Genetic Variation in the Population of Ibiza (Spain): Genetic Structure, Geography, and Language

ANTONIAPICORNELL,' ANA MIGUEL,' JOSE A. CASTRO,' M. MISERICORDIARAMON,' RECTOR ARYA,2 AND M.H. CRAWFORD2

Abstract A sampleof 203 individualsfrom Ibiza (BalearicIslands, Spain)were tested for blood group and serumprotein genetic variation andcompared with other circum-Mediterranean populations. Allele fre- quencieswere calculated for the following blood group and serum systems:ABO, Rh,MNSs, P, Lewis,Duffy, Kell, ORM, GC, TF, PI, and HP. The allelefrequencies from Ibiza werecompared with those from otherBalearic Islands (Majorca and Minorca) and with related European andNorth African groups using an assortmentofanalytical methods (geneticdistances, R matrixanalysis, and Manteltests). R matrixanalysis revealedthat Ibiza is geneticallydifferent from the other Balearic pop- ulationsand, because of geneflow from Spain, clusters with European groups.The levelof geneticmicrodifferentiation of the Mediterranean populations,measured by RST (average of the R matrixdiagonal elements, rl7),is 0.028.An examinationof therelationship between genetic, geographic,and linguisticdistances by Manteltests revealed that genetic distancesare significantlycorrelated with linguistic distances, whereas thegenetic distances are not significantly correlated with geographic distances.The plotof meanper locus heterozygosityversus the genetic distancefrom the centroid of distribution revealed that all threeBalearic Islandshave experienced considerable gene flow but that Ibiza has been mostaffected by the action of stochasticprocesses.

Island populationshave been of greatfascination to populationgeneticists and biologicalanthropologists because thesereproductively isolated and often small human aggregatesoffer an opportunityto investigatethe effectsof stochasticprocesses and unique historicalevents. For example,the research on thepopulation of Tristanda Cunhaby Roberts(1968) yieldedconsiderable

1 LaboratorideGenetica, Department deBiologia Fonamental i Ciencias dela Salut, Universität de lesliles Balears, Carreterra Valldemossa, km7.5, 07071 Palma de Mallorca, Spain. 2Laboratory ofBiological Anthropology, Department ofAnthropology, University ofKansas, Law- rence,KS 66045. Correspondenceshouldbe sent to M.M. Ramon. HumanBiology, December 1996, v. 68, no. 6, pp. 899-913. Copyright© 1996 Wayne State University Press, Detroit, Michigan 48201-1309 KEYWORDS: BLOOD GROUPS, SERUM PROTEINS, R MATRIX, IBIZA, BALEARIC IS- LANDS 900 /PICORNELL ET AL. insightinto the evolutionary effects of uniquehistorical events on subsequent generations.Similarly, the influence of geographyon thegenetic structure of theÀland Islands was revealedusing island populations (Mielke et al. 1982). More recently,Crawford et al. (1995) demonstratedthe effectsof the inter- actionof religion,economics, and geographyon the geneticstructure of the islandisolates of Newfoundland. In contrastto thesegeographic isolates, Ibiza and theother two Balearic islandswere located on thecrossroads of historical populationmovements in the circum-Mediterraneanregion. They offeran opportunityto observethe interactions between successive migrations and the actionsof stochasticprocesses on a small foundingpopulation.

