Surnames in Honduras: A Study of the Population of Honduras through Isonymy

Authors


  • The work was supported by grants of the University of Ferrara to Chiara Scapoli.

Summary

In this work, we investigated surname distribution in 4,348,021 Honduran electors with the aim of detecting population structure through the study of isonymy in three administrative levels: the whole nation, the 18 departments, and the 298 municipalities. For each administrative level, we studied the surname effective number, α, the total inbreeding, FIT, the random inbreeding, FST, and the local inbreeding, FIS. Principal components analysis, multidimensional scaling, and cluster analysis were performed on Lasker's distance matrix to detect the direction of surname diffusion and for a graphic representation of the surname relationship between different locations. The values of FIT, FST, and FIS display a variation of random inbreeding between the administrative levels in the Honduras population, which is attributed to the “Prefecture effect.” Multivariate analyses of department data identified two main clusters, one south-western and the second north-eastern, with the Bay Islands and the eastern Gracias a Dios out of the main clusters.

The results suggest that currently the population structure of this country is the result of the joint action of short-range directional migration and drift, with drift dominating over migration, and that population diffusion may have taken place mainly in the NW-SE direction.

“Gracias a Dios que hemos salido de estas honduras”

From the Admiral's log of the 4th voyage, 1502

Introduction

Studies of the genetic structure of the population of the Centro-American states are recent and refer mainly to the frequencies of traditional blood group markers. This is true also for Honduras, where traditional and DNA markers (Herrera Paz et al., 2008; Matamoros et al., 2008) and the local distribution of surnames (Herrera Paz et al., 2010) have been studied.

Among the Centro-American States, Honduras (“The Deeps” in Spanish) extends from the Atlantic to the Pacific. It has the approximate shape of a flat inverted triangle with the base limited in the north by the Caribbean coast and the vertex in the south by the Gulf of Fonseca in the Pacific. At its maximum, it is about 700 km long and 330 km wide, for an area slightly larger than 112,000 square km, inhabited by approximately 8.2 million persons.

In this work, we investigate the Honduran population with the aim of detecting its structure through the study of isonymy (Crow & Mange, 1965) in the country as a whole and in the two administrative levels of the nation, namely 18 departments and 298 municipalities. We report here how in Honduras isonymic distance varies with geography, as we observed in other American countries. We obtain signs of the direction of migrations by studying the geographic heterogeneity of surnames. For each level, we study the surname effective number, α, and the value of total inbreeding, FIT, random inbreeding, FST, and local inbreeding, FIS. Thus, our aim in this work is a comprehensive study of the present isonymic structure of Honduras resulting from surname drift and population movements in an area bordered by the Atlantic and the Pacific Oceans, east of Guatemala and El Salvador, and north of Nicaragua.

Materials and Methods

Administrative Subdivisions of Honduras

In 2012, one of the authors (EFHP) obtained from the Central Election Commission (CEC) of Honduras the data suitable for describing the isonymic structure of the country with the methodologies developed by us. In the data that were made available, the list of electors of the 2009 general election had a total of 4,348,021 electors, which were distributed in 18 departments and 298 municipalities of the country. About 90% Hondurans are Mestizos, 7% Amerindians, 2% African, and about 1% Caucasians (The World Factbook, 2013–14, 2013). These proportions, however, often depend on self-declarations so their degree of accuracy may be questioned. Since the ethnic origin is only weakly related to surnames, in this analysis, we decided to ignore ethnicity and use the two administrative subdivisions as statistical units. The geography of both levels is well defined, and all the individuals in the available sample are classified accordingly; municipalities inside departments inside Honduras. Since in this country the Hispanic dual surname system is used, for the analysis we had available 4,348,021 paternal and 3,577,733 maternal surnames, given that in only 82% of the individuals the maternal one was registered. This difference in registration is common in several of the South American countries we have studied, and it may be due to local cultural factors, one of them the disregard of maternal surnames in some documents.

