close

Вход

Забыли?

вход по аккаунту

?

Efficiency of cranial bilateral measurements in separating human populations.

код для вставкиСкачать
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 83:307-319 (1990)
Efficiency of Cranial Bilateral Measurements in Separating
Human Populations
I. HERSHKOVITZ, B. RING, AND E. KOBYLIANSKY
Department of Anatomy and Anthropology, Sackler Faculty of Medicine,
Tel Aviv University, Israel
KEY WORDS
Isolates, Skull discrimination, Asymmetry
ABSTRACT
A set of 31 nonconventional paired cranial measurements, as
well as six Conventional nonpaired measures, were Laken o n 266 skulls,
representing two related populations: Bedouins of the Israeli Negev Desert
and Bedouins of the Sinai. The data were subjected to univariate and discriminant analyses to determine the relative efficacy of paired vs. conventional
measures in sorting individuals according to tribal and sex affiliation. It was
found that paired measures have greater discriminatory power (87%) than
conventional ones (47%) in terms of classifying individuals belonging to
human isolates derived from a common ancestor and sharing similar environmental conditions. This greater discrimination attests to the value of the level
of “developmental noise” (a measure provided by fluctuating asymmetry) in
sorting human populations. Possible explanations are proffered for the above
finding.
As pointed out by Woo (1931) many years
ago, “Racial differences were approached
from the standpoint of appearance, in other
words from the conception of portraiture.
Anthropometricians endeavoured to give
quantitative value to the differences that
were obvious to them at first sight.” It is not
surprising, then, that a large set of cranial
measurements deemed to be most representative of skull morphology was proposed, and
that these were used in intrapopulation and
interpopulation analyses in anthropology for
over a century (for early literature review,
see Falk and Corruccini, 1981).In time, however, selection of the best discriminatory
traits was im roved by using multivariate
techni ues ( owells, 1969; Relethford,
1988).%actorand principal component analyses supported the prevailing notion (e.g.,
Young, 1957; Moss and Young, 1960) that
the crania of humans and animals consist of
a number of varying units that res ond in
s ecial ways to a variety of externa forces
owells, 1973; Key and Jantz, 1981; Cheverud, 1982). This is consistent with the results of twin studies (Nakata et al., 1974)
that evinced the existence of several genetic
and environmental components from the
craniometric data.
a
(4
0 1990 WILEY-IXS, INC
B
The success of discriminant methods
based on selected cranial morphological
traits depends on the heritability level of the
latter. Indeed, the belief in a strong genetic
component of craniofacial traits encouraged
physical anthro ologists to use them in
drawing up racia taxonomies. However, numerous studies (e.g., Nakata et al., 1974;
Susanne, 1977; Sharma, 1987) have shown
that considerable variation can be found in
the estimates of heritability for different
craniofacial traits (e.g., head length = 0.90;
head breadth = 0.72; head height = 0.66;
estimates of heritability from twin studies,
Sharma, 1987).
Studies that used traditional measurements (Martin and Saller, 1957) usually
failed to discriminate between skulls from
closely related populations (Arensburg et al.,
1987). In fact, only opulations remote in
space and cline could e separated efficiently
by traditional measurements (Howells,
1969; Falk and Corruccini, 1981).
In the present paper, we adopted a different approach toward cranial discrimination
P
g
Received May 23.1989, accepted January 4.1990
Address reprint requests to Dr I Hershkovitz, Dept of Anatomy and Anthropology, Sackler Faculty of Medicine, Tel Aviv
University, Ramat Aviv, Israel
308
1. HERSHKOVITZ ET AL
in that we focused on skull asymmetry
rather than on general skull morphology.
Rates of fluctuating asymmetry, defined as
the random difference between both sides of
a paired trait in an individual, are believed
to reflect the magnitude of developmental
stability of an organism, that is,
tion to deviation from norma ontogeny
(Livshits and Kobyliansky, 19871,and correlate with opulational homozygosity level
(Kobylians y and Livshits, 1983). Hence we
concluded that paired
measurements may be
..
used more effiaentlv to discriminate hetween high1 inbred human populations,
even when t ey are derived from the same
ethnic stock and live under similar environmental conditions. This conclusion was
based on the assumption that, due to geographic and social isolation and a long inbreeding process, the human groups, in the
course of time, would tend to become genetically divergent. The small size o f the groups
and their strong emphasis on consan
ous marriages will result in a s t e a Y E 1
crease in the homozygosity of indivi uals
within each group. These homozygous individuals would be less able to buffer against
developmental disturbances and hence will
tend to deviate from perfect symmetr more
than heterozygotes (Soule, 1979; KO yliansky and Livshits, 1983, 1985; Leary et al.,
1984; Livshits and Kobyliansky, 1984;
Palmer and Strobeck, 1986). It would follow,
then, that individuals of different Bedouin
tribes who vary (due to different inbreeding
levels among the tribes! in their levels of
developmental homeostasis will manifest
ciifferent rates of cranial morphoiogicai fiuctuating asymmetry-a feature that would be
of discriminatory value.