Population Ibiza is a small island in theMediterranean Sea locatedapproximately 70 km fromSpain and 100 km fromMajorca (Figure 1). Ibiza is thesmallest of the threemain Balearic Islands withan area of 567 km2,compared with Majorca (3640 km2) and Minorca (702 km2). Ibiza differsfrom the other two islands in landscape,vegetation, and especiallythe originsof founding settlements. Thereis no archeologicalevidence for human habitation on Ibiza before the arrivalof theCarthaginians in 654 b.c. If theisland was occupiedduring the prehistoricperiod, its residentsfailed to leave evidenceof the"talaiotic" (stone builder)culture that is prominentin bothMajorca and Minorca.The Carthaginiansremained in Ibiza forat least fivecenturies. They foundedthe largestcity, Ibusium (currently Ibiza), namedafter the Egyptian-Phoenician- Carthaginiangod Bes. The Carthaginianswere notrestricted to coastal areas but also colonized the interiorof theisland. Ibiza was annexedby theRoman Empire in 123 b.c. as partof a political pact. Apparently,the Romans failed to occupy Ibiza, as evidencedby the absence of any Romanremains on theisland. The Balearic archipelagocame underMoslem dominationfrom the seventhto the twelfthcentury, leaving behindboth cultural and possiblybiological influences. In a.D. 1229 theCat- alans expandedinto the Balearic archipelagofrom what is now Spain, first withan invasionof Majorca and thenin 1235 by theoccupation of Ibiza. The autochthonouspopulation of Ibiza has grownfrom approximately 3,000 personsin 1392 to 9,596 inhabitantsin 1652 to approximately40,000 people in 1990. In 1990 the totalpopulation of Ibiza was 80,538, but only 50% of theseinhabitants were autochthonous. Duringthe last seven centuriesthe Ibiza populationhas been reproduc- tivelyisolated and thusreceived little gene flowfrom outside. This apparent reproductiveisolation was disturbedby a 1970s touristinflux that promoted immigrationfrom mainland Spain and doubledthe total population. The small size of the autochthonouspopulation and its reproductiveisolation resulted GeneticVariation in Ibiza /901

Figure1. LocationofIbiza, Minorca, and Majorca in the Mediterranean Sea.

in a moderateincidence of consanguineousmarriages (Vails 1969). The per- centageof consanguineousunions varies from1% to 16% in parishesscat- teredthroughout the island withan average of 5.59% forthe entireisland. Wright'sF inbreedingcoefficient for Ibiza was computedto be 0.0019 (Vails 1969). The historyof Ibiza, whichcan be characterizedby its gene flowfol- lowed by reproductiveisolation, is reflectedin the gene pool of thecontem- porarypopulation. Thus a geneticanalysis of Ibiza and comparisonwith other humansettlements of the Mediterraneanshould contributemuch to our un- derstandingof Ibiza' s historyand the actionsof evolutionaryprocesses. To date, the populationof Ibiza has been characterizedgenetically by a few blood groups(Mourant et al. 1976) and erythrocyticenzymes (Miguel and Petitpierre1989). The aims of thisstudy were (1) to extendthe genetic 902 /PICORNELL ET AL. characterizationof the island populationby testingfor additionalblood groupsand serumproteins; (2) to comparethe allelic frequenciesobserved in Ibiza withthose of othercircum-Mediterranean populations; (3) to measure any possible geneticdifferentiation of the Ibiza populationfrom the other Balearic islands and the populationsof mainlandSpain and NorthAfrica; (4) to estimatethe differentialcontributions of European,Middle Eastern, and NorthAfrican populations to contemporaryIbiza; and (5) to determine the relativeroles of geographyand linguisticsin the migrationpatterns to Ibiza.

Materials and Methods

Blood Analysis. Blood samplesfrom 203 individualsof the autochthonous populationwere collectedin a local hospitalin Ibiza city.These specimens were immediatelyshipped (at 4°C) to the Laboratoryof Geneticsin Palma, Majorca. For theABO and Rh systemsthe sample size was larger,with data on 487 individuals.These additionalsamples were made available froman earlierinvestigation. Blood groupingwas done immediatelyupon receiptof the specimens, - whereasthe serumsamples were keptfrozen at 20°C and analyzedlater. Blood groupswere typed using the standard techniques recommended by the antiserummanufacturers (Behring and Marburgof Germanyand Grifolsof Barcelona,Spain). Haptoglobinswere typedby horizontalstarch gel electrophoresisfol- lowing the methodsof Smithieset al. (1962). The systemsGC, ORM, PI, and TF were typedby isoelectricfocusing using the automatedPhastsystem (Pharmacia,Uppsala, Sweden) on miniaturizedgels in a pH gradientof 4.0- 6.5. The separationwas followedby immunofixation(GC and ORM) or by proteinstaining (PI and TF) using the methodsof Carracedoet al. (1986), Moral (1987), and Montielet al. (1988). BeforeTF and ORM typing,samples were treatedwith ferrous ammonium sulfate for TF typing,as describedby Constanset al. (1980), or withneuraminidase for ORM typing(Carracedo et al. 1986).