The area studied covers the entire nation. The 18 departments differ in position, area, and population (Fig. 1). The most isolated is one maritime department, made by a streak of four main islands in the Atlantic about 60 km north of the coast, the Islas de la Bahía in the Caribbean, where British surnames are prevalent. There are four coastal departments in the north; Atlántida, Cortés, Colón, and Gracias a Dios, the latter an area of recent immigration (Herrera Paz & Mejia, 2010). Then, immediately south of Atlántida and Colón, are the two large departments of Yoro and Olancho, and from west to east, a belt of nine departments; Santa Bárbara, Copán, Ocotepeque, Lempira, Intibucá, Comayagua, La Paz, Francisco Morazán, and El Paraíso. Then come two departments which delimit the Gulf of Fonseca in the Pacific, Valle, and Choluteca.

Figure 1.

Distribution of the 298 municipalities in the 18 departments for the 2009 elections in Honduras.

The data summarized by department are given in Table 1. The data for municipalities (and further supplementary material) are also available and downloadable at our Web site: http://web.unife.it/utenti/alberto.carrieri/ricerca.htm.

Table 1. Department of Honduras: number of paternal surnames, N; number of different surnames, S; α, as effective surname number; isonymy, I; Karlin-McGregor ν; ratio of number of different surnames to sample size, S/N; proportion of surnames present only once to sample size, hapax
DepartmentNSαIνS/Nhapax
Atlántida231,9143813144.10.0069440.0006170.0160.0065
Colón159,6112540135.30.0073980.0008410.0160.0062
Comayagua233,1131972113.70.0088000.0004830.0080.0030
Copán190,620171188.90.0112550.0004610.0090.0031
Cortés810,7496179135.40.0073870.0001660.0080.0031
Choluteca252,9451840105.30.0095010.0004120.0070.0025
El Paraíso231,743177190.90.0110100.0003880.0080.0029
Francisco Morazán861,8335498131.60.0076010.0001510.0060.0025
Gracias a Dios24,7702502254.80.0039650.0101410.1010.0475
Intibucá102,69485848.60.0205820.0004630.0080.0028
Islas de la Bahía30,7461918193.40.0052030.0062190.0620.0262
La Paz94,81298745.80.0218660.0004720.0100.0033
Lempira142,19796862.80.0159280.0004340.0070.0023
Ocotepeque71,71199282.50.0121310.0011360.0140.0046
Olancho259,9121902121.00.0082650.0004620.0070.0027
Santa Bárbara236,534171881.90.0122210.0003420.0070.0026
Valle110,029125493.00.0107600.0008360.0110.0038
Yoro302,0882802113.60.0088030.0003730.0090.0036
Mean241,5572290113.50.0105340.0013550.0180.0072
SE54,55535211.90.0011280.0006090.0060.0027
Min24,77085845.80.0039650.0001510.0060.0023
Max861,8336179254.80.0218660.0101410.1010.0475

In the following subsections, we recall the definitions of some of the statistics derived from the surname distributions and their meaning in the study of microevolution in human groups.

Isonymy Theory

The main statistics derived from the surname distributions are: Isonymy within (Ijj) and between (Iij) groups, Alpha (α), Karlin-McGregor ν (ν), and Isolation by distance as measured by Lasker's (L), Euclidean (E), and Nei's (Nd) indices. For the definitions of these statistics and their meaning in the study of microevolution in human groups, we refer to our previous papers (Barrai et al., 1996, 2000; Dipierri et al., 2005; Rodriguez-Larralde et al., 1998, 2011) and for an exhaustive review, to Relethford (1988). In the following subsections, we briefly recall the formulation of two estimators of α and the definitions of random kinship.

Alpha (α) was initially estimated by Fisher (1943) as Nν/(1 − ν); subsequently, Barrai et al. (1996) proposed a new formulation by setting α equal to 1/Σp2. Since this estimation corresponds exactly to the allele effective number in a genetic system, we used it instead of Fisher's formulation and called it the “surname effective number.” The difference between the two estimates (1/ Σp2 vs. Fisher) is of the order of 1% for departments and of 3% for municipalities.