The rate and magnitude of skull symmetry
has never been used in sorting human skulls
into group affiliations. Rather, the vast
amount of research carried out on skull
asymmetry in humans and nonhumans has
been focused mainly on the different functions of the right and left hemispheres of the
brain or on the masticatory mechanism (e.g.
Woo, 1931; Vig and Hewitt, 1974).The skull
complex, however, consists of several paired
and unpaired bones and therefore manifests
characteristics that display some degree of
asymmetry (such as size differences between
homologous bones or between bilateral
parts 1.
The relative contribution o f genetic and
environmental factors t o the variation in the
predisposi-
E
K
t
level of fluctuating asymmetry is as yet unclear (Palmer and Strobeck, 1986; Livshits
and Kobyliansky, 1988). Many reports have
shown that environmental stresses can significantly elevate fluctuating asymmetry in
humans as well as in animal species (e.g.,
Siege1 and Smookler, 1973; E es er et al.,
1986). Estimates of fluctuating asymmetry
heritability are usually found to be low and
not statistically significant (e.g., Leamy,
19861, but recently Livshits and Kobyliansky (1988) found that the mean FA values of
Several metric traits indeed reveal a statistically significant genetic component.
The present paper describes a method for
discriminating between related inbred populations by utilizing the asymmetrical nature
of the skull.
MATERIALS AND METHODS
Two Bedouin populations, one from the
Israeli Negev and the other from South
Sinai, were studied. From a historical point
of view, the Bedouins who today inhabit the
Negev and South Sinai regions are only a
fraction of the total Bedouin population of
the Near East and Arabian Peninsula. They
began to settle in the region of the Negev and
Sinai between the 10th and 14th centuries
C.E. (Sharon, 1977).The Bedouin tribe of the
Negev included in the present study is
within the social framework of the “Tayaha”
tribal affiliation, whereas the studied Sinai
tribes belong to the “Towara” tribal affiliation (Hershkovitz, 1985). These two groups
of tribes are geographically and socially isolated and have no intermarital contacts.
Aimost 97% of Sinai Bedouin males (Her siikovitz, 1985) and 86% of Negev males (Muhsam, 1966)find their mates within their own
tribes. Marriages within the extended family, with preference for first-cousin matings,
are very common among Sinai and Negev
Bedouins, and may comprise more than 30%
of the total (Hershkovitz, 1985).Unlike other
Bedouin tribes of the Near East and Saudi
Arabia, the Sinai and Negev tribes are small,
the Negev tribes being generally larger than
those of the Sinai (Muhsam, 1966; Hershkovitz, 1985).
Concerning the origin of the material and
its affiliation, the skeletal remains of the
Negev Bedouins were recovered in ancient
burial caves from the Roman-Byzantine periods near Kibbutz Lahav (Fig. l ). The reuse
of ancient burial caves by Israeli Bedouins is
quite common. From coins, beads, metal
309
SEPARATING HUMAN POPULATIONS
bracelets, and other artifacts accompanying
the osteological material, the skeletons were
dated to the late 19th century. Historical
documents (Muhsam, 1966; Sharon, 1977)as
well as direct testimony of the present Bedouins of the region (Marx, 1974; Bailey,
1980) confirm the origin of the material and
its affiliation to the Ramadin tribe (still living in this region). The skeletal material
from the Sinai Desert came from two main
sites, Wadi Qid and Wadi Slaf (Fig. 1). From
the artifacts and testimonies of the Bedouins
themselves, the skeletons are affiliated to
the Muzeina and Awlad Said tribe.. (amnng
the larger tribes of the Towara). According to
the literature (Ben-David, 19781, there is
documentation that a severe epidemic
spread through these regions during the
later 19th century, an account of which also
a pears in many Bedouin folk stories
ershkovitz, 1985). The existence of mass
burials seems to sup ort this notion.
Because the tribafaffiliation of both the
Negev and Sinai skeletal sample could be
determined, we were able, through Bedouin
history books and documents, as well as
other sources (Hershkovitz, 19851, to follow
their tribal history almost until their time of
(fi
______
---I
7
Fig. 1. Locations of sites in the Negev and Sinai
Deserts
separation from a common mother population more than 600 years ago.