Analytical Methods. Allele frequenciesof 13 loci from15 populations wereused to constructa variance-covarianceR matrix,which underwent prin- cipal componentsanalysis (Harpendingand Jenkins1973). The average of the diagonal elements(ru) provides an estimateof RST, the equivalentof Wright'sFST and Wahlund's F (Harpendingand Jenkins1973). The eigen- vectors,scaled by thesquare roots of theirrespective eigenvalues, were plot- ted to producegenetic maps. The allele frequenciesused to constructthe R matrixprimarily came froma numberof publishedcompilations (Mourant et al. 1976; Roychoud- GeneticVariation in Ibiza /903

huryand Nei 1988; Tills et al. 1983). Otherrecent sources of gene frequency data on Minorca,Majorca, and Ibiza are Miguel and Petitpierre(1989) and Moral (1987). Matriceswere constructedbased on geographic,genetic, and linguistic distances.The geographicdistances were measuredin kilometersas straight- line distances(as thecrow flies)between populations. Euclidean genetic distances were derivedfrom the R matrixusing the formuladerived by Har- pendingand Jenkins(1973) foreach pair of populationsi andj: = - 4 rn + rjj lrU- (!)

Linguisticdistances were based on a hierarchicalclassification, which permits theconversion of theorder of relationshipsinto scores or distances(Crawford and Duggirala1992). Populationsspeaking different dialects of thesame sub- groupof languageswere assigneda distanceof 1. Populationsspeaking languages thatbelong to differentsubgroups within a givengroup of languages were assigneda distanceof 2. A distanceof 3 was assignedto populations speakingdifferent groups of languagesthat are in a given subbranch.Lan- guages thatfall into differentsubbranches of a given branchare separated by a distanceof 4. If the languagesbelong to two differentbranches within a subfamily,they are separatedby a distanceof 5. If the languagesbelong to two separatelinguistic families, a distanceof 6 is assigned.The scheme of Bee (1971) was used as the basis forthe classificationof the Romance languages. To examinethe interaction between genetic, geographic, and linguistic distance matrices,we computednormalized product-moment correlations, partialcorrelations, and multiplecorrelations using the mantel program(Re- lethford1990). Mantel's (1967) permutationprocedure, based on a general regressionapproach, is a valid tool formeasuring the correlation or associationbetween the elements of any two matrices.As such,the correspondence betweenthe two givenmatrices A and В can be measuredby usingthe Mantel teststatistic (Z): = Zab 2 AijBij, (2) ij

whereZab is the sum of cross-productsbetween Atj and Btpthe elementsof row i and columnj of matricesA and B. Because ZABis an unnormalized correlationcoefficient, it has to be normalizedinto a product-momentcor- - relationcoefficient that ranges from 1 to +1 (Dow and Cheverud1985; Smouse et al. 1986; Dow et al. 1987). In thismethod a samplingdistribution of the Mantel statisticis generatedthrough repeated simultaneous permuta- tionsof therows and columnsof one of the matrices.The significancelevel is obtainedas the proportionof the generatedMantel statistic(ZAB) that is greaterthan or equal to theobserved correlation (Relethford 1988). 904 /PICORNELL ET AL.

Distance measureswere comparedby computingPearson's product- momentcorrelations between the two given matrices,and the significance levels were ascertainedusing Mantel's permutationtest (Mantel 1967). The partialcorrelation approach was based on the regressionof each elementof the two distancematrices on a controlmatrix. Therefore partial correlationis thecorrelation between the correspondingelements in thetwo residualmatrices (Rx and R2) thatremain after the removalof the effectsof thecontrol matrix (Dow et al. 1987). Giventhree distance matrices A (genetic distance),В (geographicdistance), and С (linguisticdistance), three partial correlationcoefficients, rAß(C), rAC{B) , and rßC(A),were obtainedby using a least-squaresregression method, wherein the partialcorrelation rAB(C) indi- cates the associationbetween matrices A and В while keepingthe С matrix constant(Dow and Cheverud1985; Dow et al. 1987). Significancelevels for the partialcorrelations between the dependentdistance matrix (genetic distance) and one of the two independentmatrices (geographic or linguistic distance) were obtainedby controllingthe effectsof one of the two independentmatrices. Therefore the associationbetween the geneticand geographicdistances was exclusivelymeasured by keepingthe linguistic distance constant. Multiplecorrelations yield the relativeeffects of the two independent matrices( В and C) on thedependent matrix (A). In thismethod the expected values are computedthrough multiple regression analysis. The product- momentcorrelations are computedbetween the expected values and the ob- servedvalues of thedependent variable (Relethford 1990). Linkage disequilibriawere calculatedusing the programsLD79.FOR and LD86.FOR (Weir 1990).