Random kinship ΦIJ (x) between any two localities I and J at distance x is given by

display math

where K is the average kinship at geographic distance x = 0, say average FST, and B is a function of average mutation rate and of the variance of x (Malécot, 1955; Kimura, 1960). Then, ΦIJ(x) is always positive and is expected to decrease exponentially to 0 with increasing distance. Here, we define random kinship as

display math

with average FST as the average kinship at distance x = 0.

As geographical coordinates, we used the centroids of departments and municipalities areas obtained from the ArcGis (ESRI) map downloaded from Global Administrative Areas site (http://gadm.org/).

The significance of correlations was assessed with the Mantel's test using 1000 permutations (Mantel, 1967; Smouse et al., 1986). For a graphic representation of the surname relationship between different locations, these were mapped on the first and second axis of the multidimensional scaling (MDS) of Lasker's distance matrix. To this purpose, we used the R software package. In order to detect the direction of surname diffusion, following Menozzi et al. (1978), the first three components from the principal component analysis (PCA) of the same matrix were also projected individually on the Hondura's map, again with the ArcGis (ESRI) software package. To complement and clarify the clustering, we built dendrograms of departments and municipalities. These were obtained from the matrix of Lasker distances between administrative sections. They were considered only as a help to the clustering; we do not imply that the present situation was generated by subsequent splits of pre-existing clusters.

Results and Discussion

The Most Frequent Surnames

The distribution by department of the surname numbers used in the analysis, with the main parameters derived from the isonymy theory, are given in Table 1. The corresponding information for municipalities is presented in Table S1. The distribution of the logarithm of the number of surnames over the logarithm of the number of times they occur (Fox & Lasker, 1983, but see also Adamic & Huberman, 2002) is fairly linear, both for paternal and maternal surnames. The graphs (Figs. S1a, b) are visible and downloadable at our Web site. Some concavity of the distribution is apparent both for paternal and maternal surnames. This trait of the distribution indicates that there is a deficiency of surnames having intermediate frequencies. Note that the average number of persons carrying the same paternal surname is 296.5, the largest type-token ratio (Adamic & Huberman, 2002) observed by us in our studies which overall cover more than 125 million surnames (Table 3). In Table 1, we also give the ratio of the number of different surnames to sample size (S/N) for departments, and the proportion of surnames present only once (hapax legomenon, “present once” in Greek). For municipalities, the same values are given in Table S1. For departments, S/N varies from 0.006, meaning an average of 157 individuals per surname, to 0.101, meaning about 10 individuals per surname. For municipalities, S/N varies from 0.008 to 0.217. We recall that for the whole of Honduras S/N is 0.00337. The hapax for departments is 0.0072 ± 0.0027 and for municipalities it is 0.0188 ± 0.0008. Both values are consistent with the values of S/N, which is 0.018 for departments and 0.052 for municipalities.

In Honduras, surnames originated and have been established generally in the same way as in most South American countries. Honduras makes no exception to the Spanish biparental surname system, the paternal surname being regularly transmitted in the first position and the maternal in the second position, so that in a line it is lost after two generations. We analyzed the surnames as three groups, paternal (4,348,021), maternal (3,577,733), and pooled (7,925,754). We recall that the maternal surname is a paternal surname in the previous generation, so that the paternal versus maternal comparison today is indicative of differences in the male migration, if any.

We studied in some detail the 50 most frequent surnames in both parental series. Differences in some surnames’ characters, like ñ, the tilded n, and the accented vowels, were maintained through the proper ASCII codes. The data are shown in Table S2, which in the paternal series comprise 2,036,772 surnames equal to 46.8% of the total number of individuals with paternal surnames used here.

