzed in this work, up to now there is no available geolinguistic data source that allows such detailed analyses, as discussed above. However, we consider that the classification of the varieties spoken in Asturias, based on the analysis of the most conservative rural varieties and established from phonetic features, which we take as reference, is endorsed by the Romance tradition of dialectal studies.Footnote ⁵

2.2. Surname data

The basic origin of Asturian surnames, like the rest of the names of Spain and Portugal, is the naming system that became widespread in Europe from the 13th century onwards. The Spanish Civil Code of 1870 made it official for citizens to maintain a uniform surname. Thus, the custom of transmission of surnames was formalized, a custom which, with few variations, traditionally had been used since medieval times (Kremer, Reference Kremer2004). Official surnames in Spain consist of two parts: the father’s family name and the mother’s family name, in this order.

Basic etymological categories of surnames of the Iberian Peninsula are patronymics, habitational names, and lexically derived names (Kremer, Reference Kremer and Hanks2003). The Asturian territory is part of the north of the peninsula, an area that some students point out as the birthplace of most of the surnames extended in Spain during and after the Reconquista (Mir, Reference Mir de la Cruz1981). Proof of this statement is the fact that the most common surnames in all of Spain are also those that show the highest frequencies in the northern regions (García, Fernández, González, Rodríguez, López, etc.). As northern Atlantic Spain, the Asturian population is also characteristic in the abundance of surnames that originate in place names (Faure, Ribes & García, Reference Faure, Ribes and García2001) and in a high frequency of isonymy, a measure of surname similarity inside a group (Rodríguez-Larralde et al., Reference Rodríguez-Larralde, Martin, Scapoli and Barrai2003; Scapoli et al., Reference Scapoli, Mamolini, Carrieri, Rodriguez-Larralde and Barrai2006).

The list of Asturian surnames was formed mainly in the middle ages and the comparative studies with the current surname corpus show that, essentially, the repertoire of more common forms has hardly changed (Viejo, Reference Viejo Fernández1997). The surnames of patronymic origin (the greater part ending in -ez) still are the most frequent ones in the population. Furthermore, they are the same, with little variation, as those that are found in the 15th century, at the end of the 19th (Viejo, Reference Viejo Fernández1997), and in the first decades of the 21st century. Table 2 shows the 50 most common surnames in the population analyzed and the number of councils in which each one was registered, the percentage of occurrences and the percentage of municipalities in which it appears. It may be verified that most of these surnames belong to the class of patronymics and that are common in all or in most of the Asturian councils. Maps 3 and 4 show the onomastic diversity of the Asturian councils on maps. Map 3 presents the distribution of onomastic diversity based on the total data; in Map 4 the data for each municipality has been weighted based on the number of inhabitants of the municipality, which corrects the effect produced by the most populated municipalities. In both maps, the counties with darker colors are those with a greater number of different surnames. The map on the left allows you to appreciate the greater onomastic diversity of the eastern and western extremes of the territory.

Table 2. The 50 most frequent names in Asturias (INE, 2012)

Map 3. Surname diversity in Asturias (total data).

Map 4. Surname diversity in Asturias (weighted data).

The onomastic information used in our research is more complete and detailed than most of the research of the same type carried out on the whole or specific areas of the Iberian Peninsula (Scapoli et al., Reference Scapoli, Mamolini, Carrieri, Rodriguez-Larralde and Barrai2006; Rodríguez Díaz & Blanco Villegas, Reference Rodríguez-Díaz and Blanco Villegas2010; Rodríguez-Díaz, Blanco-Villegas & Manni Reference Rodríguez-Díaz, Blanco Villegas and Manni2017).Footnote ⁶ Onomastic data were taken from the 2012 registry of inhabitants of Asturias, provided by the Sociedad Asturiana de Estudios Económicos e Industriales (SADEI). This database contains onomastic information of 1,077,360 people with reference to each of the 78 municipalities of Asturias. The total number of different surnames contained in the database is 18,418.Footnote ⁷ As a geographical area of reference, most of the studies on surname distribution carried out on Spain or on Spanish communities use the province, an administrative entity that is much larger than the municipality.Footnote ⁸ The use of the municipalities,Footnote ⁹ the fine scale and discrete unit, allows us to have more in-depth knowledge of the geographical distribution of surnames, especially when dealing with geographical areas of reduced extension, such as Asturias.

2.3. Method

The basic procedure to identify the onomastic structure of a population and to discover the regions of surnames is the use of isonymy measures. Surnames, as a cultural element, are transmitted from generation to generation according to precise rules similar to those governing the transmission of Y chromosomes, especially in those cultures in which surnames pass from parents to children. In general, the identity of surnames involves the existence of family ties; as a consequence, the measures of isonymy serve to account for the inbreeding levels of a population. In 1875, George H. Darwin (Reference Darwin1875) proposed using surnames to determine the proportion of marriages between cousins; in the early 20th century, surnames were also used to analyze consanguinity in marriages in the United States (Arner, Reference Arner1908). From the 1960s onwards, the analysis of isonymy was extended as a procedure for the study of population structure (Shaw, Reference Shaw1960). In recent years, the filtering methods of the data analyzed have been refined (Cheshire, Reference Cheshire2014), although the most commonly used procedures are essentially the same (Lasker, Nei, isonymy between groups, etc.).