Results

Phenotypenumbers and allele frequenciesof blood groupsand serum proteinsare summarizedin Tables 1 and 2. Withthe exception of theDuffy system,all loci were in Hardy-Weinbergequilibrium. The 2 Duffysystem showed a significantexcess of homozygotes(% = 5.5, d.f. = 1,/?< 0.05). A slightexcess of homozygoteswas observedin fiveof seven systems(MN, Ss, ORM, TF, and HP). However,a one-tailedsign testwas carriedout and showed no significance( p = 0.14). This slightexcess of homozygotesmay be the resultof inbreeding(Vails 1969). The significantexcess of homozy- gotesat theDuffy locus maybe theresult of selectionfor the FY silentallele, whichis not detectablewith a singleantiserum. High prevalenceof malaria on Ibiza could have acted as a selectiveagent. Mac Arthurand Wilson (1967) hypothesizedthat populations living in reduced geographicareas have reduced levels of heterozygosity.Thus the lower heterozygosityin Ibiza may be attributedto gene frequencydrift. To GeneticVariation in Ibiza /905

Table 1. DistributionofPhenotypes and Gene Frequencies of Blood Groups in Ibiza a System Phenotype N Allele Frequency /2 d.f. Significance MNSs MS 4M 0.517± 0.029 0.0 1 n.s. Сn = 150) MS,LS 11 N 0.483± 0.029 MLS 16 S 0.272± 0.028 3.1 1 n.s. M,NS 8 LS 0.728± 0.028 M,NS,LS 24 M,NLS 28 MS 0.180± 0.027 6.2 3 n.s. NS 1 MLS 0.316± 0.032 N S,LS 6 NS 0.093± 0.022 N LS 25 NLS 0.411± 0.034 M 9 M,N 15 N 3 P PI + 62 PI 0.234± 0.026 (n = 150) PI- 88 P2 + P 0.766± 0.026 Lewis A 31 Le 0.569± 0.041 (n = 124) В 70 le 0.431± 0.041 23 Duffy В 74 FY*A 0.382± 0.027 5.5 1 p < 0.05 (n = 146) A 21 FY*B 0.618± 0.027 А,В 51 Kell K+ 16 0.055±0.013 (n = 150) К- 134 Кс 0.945±0.013 ABO О 286 АВОЮ 0.786± 0.014 (п = 487) А 175 АВО*А 0.107± 0.014 В 20 АВО*В 0.107± 0.004 А,В 6 Rh D+ 384 D 0.540± 0.020 (п = 487) D- 103 d 0.460± 0.020 a. n.s.= notstatistically significant.

testthis hypothesis, we performeda correlationanalysis between mean per locus heterozygosityand the logarithmof the geographicarea using several Mediterraneanislands. The mean heterozygositiesof Alonissos (0.33), Ibiza (0.34), Minorca(0.35), Majorca (0.36), Sardinia(0.35), and Sicily(0.36) were regressedagainst area. The observedcorrelation (r = 0.485) was positivebut not statisticallysignificant. Ibiza, one of the smallestislands, had one of the lowest levels of heterozygosity.Thus the observeddeficiency of heterozy- gotes shouldnot be attributedto the island environment. The allele frequenciesfor Ibiza were comparedwith data fromother circum-Mediterraneanpopulations (Mourant et al. 1976; Roychoudhuryand Nei 1988; Tills et al. 1983; Alonso et al. 1990; Gameroet al. 1988; Sebentan and Sagisaka 1988). Based on its history,Ibiza is an admixedpopulation of NorthAfrican and Europeanancestry ; therefore this population should exhibit 906 /PICORNELL ET AL.