In the paternal series, the most frequent surnames are Hernandez with 131,312 occurrences, Lopez with 119,449, Martinez with 112,984, Rodriguez with 102,299, and Garcia with 90,200. After these, one finds Mejia (81,525), Flores (66,748), Cruz (65,700), Reyes (59,939), and in 10th place Sanchez (59,145). Overall, the first 10 paternal surnames comprise 889,301 individuals, or 20.4% of the total number of electors. The first 10 maternal surnames are exactly collinear with the paternal, from Hernandez (107,974) to Sanchez (45,337). There are 727,845 such surnames which correspond to as many individuals in this series. They represent 20.3% of persons being registered also with the maternal surname. The frequency agreement between the paternal and the maternal percentages for the first 10 surnames is very close. The same is true for the first 50 surnames. For these, the average ratio for the same surname paternal/maternal is 1.217 ± 0.007, quite uniform.

We note that the most frequent surnames are consistently Spanish. However, outside the first 50 surnames, there are quite a few interesting deviations from traditional Spanish. The main difference is represented by British surnames, very frequent in the Caribbean coast and in Islas de la Bahía. So, irrespective of parental origin, surnames like George (3329), Bodden (2364), Nicolas (1276), Williams (1039), and Wood (955) are highly represented. As an example, the frequencies of the first 50 British surnames are listed in Table S3.

Isonymy Parameters in Departments and Municipalities

In the following, we give the average values of the isonymy parameters in the country as a unit and in the two administrative levels.

Alpha

The effective surname number, α, in Honduras was estimated at 136.3 for the Country, considered as a unit. The average for the 18 departments was 113.5 ± 11.9, and for the 298 municipalities it was 54.7 ± 2.2. The difference between the estimates of α, then of FST, in the two administrative levels, and for the country as a unit, is observed when different subdivisions of the same area and population are considered. This constitutes the “Prefecture Effect,” identified for FST by Nei & Imaizumi (1966) in Japan, and so named by Scapoli et al. (2007). Nei and Imaizumi observed that, for the same area and population, small subdivisions have larger FST, and larger subdivisions have smaller FST. In their study, the effect was seen in towns and in the Japanese prefectures where the towns were located; hence the name. In Honduras, departments and municipalities are analogous to Japanese prefectures and towns. In departments, the major component of total inbreeding, FIT, is local inbreeding, FIS (Fig. 2). The F's are standardized variances; a larger F means a larger dispersion of surname frequencies. In Figure 2, we note that FIS, which is a within group variance, in this case a variance within departments, is consistently larger than FST, the variance between groups. On the average, for Honduran departments, the within variance is 70% and the between departments variance is 30% of the total (Table 2). Thus, the variation of surname frequencies within departments is larger than the variation between departments. On the contrary, for municipalities, the average fraction of within variance is 36%, and the between variance is 64%. For Honduras as one entity, the fraction of within variance is 77%. We have here an analogy with the prefecture effect, namely, the variance between groups is larger for small groups (64%), intermediate for larger groups (30%), and minimal for the undivided sample (23%) (Table 2). We are not in a position to comment on the genetic differentiation of groups on the basis of these results; we deal with frequencies of surnames, which are epigenetic structures. For important discussions on these issues, see Lewontin (1972) and Edwards (2003).

Table 2. Components of inbreeding in the undivided country and in administrative sections of Honduras. Averages
    Variance
SectionFITFSTFISWithinBetween
Country0.00790.00180.006177%23%
18 Departments0.00880.00260.006270%30%
298 Municipalities0.01150.00730.004236%64%
Figure 2.

Variation of random and local inbreeding over total inbreeding inside 18 departments of Honduras.

We observe, however, that for identification and classification purposes, whatever the ratios between standardized variances, rare alleles are useful discriminators. In Honduras, a Hernandez might come from anywhere in the country, in fact, from anywhere in South America, but a McLaughlin has a high chance of coming from the Bay Islands.