In the course of the present investigation,
cranial measurements were taken on 266
adult Bedouin skulls, of which 164 were from
the Negev and 102 from the Sinai; the skulls
are currently housed in the Department of
Anatomy and Anthropology, the Sackler
Faculty of Medicine, Tel Aviv University.
All measurements were taken by one of
the authors (B. Ring). The measured coefficient of repeatability was high (r = 0.981.
The anthropometric criteria comprised 31
new paired measures (chor6s 2nd XCC) mr!
six traditional single measures (Table I,
TABLE I. List of variables used in the present study'
Code
Variable
Measures of
1. PBC
PBA
2. SBC
SEA
3. PLC
PLA
4. SLC
SLA
5. ALE
ALA
6. ABC
ABA
7. NAC
NAA
8. OAC
OAA
9. PNSS
10. PSQ
11. SON
the cranium2
= porion-bregma chord
= porion bregma arc
= sphenion-bregma chord
-= sphenion-bregma arc
= porion-lambda chord
= porion-lambda arc
= sphenion-lambda chord
= sphenion-lambda arc
= asterion-lambda chord
= asterion-lambda arc
= asterion-bregma chord
= asterion-bregma arc
= nasion-asterion chord
= nasion-asterion arc
= opisthion-asterion chord
= opisthion-asterion arc
= parietal notch-sphenosquamosal suture
= porion-squamosal suture
= supraorhital notch-midcoronal suture
(lateral frontal arc)
L
\
* ...... .
edbures of the
12. NEL = nasion-ectoconchion length
13. ORBH = orbital height
14. ORBW = orbital width
15. MZL = maximum zygomatic length
16. MZH = maximum zygomatic height
17. PZA
= prosthion-zygomaxillare anterior chord
18. NSA = nasal spine-alare chord
Measures of the base of the skull'
19. BML = basion-mastoid length
20. RSS
= basion-sphenoid spine chord
21. SSF
= sphenobasion-stylomastoid foramen
chord
22. SBP
= sphenobasion-pterion chord
23. FIM
= foramen incisivum-tip of mastoid chord
Conventional measures
CIND = cephalic index
FIND = facial index
RIND = basal index (sphenobasion-Op/Po-Po)
BIAB = biauricular breadth
MAXL = maximum cranial length
MAXR = maximum cranial breadth
'For location of measures, see Figures 1 a n d 2.
'Measured on both sides of the skull.
310
I. HERSHKOVITZ ET AL.
Fig. 2. Paired measures taken on skull, lateral view.
Explanation of numbers is given in Table 1.
Fig. 3. Paired measures taken on skull, basilar view.
Explanation of numbers is given in Table 1.
Figs. 2, 3). The paired measures were selected in order to characterize the three main
anatomic regions of the skull, as follows:
calvarium, 19 measures; face, seven measures; and base of the skull, five measures.
Of these 31 measurements, half were corresponding measurements on single homologous bones, and the rest spanned t,wo or
more bones. For each skull, age and sex were
determined independently by three experienced anthropologists. Where opinions diverged, the skull was reexamined and the
majority decision accepted.
All skulls were assigned into three age
groups according to suture closure, as follows: 1)18-30 years (117 crania), all ectocranial sutures still open; 2) 30-45 years (106
crania), ectocranial sutures partially closed;
and 3) 45+ years (43 crania), ectocranial
sutures synostosed. The degree of agreement
regarding assignment of the skulls into the
above age groups was over 90%. We found
the age distribution to be similar in both the
Sinai and Negev samples.
Determination of sex was based on size of
the glabella and supraciliary eminence, inclination ofthe frontal bone, frontal bossing,
shape of the orbit, thickening of the orbital
311
SEPARATING HUMAN POPULATIONS
margin, and general gracility of the skull.
Size of the mastoid process and the relief of
the planum nuchale, which are usually used
in sex determination, were ignored here because Bedouin women tend to resemble
males in this respect, as they regularly carry
heavy burdens on their heads. Sex ratio
(maledfemales x 100)for the Negev sample
was 102.5 and for the Sinai group 142.9.
Agreement achieved in our sex classification
was over 80%. It is noteworthy that anthropological studies on living Bedouin populations in the Negev (Muhsam, 1966)and Sinai
(IIei SlikuVitL, ISS5) h a v e reveaied simiiar
results regarding sex ratio.