Table2. DistributionofPhenotypes and Gene Frequencies of Serum Proteins in Ibiza a System PhenotypeN Allele Frequency /2 d.f. Significance ORM Fl 59 ORM*Fl 0.518± 0.025 3.3 1 n.s. (n = 194) F1,F2 1 ORM*F2 0.008± 0.004 F2 0 ORM*S 0.474± 0.025 Fl,S 82 F2,S 2 S 50 GC 1 92 GC*1 0.684± 0.023 0.05 1 n.s. (n = 198) 1,2 87 GC*2 0.316± 0.023 2 19 TF CI 114 TF*C1 0.742± 0.020 1.6 1 n.s. (n = 198) C1,C2 55 TF*C2 0.210± 0.020 C2 11 TF*C3 0.020± 0.007 C1,C3 3 TF*B 0.028± 0.008 C2,C3 4 C3 0 Cl,В 8 C2,B 2 C3,B 1 В 0 PI Ml 121 Pimi 0.793± 0.020 1.29 1 n.s. (n = 198) M1,M2 34 PI*M2 0.096± 0.014 M2 1 /7*5 0.103± 0.015 Ml,S 35 PI*Z 0.008± 0.004 M2,S 2 S 2 Ml,Z 3 M2,Z 0 S,Z 0 Z 0 HP 1 30 НРЧ 0.381± 0.024 0.13 1 n.s. (n = 198) 1,2 91 HP*2 0.619± 0.024 2 77 a. n.s.= notstatistically significant.

intermediatefrequencies of geneticmarkers compared with its parentalpop- ulations.Some systemsdo indeedshow frequenciesthat resemble North Af- rican and Middle Easternpopulations. For example,the FY*A allele has a low frequencyin NorthAfrica and theMiddle East (0.177-0.293) buta higher incidencein Europeanpopulations (0.327-0.488). Ibiza exhibitsthe lowest FY*A frequencyobserved to date in Europe (0.319). The TF*C2 allele has one of thehighest values (0.210) in Europe,similar to values foundin some NorthAfrican populations (0.215). Populationsof the Middle East have the highestfrequency of the TF*C2 allele (0.275). The TF*C3 allele has a low GeneticVariation in Ibiza /907 incidencein theIbiza population,similar to thefrequencies found in southern Spain and in NorthAfrica. Some systemsindicate that Ibiza differsfrom European populations. For example,the frequencyof the MS haplotype(0.180) is low forEurope but does not differsignificantly from the frequencyin Turkey(0.190), Sar- dinia (0.195), and Crete(0.190). In theRh systemthe D allele has thelowest frequency(0.540) forEurope withthe exceptionof the Basques (0.489). In the PI systemthe P/*M2 allele frequency(0.096) is the lowest observedin Europeand otherMediterranean populations (0.140-0.200) butresembles the values seen in France (0.092) and Egypt(0.082). The PI*S allelic frequency is distributedalong an east-westcline in theMediterranean populations. The PI (P system)allele has a verylow frequencyin Ibiza (0.234), resembling the frequencyfor the populationof Malta (0.290). The frequencyof the АВОЮ allele in the Ibiza population(0.766) is amongthe highest observed in the Mediterraneanarea, with only Sardinia (0.779) and Trentino,Italy (0.769), surpassingit. Althoughthere is littlecomparative population data on the ORM system,Ibiza appears to have a low frequencyof the ORM*Fl allele. The frequenciesof the othergenetic markers studied in Ibiza are well withinthe ranges observed in Mediterraneanpopulations. The blood geneticdata were testedfor linkage disequilibria; however, no statisticallysignificant values were observed. In the principalcomponents analysis plot of the R matrix,the firstei- genvector(which accounts for 34.2% of the variance) separatesthe North Africangroups from the European populations (Figure 2). The secondeigen- vector(16.24% of the variance)distinguishes the Middle Easternand Greek populationsfrom Italian and Sardiniangroups. Minorca and Majorca appear to clusterclosely withsouthern Spain, Catalonia,and Corsica. Ibiza is inter- mediatebetween Spain, theother Balearic islands,and Sardinia.This plot is suggestiveof Ibiza' s uniquenessfrom Majorca and Minorca,thus reflecting the influenceof unique historicalevents. Figure3 shows whichalleles contributedto the dispersalof the popu- lationsin theprevious two-dimensional gene map (Figure2). Ibiza is distinguishedfrom the otherBalearic island populationson the basis of the frequencies of the TF*C3 and P2 + p alleles. The NorthAfrican populations are distinguishedfrom the Europeangroups by the highincidence of GC*7, F7*2 + FF*?, and ACP*B alleles. The RSTvalue forthe Mediterraneanpopulations, a measureof genetic heterogeneity,is relativelylow, if viewed on a worldwidebasis, buthigh for Europeanpopulations (Jorde et al. 1982). The RSTvalue of 0.028 is slightly lower thanthe values recordedfor Australian aboriginal populations (0.04) and for East HighlandsNew Guineans (0.038). The European populations have values in the range0.0002-0.0136 (Jordeet al. 1982). The relationshipbetween mean per locus heterozygosityH and distance fromthe centroid ru is shownin Figure4. Ibiza exhibitsboth a highhetero-