Inbreeding by Isonymy

Departments

Values of the inbreeding coefficients and of α are given in Table 1 for departments and in Table S1 for municipalities. The highest value of departmental α (254.8) is observed in the department of Gracias a Dios, a large swampy area in the east of the country, which has the smallest population. In this department, we counted 24,770 electors, while, as an example, the department of Francisco Morazán, where the capital Tegucigalpa is located, has 861,833 about 30 times more electors than Gracias a Dios. The next value of α, 193.4, was observed in Islas de la Bahía, with 30,746 electors. Both have a relatively small population. We cannot invoke recent migration to Islas de la Bahía for an indication of their low level of inbreeding because, with the exception of Atlántida and Colón, migration from the rest of the country toward these islands has been scarce, but we might do so for Gracias a Dios, where settlements are more recent. Both these departments are multiethnic and have an important English contribution, thus the bicultural status may be the cause of the high abundance of surnames.

The smallest values of departmental α, 45.8, and 48.6, were observed in La Paz and Intibucá, respectively, which are mainly mountainous, and border with each other and with El Salvador to the south. Next is Lempira (α = 62.8), also mountainous and bordering with Intibucá.

Municipalities

The largest values of α, the inverse of isonymy, were seen in the large municipalities which are also capitals of departments. Highest α's for municipalities were 242.6 in Puerto Lempira, capital town of Gracias a Dios, 176.4 in Roatán, capital of the largest of Islas de la Bahía which has the same name. Then 149.8 in La Ceiba, head town of the Atlántida department and 148.2 in Puerto Cortés, close to the border with Guatemala, which is presently the most important port of Honduras.

In San Pedro Sula, the second largest town of Honduras, α is 142.2; this municipality currently has the highest immigration rate of the country (it is called the “Industrial City”) and it is near the coast, about 60 km from Puerto Cortés. The α is 141.1 in La Lima (it is located in Cortés department and it is called “Little New York” because of the intensity of its urban life). The Municipality of Distrito Central, where Tegucigalpa is located, has a population of more than 800,000 and an α of only 139.0. Several of these large municipalities take the name of the department where they are located. Guanaja, in Islas de la Bahía, has less than 5000 electors and α = 137.6. Some of these municipalities have industrialized agriculture, particularly of fruits, and also industries of clothes manufacturing (called “maquila”). This is true especially for (mainly) San Pedro Sula and (secondarily) La Lima and Choloma, and might have resulted in recent immigration toward these areas.

The lowest α (and highest FST) are observed in the municipality of Santa Cruz in Lempira, (with α = 5.6), Opatoro in La Paz with 5.7, Guata in Olancho with 7.6, and Guajiquiro in La Paz (α = 8.5), and Dolores in Intibucá with α = 8.9. The sixth and seventh municipalities are Lauterique (9.5) and Mercedes de Oriente (11.0) both in La Paz. These municipalities are located in mountainous areas.

Effect of Geographic Distance

We studied isolation by distance through the correlation of geographic with surname distances at the department and municipality levels. We found that Lasker's, Euclidean, and Nei's distances between the 18 departments were significantly correlated with linear geographic distance, with r = 0.65 ± 0.050, 0.63 ± 0.068, and 0.55 ± 0.086, respectively. The same tendency, but with lower correlations, was observed among the 298 municipalities, where we observed 0.37 ± 0.009, 0.28 ± 0.009, and 0.08 ± 0.011 for Lasker's, Euclidean, and Nei's, respectively. As an example, the variation of Lasker's distance between departments is given in Figure 3 (see Fig. S2 for the distribution of Lasker's distances between municipalities). We observe that the correlation of Nei's distance with geographic distance, although significant, is very low. This may indicate large movement between municipalities at very near distances and this might also be reflected in the low correlations of Lasker's and Euclidean distances at the municipality level. In fact, such movement was observed in recent times (Flores-Fonseca, 2006, 2011).

Figure 3.

Variation of Lasker's distance between departments with geographic linear distance.