In the last decade, there have been many published studies that combine the methods of isonymy determination and cartographic results visualization in order to offer detailed analysis of the spatial distribution of surnames. Cheshire, Longley & Singleton (Reference Cheshire, Longley and Singleton2010) show a strong relationship between district surname and geographic locations in Great Britain, constructing clusters from surrounding districts based on Lasker distances. In Italy, some scholars (Boattini et al. 2010; Reference Boattini, Lisa, Fiorani, Zei, Pettener and Manni2012) analyzed the geographic location of different Italian surnames using neural networks, which allow for distinguishing monophyletic and polyphyletic surnames. Novotný & Cheshire (Reference Novotný and Cheshire2012) have studied the surname space of the Czech Republic, finding clear parallelism between their network representation and ethno-cultural boundaries in the country. Recently, Mikerezi et al. (Reference Mikerezi, Xhina, Scapoli, Barbujani, Mamolini, Sandri, Carrieri, Rodriguez-Larralde and Barrai2013) describe the isonymic structure of Albania. Using a different perspective, Cheshire & Longley (Reference Cheshire and Longley2012) turn to kernel density estimation in order to produce heat maps that detect those areas in Great Britain in which certain surnames are more concentrated.

In order to obtain the regions of the surnames of the Asturian population, we analyzed surnames from the 2012 census across the 78 councils. For the result to provide a better account of population structure, the number of surnames analyzed has been shortened, following procedures used in other works (Cheshire, Reference Cheshire2014, for instances). As in previous studies, only the data referring to the first surname of each person were considered. In the first place, those names that appear in one municipality and those that are below the 5% quartile, that is to say, names that because of their low diffusion do not show an interesting spatial distribution, have been eliminated from the base; some examples are Abellas, Lemiña, Terroba, and a large number of surnames of foreign origin (Abbassi, Chapman, Kadlec, Mailhe, Saffar, Woodrow, Zielinski, etc.). In addition, also removed from the base were surnames above the 95% quartile, which are the most repeated surnames in the population, present in most Asturian councils and most of them of a polyphyletic origin. Exclusion of polyphyletic surnames is quite common in the studies of surname regionalization as they are so widespread. Moreover, due to their diffuse origin, they make it difficult to identify the patterns of regional organization of local surnames. Among the removed surnames are those with at least one representative in each council (Álvarez, Díaz, Fernández, García, González, López, Martínez, Pérez, and Rodríguez). Once these reductions have been made, the final number of different surnames analyzed is 6,502.

2.4. Isonymy measures

Surname (dis)similarity among regions can be quantified by different measures. Consider index $i=1\comma \ldots \comma n$ for denoting a certain geographical region (for two regions, $\left( {i\comma j} \right)$ ). Each region has an associated collection ${S_i}$ of surnames, and for a pair of regions, the collection of all the surnames in them is denoted by ${S_{ij}}$ ( ${S_{ij}} = {S_i} \cup {S_j}$ ). The total number of surnames in a certain region is denoted by ${n_i}$ .

Isonymy refers to the possession of the same surname, a premise in genetics being that individuals with the same surname are more likely to share the same family lineage, so isonymy indicates biological relation. With the notation introduced above, isonymy (as an internal measure, within a region $i$ ) is defined as ${I_i} = \mathop \sum \nolimits_{k \in {S_i}} p_{ki}^2$ where ${p_{ki}}$ denotes the relative frequency of surname $k$ in region $i$ . High values of isonymy are possible in a population where there are few surnames, and low values of isonymy are obtained when the number of surnames is large. Using an analogy with genetics, as with alleles, surname drift is proportional to time, therefore, small isonymy values suggest recent immigration or settlement.

Isonymy can be extended as a measure of population similarities between groups. Under the assumption of a common origin, isonymy between two regions $i$ and $j$ is defined as ${I_{ij}} = \mathop \sum \nolimits_{k \in {S_{ij}}} {p_{{k_i}}}{p_{{k_j}}}$ (1). Other different measures of the isonymic distance between a pair of locations can be derived from (1). For instance, the Lasker distance is given by $L = - \log ({I_{ij}})$ .

Lasker distances can be interpreted as a measure of similarity or difference between two areas, where large distances show less similarity in surname composition. Nevertheless, Lasker distance is not the only choice to quantify surname similarity. Other common coefficients are the Euclidean distance, introduced by Cavalli-Sforza & Edwards (Reference Cavalli-Sforza and Edwards1967) and Nei’s distance (Nei, Reference Nei1973), both given by:

$$E = \sqrt {1 - \mathop \sum \limits_{k \in {S_{ij}}} \sqrt {{p_{ki}}{p_{kj}}} } \quad {\rm{and}}\quad N = - \log\left( {{{{I_{ij}}} \over {\sqrt {{I_i}{I_j}} }}} \right)\comma\ {\rm{respectively}}.$$

Euclidean and Nei’s distances have been developed for purely genetic data, but they can be applied to the frequencies of surnames, as those carried out by Mikerezi et al. (Reference Mikerezi, Xhina, Scapoli, Barbujani, Mamolini, Sandri, Carrieri, Rodriguez-Larralde and Barrai2013). In addition, to detect isolation by distance between locations $i$ and $j$ , the linear correlation of surname distances (Lasker’s, Euclidean and Nei’s) with their geographic distances can be computed.

Once the measures are obtained, the final output is a graphical representation of the different surname regions obtained by Multivariate Analysis. This is usually done by representing the clusters given by dendrograms constructed from the matrices of Lasker’s distances (see Cheshire et al., Reference Cheshire, Longley and Singleton2010), so the basic information of splitting or merging clusters is the similarity or isonymic distance between areas. The basic information for splitting or merging clusters is the similarity or distance between the clusters, and this distance can be obtained by different methods, such as complete linkage or Ward’s procedure. Ward’s method is a hierarchical agglomerative clustering procedure successfully used in linguistic variation analysis (Nerbonne & Heeringa, Reference Nerbonne and Heeringa1997; Goebl, Reference Goebl2006; Szmrecsanyi, Reference Szmrecsanyi2012; Strauss & von Maltitz, Reference Strauss and von Maltitz2017).