This content downloaded from 129.237.46.100 on Mon, 22 Dec 2014 14:36:38 PM All use subject to JSTOR Terms and Conditions 908 /PICORNELL ET AL.

Figure2. Least-squaresreduction genetic map of 15 Mediterranean populations based on allele frequenciesforthe ABO, Rh, MNSs, Kell, P, FY, PGD, PGM, GC, ACP, HP, and TF loci.A totalof 50.44% of the variance isexplained bythe plot of the first two eigen- vectors.

zygosity(possibly reflecting the admixtureof thefounding populations) and relativelygreat distance from the centroidof distribution(most likelythe resultof stochasticprocesses). By far,the mostisolated population appears to be the Algeriansample, which has a highdistance from the centroidand a low heterozygosity.

Concordances betweenDistances. Table 3 summarizesthe results of the pairwiseMantel tests.At the populationlevel the product-momentcorrela- tionsbetween two of the threedistance matrix comparisons are statistically significant.A relationshipexists between linguistic distance and bothgenetic and geographicdistance, with a correlationbetween genetics and language of 0.152 (p > 0.05) and a correlationbetween geography and languageof 0.178 ( p = 0.05). Yet thereis no apparentrelationship between genetics and geography. Partialcorrelations were made on genetic,geographic, and linguistic distancematrices. When either the language or geographywere kept constant, therewas no statisticallysignificant association between genetics and geog- raphyor betweengenetics and language(Table 4). GeneticVariation in Ibiza /909

Figure3. Least-squaresreduction genetic map of the 35 allelesof the 12 loci corresponding to themap in Figure 2.

A multipleregression analysis of one distancematrix on theother two matricesalso failedto indicateany statisticallysignificant relationships be- tweengeography, genetics, and language (Table 5). Apparently,the effects of migrationand conquesthave swampedmost of the effectsof stochastic processesthat under genetic and geographicisolation would accentuatedif- ferencesbetween populations. The only statisticallysignificant relationship, althoughmarginal, exists between language and geneticsand geography. Thus language servesas the best predictorof geneticvariation in the Medi- terraneanpopulations.

Discussion

Our resultsindicate that genetic differences exist between Ibiza and the othermain Balearic islands (Majorca and Minorca).These threepopulations have similarhistories except for the Carthaginian original settlement of Ibiza. As a result,Ibiza has closer geneticaffinities to the Middle East and North 910/ PICORNELLET AL.

Figure4. Plotof mean per locus heterozygosity H against distance from the centroid ofdistribution ruof the relationship matrix for 15 Mediterranean populations.

Table 3. Pearson'sProduct-Moment Correlations for Genetic (GEN), Geographic (GEOG),and Linguistic (LANG) Distance Matrix Comparisons among Ibiza and Other MediterraneanPopulations Testof Relationship Correlation Significance(p) GEN*GEOG 0.119 0.127 GEN*LANG 0.152 0.068a GEOG*LANG 0.178 0.058" a. p > 0.05. b. p = 0.05.