The signal extracted from the scatter diagram of Lasker's distance over kilometers for municipalities is given in Figure 4. Although it rises sharply below the first 100 km, a clear tendency toward an asymptote is not observed, as it was in Spain, Bolivia, and Chile (Rodriguez-Larralde et al., 2003, 2011; Barrai et al., 2012). In these countries, the relation between isonymic and geographic distance flattens after 100 km. In Honduras also, there is an increase of surname distance up to 100 km, which gives indication of the presence of some isolation and drift below that distance. After that the increase in isonymic distance becomes continuous but not clearly asymptotic.

Figure 4.

The signal extracted from the variation of Lasker's distance (±SD) between 298 municipalities over geographic distances.

Kinship

We plotted random kinship between municipalities as previously defined, as a function of geographic distance (Fig. 5). The decrease of kinship with distance is significantly exponential, as predicted by Malécot (1955), see also Kimura (1960). Note the symmetry with Figure 4. Specifically, the exponential decay should be characteristic of structures more linear than Honduras, for example, as observed by us in Chile. However, there is considerable and significant agreement between Malécot theory and kinship decay also in Honduras. Therefore, the model is very strong and, possibly due to the large number of pairwise distances available to us, it is also applicable to a geographic structure which, like Honduras, is elongated from east to west, but which is poorly linear. A similar agreement was observed in Albania (Mikerezi et al., 2013). In Figure 5, we plot the observed decay of kinship (Obs Kinship), and the decay obtained from its exponential regression on all geographic distances (Exp Kinship). In Figure S8, we also indicate all the observed points. The correlation of these with distance is 0.45. The Exp Kinship is given by the function

display math

which is significant (F[1,∞] = 11,106; P << 0.0001). We recall that the 44,253 distances are not independent, since they are due to the convolution of 298 independent observations; we therefore cannot assess exactly the probability of F, which would be vanishing at this value of the indicator.

Figure 5.

Exponential decay of random observed kinship (±SD) and expected kinship in Honduras over geographic distance. Pairwise distances between 298 municipalities.

From this function, and recalling that the exponent is inline image, here σ2 is the variance of geographic distance, for the present 44,253 distances σ2 = 8668.7, we obtain

display math

which estimates the rate of change of surnames per generation at a little more than 2%. In Table 3, we give the linear correlation between isonymic and geographic distances for all the areas studied to date. For Honduras, the correlation between Lasker's and geographic distance is 0.44 and we consider the correlation an indication of isolation by distance. If there is a positive correlation between geographic and surname distances, there is a gradient of these distances, possibly indicating increasing isolation.

Table 3. Comparison of isonymy parameters in nine European countries, in five South-American countries, in the USA and Texas, and in Yakutia. Overall, more than 125 million surnames were analyzed
CountrySample size (SS, millions)Surnames (S)α (average)Isolation by distanceaSS/ S(%)
  1. a

    Isolation by distance is measured through the correlation of geographic with surname distances.

Europe     
Austria1140,7668540.597.1
Albania3.037,1841230.7182
Belgium1.1137,4429970.748
France6495,1041,6150.6912.1
Germany5.2462,5261,5960.5111.2
Holland2.4126,4857870.4619
Italy5.1215,6231,2360.6123.7
Switzerland11.7166,1168910.7210.2
Spain3.6    
Paternal 94,8861340.2138
Maternal 110,0341440.2633
Asia     
Yakutia0.544,6251070.6911.1
NORTH AMERICA:     
USA18899,5851,3660.2420
Texas3.6235,7407340.4215.3
South America     
Argentina322.6414,4414220.4754.5
Venezuela23.968,6651220.7856.8
Bolivia423.2174,9221220.5144.6
Paraguay34.839,0471080.42122.9
Chile416.472,6672090.63224.0
Central America     
Honduras34.314,6621130.44296.5
1Cantons     
2States     
3Districts     
4Provinces     
Elsewhere towns     

Relations between the Administrative Sections of Honduras

In order to obtain a general idea on the movements of population groups in Honduras, we conducted MDSs and PCAs on the matrix of Lasker's distances between departments and between municipalities (tables and figures are reported as Supplementary Information).

