PREVENCIONES ESPECIFICAS

Analyses of surnames were used in two phases during the course of this thesis. Firstly, we used this approach using data from the 20 most populated villages, to identify the three locations with the highest variability of surnames and highest estimated distances, and thus avoiding samples from closely related individuals as much as possible2_.

During the main fieldwork, for practical reasons such as difficulty of access, proximity to other villages, and interest of the villagers in our research, surname information was not prioritised in deciding locations. Indeed, two locations included in our final sample were not among the 20 most populated villages used for the first analysis. We therefore repeated the analysis, but this time using data of surnames from the villages included in the actual sample. This second analysis is reported in this Chapter, and the data used are presented in Table 4.1. A report with the results of the first analysis, used to decide a sampling strategy, can be found in Appendix F.

The data on surnames were readily available in a convenient form for these purposes as a result of a public policy, in place for the last 20 years: due to the particular system of land tenure of these communities, governments have been encouraging members to register their individual rights as commoners in a dedicated section of the Land Registry. Since these lists are recorded at the com- munity level they are particularly useful for our purposes. However, many commoners have not yet registered their rights and some commoners who have migrated out of the villages might still have their rights registered. Additionally, each list contains only the name of the person who is registering the rights on behalf of the whole household, thus does not include partners, children, or any other people living with them. Therefore, these lists are not accurately representative of

4.1. Population Structure and Surnames 98 the real populations, and the ratio of people registered varies widely between communities and is not proportional to the total population size.

Community Total names S

Puclaro 195 44 Canelilla 128 36 Monte Patria 525 84 Barraza 137 37 Espinal 108 24 Calera 335 64 Mal Paso 236 46 Canela 1319 140 Huentelauquén 391 87 Total 3275 307

Table 4.1.Number of total names and number of unique surnames (S) obtained from the Land Registry records for each community.

A second source of population data comes from a special report on Ag- ricultural Communities published by INE (Chilean National Institute of Statistics) based on data obtained from the National Census of 2002 (Vergara et al., 2005), described in section 2.2.12. This report con- tains data on population size at com- munity level, but it is out of date, and

the figures may have been affected by patterns of seasonal migration.

The data on the Census available to the public do not include names of individuals, and therefore, our surnames analysis is done based on the lists from the Land Registry authority. Land registry data were used to compare the size of the list of each community with the population size reported in the census, assuming that the latter includes all the people living under a single land registration as well as non-registered commoners. We found a weak correlation between registered commoners in the Land Registry list and the number of houses in each village (r = 0.5299), and between registered commoners and overall population size (r = 0.4937).

These communities use the Spanish system of surname inheritance; people have two surnames, a first surname from the father, and a second surname from the mother, and normally women do not change their surnames upon marriage. Since a person only has two surnames, only the paternal surnames (one from each parent) are passed down. Therefore maternal surnames are lost from one generation to the next, so that people do not share any surnames with their grandmothers. While surnames are usually described as analogous to the Y chromosome, this system would be better described as male-inherited mitochondrial DNA, since a woman does have a surname (inherited from her father instead of her husband), but cannot pass it down to her grandchildren. This feature allows an extended method of isonymy using four surnames instead of two, resulting in more precise estimations of inbreeding (Lasker & Kaplan, 1985; Pinto-Cisternas et al., 1985; Shaw, 1960), and has been exploited to study surnames in several Spanish speaking populations (Bronberg et al., 2009; Dipierri et al., 2005; Pettener et al., 1998; Rodriguez Diaz et al., 2010), including Chile (Barrai et al., 2012). However, we could not take advantage of this trait since the maternal surnames were not always recorded in the data available.

4.1. Population Structure and Surnames 99

Espinl Puclar MalPas Barraz Canlll Calera MntPtr Hntlqn Canela

α 0 5 10 15 20 25 30

Figure 4.1.Fisher’s α for each Village based on frequencies of surnames. Values of α are equivalent to the effective number of surnames in each population.

These lists of surnames have some interesting traits: Firstly, many communities are named after family names (e.g. Barraza, Castillo); and many names that are rare in other parts of Chile (and as- sociated with rich families) are very common here (e.g. Tagle, Cox, Cotapo, Ossandon). In addition, some surnames are clear mutations of others (e.g. Velasco and Velazco, Rojas and Roja, Cox and Coz, Sapiaín and Zapiaín). However, these features have been ignored in the analysis since we do not have the means to test them empirically.