Table 4. PartialCorrelation between Two MatricesWhile Controlling for the Third Matrix

Testof Relationship Correlation Significance(p) GEN*GEOG.LANG 0.095 0.189 GEN*LANG.GEOG 0.132 0.112 GeneticVariation in Ibiza /911

Table 5. MultipleCorrelation Coefficients of One DistanceMatrix on Two Other Matrices

Testof Relationship MultipleCorrelation Significance(p ) GEN*GEOG.L ANG -0.154 0.969 GEOG*GEN.LANG -0.121 0.865 LANG*GEN.GEOG -0.241 0.991

Africathan do the otherBalearic islands. Thereforeour resultsappear to confirmthe historically derived hypothesis of a Carthaginianorigin of Ibiza. In thiscase thegene pool of thepopulation is a reflectionof itsunique history. The geneticdifferentiation of Ibiza fromthe other two Balearic islands could have been influencedby a foundereffect and otherstochastic processes. The plotbetween mean perlocus heterozygosityand rureveals that, although Ibiza has levels of heterozygositysimilar to thoseof Minorca and Majorca, the distancefrom the centroidof distributionis greater.This studydemon- stratesthat unique historicalevents combined with the action of stochastic processescan explain muchof the observedgenetic variation in island populations.However, the actionof naturalselection cannot be ignoredin Ibiza because of the high incidenceof malaria and the observedstatistically sig- nificantdeviation of the Duffysystem from expectation. The significantpairwise correlations between linguistic distances with geographicand geneticdistances - withno apparentassociation demonstrated withpartial and multiplecorrelations - suggestthe effects of muchhistorical "noise" on Mediterraneanpopulations. For example,the migrationof Cata- lans to the Balearic Islands followedby the acquisitionof the Spanish language by theoriginal inhabitants of Ibiza complicatesthe simple associations thatone would expectto findin island populations.The effectsof stochastic processesare also compromisedby thehigh level of gene flow.

AcknowledgmentsWe aremost indebted to SarahNevo and R. Guerrerofor their valuablehelp in the correction of this manuscript. This research was partly supported by the DireccisnGeneral de Investigacismen Ciencia y Tecnologia(DGICYT) throughgrant SAF92-1197 awarded to M.M. Ramon.

Received27 July1994; final revision received 29 March1996.

Literature Cited

Alonso,A., G. Visedo,M. Sanchoet al. 1990.Isoelectric focusing inminiaturized gelsappli- cationto Gc,Pi, Tf, and ORM subtyping in central Spain. In Advancesin Forensic Haemogenetics3,H.F. Polesky and W.R. Mayr, eds. Berlin: Springer Verlag, 255-262. 912 /PICORNELL ET AL.