Departments

The MDS projection on the first two axes of the matrix between departments (Fig. S3) indicates one western cluster formed by Cortés, Copán, Santa Bárbara, and Ocotepeque. Then there is one central cluster from north to south, with Atlántida, Colón, and Yoro in the north and La Paz, El Paraíso, Valle, Lempira, and Intibucá in the south. Islas de la Bahía and the eastern Gracias a Dios are outliers.

In the resulting dendrogram obtained with the complete linkage method of agglomeration (Sørensen, 1948; Fig. S4), a first cluster comprising seven western-central departments is observed, with Copán and Ocotepeque which form a subcluster bordering with Guatemala, and a large cluster of five neighboring departments in a north-south belt made by Cortés, Santa Bárbara, Lempira, Intibucá, and La Paz. The second cluster comprises nine departments; Choluteca, El Paraíso, Francisco Morazán, Atlántida, Olancho, Colón, Comayagua, Yoro, and Valle. These departments are conterminous.

From the MDS projection in Figure S3, some other minor but relevant points emerge, which complement the clustering of departments. In particular, Francisco Morazán, Atlántida, and Choluteca, with Yoro and Valle, stand near the center of the bidimensional projection. As stated above, the Islas de la Bahía with their British surnames and Gracias a Dios with their recent immigrants are outliers. Choluteca and Valle, at the pacific side of the Andes, border with the Ocean.

Municipalities

We do not present here either the projection of the first two directions of MDS for the 298 × 298 matrix of the municipalities nor the dendrogram derived from it. Both are, however, given online as supplementary Figures S5 and S6. Since the individual names of 298 elements of the projections are illegible, we decided to identify all municipalities with the number of the respective department. This was done for the projection of Lasker's matrix (Fig. S5) and for the dendrogram (Fig. S6). The resulting clusters correlate very strongly with Departments.

Mapping of the first three components of Lasker's matrix

The structures revealed by the MDSs and the dendrograms are only partially indicative of the possible movements of the population; thus, in order to have a general idea of the direction, if any, of population movements in Honduras, we mapped on the nation (following Menozzi et al., 1978) the first three components of the matrix of Lasker's distance, obtained from a PCA (Fig. S7) and from the MDS. We provide the PCA components because the relative importance of each component is given by the corresponding eigenvalue, while the MDS provides the value of the stress for a judgment of the overall fitting on the dimensions.

We present the results of both methods showing that the resulting maps are similar, particularly for the first and second component. The third component, which, however, is responsible only for 5.2% of the variability, is less similar in the two methods, indicating that at this level random factors begin to dominate.

The resulting maps are given in Figures 6a, b. The intensity of color in each map is proportional to the deviation of the department on the respective axis.

Figure 6.

Projection of Lasker's matrix of surname distances on departments in Honduras by mapping (a) the first three PCA's factors (I: Factor 1 = 80.5%; II: Factor 2 = 8.4%; III: Factor 3 = 5.2%) (b) the first three MDS's dimensions (I: Dimension 1; II: Dimension 2; III: Dimension 3. Stress 10.75%).

The variation of the first component, which accounts for almost the total variability (80.5%) in the west-east direction, indicates a gradient of surname frequencies, which is compatible with diffusion of surnames from the west of the country toward the east and the center, as documented from other sources (see below). This might mean that the main immigration originated from (or passed through) the Guatemalan area. The second and third components (8.4% and 5.2%) give the same indication, although with minor intensity, given their size. Recent migration from rural western departments of Ocotepeque, Copán, and Santa Bárbara to Cortés is very high (west to east). Ocotepeque has had some influence (and thereby immigration) from Guatemala. Also, the port of Omoa, in Cortés Department and close to the Guatemalan border, was the most important site of entrance to Honduras during the colonization period until the 19th century and this last fact might be mostly responsible for the west-east direction.