The extent of inbreeding was estimated form the diversity of surnames, following a method con- ceived to study species abundance (Fisher et al., 1943) and adapted to the study of inbreeding using surnames from data on marriage (Rodríguez-Larralde et al., 1993) or lists (Barrai et al., 1996). The procedure starts by defining Unbiased Isonymy (I) as I = P_i(pi)2−1/N, where pi is the relative

frequency of the ithsurname and N is the total number of people in the list3. I is normally con-

verted into Fisher’s alpha (α = 1/I), a value homologous to the effective number of alleles, as it results in a number of surnames which, all at the same frequency, would have the observed I (Rodriguez-Larralde et al., 2011).

Values of α for each village are presented in Figure 4.1. These values are very low compared to other studies (Bronberg et al., 2009), and are in agreement with previous findings showing the Coquimbo (and particularly the Limari Valley) as the region with the smallest α in the whole country4_(Barrai

et al., 2012). The results of α as an estimate of inbreeding will be compared with the estimations of inbreeding using STR markers in section 4.2.1.

To examine isolation by distance, the surname lists were use to compute a similarity matrix (Fig- 3_{This is homologous to F}

isas shown in section 2.3.3.2

4_{The results by Barrai et al. (2012) at the province level are: Choapa’s α = 113.3, Limari’s α = 76.9, Elqui’s α = 140.2,}

4.1. Population Structure and Surnames 100 0 0.25 0.5 0.75 1 Puclaro Calera Espinal Huent elauquen Canelilla Mal Paso Mont e Patria Barraza Calera Espinal Huentelauquen Canelilla Mal Paso Monte Patria Barraza Canela

Figure 4.2.Similarity matrix of surnames frequencies (Hedrick’s method). The grid shows a pie chart at each pair of villages, representing the similarity in the distribution of surnames between two populations calculated using the method

described by Hedrick (1971) and adapted for surnames by Weiss (1980). Pairwise similarity represented by pie chart can range from 1 (same surnames at the same frequencies in both villages) to 0 (no surnames in common between both villages). Colours are used to highlight the same differences shown in the pie charts according to the colour key below

the grid.

ure 4.2) based on the method described by Hedrick (1971), adapted for surnames by Weiss (1980), using the R package “Biodem” (Boattini & Calboli, 2012).

This matrix shows low similarity between Puclaro (the northernmost village) and the other villages, and high similarity between the contiguous villages of Calera, Espinal and Barraza. Furthermore, a correspondence analysis of surnames (see Figure 4.3) divides communities geographically (from south to north) on the first dimension (17.64% of variance).

Correspondence analysis shows weak geographic structure. The first axis captures Puclaro and Huentelauquen at the northern and southern extremes, as expected, but Mal Paso and Canela should be closer to Huentelauquen, and Canelilla and Monte Patria closer to the cluster of Espinal, Barraza and Calera (see the map in Figure 2.1 to identify these locations). The second axis (16.31% of variance) cannot be interpreted geographically in the same fashion, though it roughly suggests distribution from east to west, with the exceptions of Canela, which should be in the middle, and

4.1. Population Structure and Surnames 101 -1.5 -1.0 -0.5 0.0 0.5 1.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 Barraza Calera Canela Canelilla Espinal Huentelauquen Mal Paso Monte Patria Puclaro Axis 1 (17.64%) Axis 2 (16.13%)

Figure 4.3.Correspondence analysis of surnames frequencies order villages from South to North at the first axis (17.64% of variance). The second axis (16.31% of variance) cannot be interpreted geographically in the same fashion and does not

show geographic structure.

Huentelauquen, which is the westernmost location and should be at the top.

To measure this trend, pairwise kinship coefficients were calculated using Lasker’s method (Lasker, 1977), and transformed into a distance matrix according to Ldist = − loge(2R), where R is Lasker’s

kinship coefficient (Rodriguez-Larralde et al., 1998). A Mantel test was used to assess correlation between this matrix and a matrix of geographic distances, confirming a small but significant cor- relation (r = 0.4, p–value = 0.03, based on 100,000 replicates).

To conclude, the analysis of surnames suggests some levels of inbreeding and small, but detectable, effect of isolation by distance. A comparison between geographic and surname distances can be visualised using neighbour–joined trees, as is shown in Figure 4.4. These suggestions will be tested using highly variable STR markers in the next section.

4.2. Relatedness 102