Вес,P. 1971.Manuel Pratique de Philologie Romane II. Paris,France: Picard. Carracedo,A., I. Concheiro,J.Estefanía et al. 1986.Polymorphism ofserum orosomucoid: Familyand population studies in Galicia. In Advances inForensic Haemogenetics 1, B. Brinkmanand K. Henningen,eds. Berlin: Springer Verlag, 282-285. Constans,J.,P. Kuhnl,M. Viau et al. 1980.A newprocedure forthe determination oftransferrin С ( TF*C)subtypes by isoelectric focusing: Existence of two additional alleles TF*C4 andTF*C5. Hum. Genet. 55:111-114. Crawford,M.H., and R. Duggirala.1992. Digital dermatoglyphic patterns ofEskimo and Am- erindianpopulations: Relationships between geographic, dermatoglyphic, genetic, and linguisticdistances. Hum. Biol. 64:683-704. Crawford,M.H., T. Koertvelyessy,R.G.Huntsman etal. 1995.The effects ofreligion, economics,and geography onthe genetic structure ofFogo Island, Newfoundland. Am.J. Hum. Biol.7:437-451. Dow,M.M., and J.M. Cheverud. 1985. Comparison ofdistance matrices instudies of population structureandgenetic microdifferentiation: Quadratic assignment. Am.J. Phys. Anthropol. 68:367-373. Dow,M.M., J.M. Cheverud, and J.S. Friedlaender. 1987. Partial correlations ofdistance ma- tricesin studies of population structure. Am. J. Phys. Anthropol. 72:343-352. Edwards,A.W.F. 1971. Distances between populations onthe basis of gene frequencies. Bio- metrics27: 87 3-88 1 . Gamero,J., M.V. Lareu, M.S. Rodriguez etal. 1988.Polimorfismo dela orosomucoide(ORM) enla provinciadeCadiz, mediante focalizacion isoeléctrica engel de poliacrilamida. II. Jornadas.An. Soc. Esp. Med. Legal Forense 2:155-158. Harpending,H.C., and T. Jenkins.1973. Genetic distances among southern African populations. InMethods and Theories of Anthropological Genetics, M.H. Crawford and P.L. Work- man,eds. Albuquerque, NM: University ofNew Mexico Press, 177-199. Jorde,L.B., P.L. Workman, and A.W. Eriksson. 1982. Genetic microevolution inthe Aland Islands,Finland. In CurrentDevelopments inAnthropological Genetics, v. 2, Ecology andPopulation Structure, M.H. Crawford and J.H. Mielke, eds. New York: Plenum Press,333-365. MacArthur,R.H., and E.O. Wilson. 1967. The Theory of Island Biogeography. Princeton, NJ: PrincetonUniversity Press. Mantel,N. 1967.The detection ofdisease clustering and a generalizedregression approach. CancerRes. 27:209-220. Mielke,J.H., E.J. Devor, P.L. Kramer etal. 1982.Historical population structure ofthe Aland Islands,Finland. In CurrentDevelopments inAnthropological Genetics, v. 2, Ecology andPopulation Structure, M.H. Crawford and J.H. Mielke, eds. New York: Plenum Press,255-332. Miguel,A., and E. Petitpierre.1989. Red cell enzyme polymorphisms inIbiza (Balearic Islands, Spain).Hum. Hered. 39:351-355. Montiel,M.D., A. Carracedo,J.R. Blazquez et al. 1990.Orosomucoid (ORMI and ORM2) typesin theSpanish Basque Country, Galicia, and northern Portugal. Hum. Hered. 40:330-334. Montiel,M.D., A. Carracedo,I. Lopez-Rodriguez etal. 1988.Comparison between isoelectric focusingmethods for the detection oforosomucoid phenotypes. Electrophoresis 9:268- 272. Moral,P. 1987.Estudio antrogenetico de diversos polimorfismos hematologicos enla islade Menorca.Ph.D. thesis, Universidad deBarcelona, Barcelona, Spain. Mourant,A.E., A.C. Kopec, and K. Domaniewska-Sobezak.1976.The Distribution ofthe Hu- manBlood Groups and OtherPolymorphisms. Oxford, England: Oxford University Press. Relethford,J.H. 1988. Estimation ofkinship and genetic distance from surnames. Hum. Biol. 60:475-492. GeneticVariation in Ibiza /913

Relethford,J.H. 1990. MANTEL: A microcomputerprogram for computing theMantel prob- abilitybetween distance matrix elements. Unpublished. Roberts,D.F. 1968.Genetic effects ofpopulation size reduction. Nature 220:1084-1088. Roychoudhury,A.K., and M. Nei.1988. Human Polymorphic Genes World Distribution. Ox- ford,England: Oxford University Press. Sebentan,I., andK. Sagisaka.1988. Genetic polymorphisms ofthe orosomucoid ORMI and ORM2in Libyans: Occurrence ofORMI *2.1 and three new ORM2 alleles. Jpn. J. Hum. Genet.33:439^443. Smithies,O., G.E. Connell, and G.E. Dixos. 1962. Inheritance ofhaptoglobin subtypes. Am. J. Hum.Genet. 14:14-21. Smouse,P.E., J.C. Long, and R.R. Sokal. 1986. Multiple regression and correlation extensions ofthe Mantel test of matrix correspondence. Syst.Zool. 35:627-632. Tills,D., A.C.Kopec, and R.E. Tills. 1983. The Distribution ofthe Human Blood Groups. Oxford,England: Oxford University Press. Vails,A. 1969.Inbreeding frequencies inthe Balearic Islands (Spain). Z. Morphol.Anthropol. 61:343-351. Weir,B.S. 1990.Genetic Data Analysis. Sunderland, MA: Sinauer Associates.