Thus, the sense of movement may be postulated to be from the west, since immigration in this area is documented. A concise summary of recent immigration to Honduras may be seen at http://www.nationmaster.com/country/ho-honduras/imm-immigration.

However, recent immigration is estimated at 3.6 per 1000 per year, so that any visible structure is mainly dependent on ancient or remote immigration.

Overall, the three components described above account for more than 94% of the surname variation as obtained from Lasker's distance matrix.

The mappings of the first three dimensions of the MDS are compatible with those obtained from the PCA. The possible west-east movement seems indicated by the first and second dimension, and less so by the third. So, the present isonymic structure of Honduras seems to be mainly due to migration from the north-west, with radiation toward the east and south, and with subsequent isolation and drift, with drift and short-range migration playing a major role in the generation of the present geographical variation of surnames. We believe that a few additional considerations are necessary to justify our use of these maps, obtained both from PCA and MDS, since this kind of representation is somewhat controversial (see e.g., Rendine et al., 1999; Sokal et al., 1999; Novembre & Stephens, 2008, 2010; Novembre, 2012). However, in this case, there is considerable consistency between the results of the mappings and the known migration data in Honduras. Therefore, we limit ourselves to the observation of the consistency between the results of the eigenvectorial representation and the historical migration data.

Conclusions

The methodology described in this paper was used to analyze the isonymic structure of several South American countries (Rodriguez-Larralde et al., 2000, 2011; Dipierri et al., 2005, 2011; Barrai et al., 2012). In these countries, 4 (Venezuela), 24 (Argentina), 23 (Bolivia), 4.5 (Paraguay), and 16.5 (Chile) million surnames from the registers of electors were used. In European countries and in the USA, we analyzed the surnames of telephone users (Barrai et al., 2001; Scapoli et al., 2005, 2007; Rodriguez-Larralde et al., 2007). In sparsely populated Siberia, we used half a million surnames (Tarskaya et al., 2009). The average value of α for all the cities (or states, in the case of Venezuela and the USA, or districts, in the case of Argentina and Paraguay), and the isolation by distance measured by the correlation between isonymic and geographic distances, are given in Table 3 for the countries studied up to now. Several features emerge from the comparisons reported in Table 3. First, the general similarity among European nations in profusion of surnames as measured by α, and for isolation by distance, as measured by the linear correlation. Second, the relatively small value of α in Venezuela, Bolivia, Paraguay, Spain, Chile in South America, and Spain and Albania in Europe; and third, the practical absence of isolation by distance in the USA, excluding bilingual Texas (Rodriguez-Larralde et al., 2007). In Honduras, the average number of persons having the same surname (measured by the ratio Sample Size/Surnames, given as the index SS/S in Table 3) is similar (296.5) to Chile (224). The ratio in countries where we had only the surnames of telephone users is about 25% of the ratio observed in countries where we had near-census data. We begin to believe this might be a “census effect” but we recall that countries where we had surnames of telephone users are the ones with higher α, i.e., higher richness in surnames. Thus, we wait to explore further the effect when we have more data available from complete or partial national censuses, because for the time being, barring Yakutia and Albania, the effect is confounded with the small number of different single surnames in the Spanish language, and Honduras seems to offer no exception to this peculiarity of that language.

In Honduras, inbreeding estimates were lowest (and α highest) in the sparsely populated oriental area, and in Islas de la Bahía. The oriental Gracias a Dios, a department of recent immigration, has the highest α value. However, at present, most internal migration seems to take place toward the capital and the other main towns. Consequently, we may conclude that currently the population structure of this country is the result of the joint action of short-range directional migration and drift, with drift dominating over migration, as suggested by the rise of Lasker's over geographic distance below 90–100 km and by its tendency to decrease above that distance.

Acknowledgments

The authors are grateful to the CEC of Honduras, especially to Mr. Carlos Humberto Arita Mejía, who conceded the data.

Ancillary