Statistics
For each characteristic, the mean (X) and
standard deviation (SD) were calculated. By
analysis of variance with one factor interchange (ANOVA),the variability of the studied traits in two Bedouin groups was determined and compared separately for each sex
(Nie et al., 1975).
The subprogram Discriminant was used to
generate canonical discriminant functions
(Nie et al., 19751, which in turn allowed
classification of individuals on the basis of
their proximity to the various group centroids. We carried out three separate classifications: of male skulls according to tribal
affiliation; of female skulls according to
tribal affiliation; and of all skulls together
according to tribal affiliation and sex. Individuals with any missing data were excluded
from the analysis.
RESULTS
Although there were some differences in
ANOVA results between sets ofright and left
measures, we present here only the mean
values and standard deviations for the measures taken on the right side of the skull
(Table 2). The Sinai and Negev females differed significantly in six of 31 right side
traits (SBC, SBA, NAC, NAA, SON, and
SSF) and the males in five traits (SON,ALC,
ABC, ABA, and SSF).Only two traits (SON
and SSF) were significantly different in both
the male and female intratribal comparisons. For left-side measures, the males differed significantly in nine traits (PBC, PLC,
SON, ALC, ALA, ABC, ABA, ORBH, and
SSF),but in females no significant discriminating left trait was found. Of the six conventional measures (Table 11, cephalic index
and maximum cranial breadth showed sig-
nificant differences among the females; facial index was significantly different among
males (Table 2).
Of the total 68 measures (31paired and 6
single) used in the discriminant analysis
(mimimum F to enter = l),33 manifested
significant discriminatory ower for the
combined sample (Table 3, le t column). Conventional measures selected for cranial identification, such as cephalic index and basal
index, manifested low Wilks lambda values
(Table 3). The contribution of paired measures to Bedouin skull differentiation is evident from the results ofthe stepwise variable
selection; i.e., in nine variables, both the
measures on the right and left sides were
selected by the stepwise procedure. Correlations of bilateral measures in order of decreasing intensity appear in Table 4.
If we examine the composition of the 15
variables with the highest Wilks lambda
values (Table 31, we find the following: 1)
nine of the 15 variables are measures of
individual bones; 2) none of the nonpaired
variables or indices appear in the list of
traits; 3) nine of the variables are measures
taken on the right side and six on the left; 4)
although about half of the 31 paired measures manifest correlation coefficients
higher than 0.800, of the group of 15 variables with the highest Wilks lambda, 67%
manifest a correlation coefficient lower than
0.800, which indicates a clear tendency toward the selection of traits with higher FA
levels; 5) nine of the 15 variables are measures of the calvarium, three are of the face
and three of the base of the skull: and 6) two
pairs of right-left measures a pear amorg
the first 15 variables (OAC ancf ABC).
Analysis of the first 15 measures with the
highest FA level (0.379-0.592), calculated as
1-r2,(Table 4),shows that nine are measures
of the calvarium, four are of the base of the
skull, and only two of the face, thus reiterating the well known fact that the face is the
most symmetrical part of the skull (Woo,
1931).
In the analysis, three functions were obtained, the first two accounting for 90% of
the between-group variance. Function 1 explains 51% of the between-group variance
and Function I1 39%.
The pooled within-group correlations between canonical discriminant functions and
discriminating variables are given for all
three functions in Table 5. Function I has its
highest correlation with the right asterion-
fp
312
1. HERSHKOVITZ ET AL,
T A B L E 2. Basic descriptive statistics for two Bedouin populations, b y sex
Group'
___
N
4
80
76
57
35
1
2
3
4
78
77
56
35
I
2
3
4
'1'1
1
2
3
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
X
RPBC
124.4
121.5
125.6
121.1
RPLC
115.2
112.0
115.9
112.5
RALC
84.2"
SD
N
4.7
7.8
5.1
4.0
79
76
53
33
5.0
5.5
4.8
3.7
79
78
56
36
b.1
78
83.0
4.2
86.2*
5.0
83.0
4.0
RNAC
78
147.2
4.5
75
140.6"
4.6
52
148.1
5.7
32
143.0*
4.7
RPNSS
75
65.4
4.4
76
62.2
4.2
52
65.3
4.9
35
62.6
3.8
RORBH
78
32.9
2.2
2.2
78
32.9
56
32.3
1.8
37
32.6
2.4
RPZA
_______________
74
54.8
3.0
73
53.6
3.3
55
54.6
3.5
35
53.5
3.9
RSSF
81
50.2*
2.4
80
48.9*
2.3
55
51.9*
3.1
2.3
36
50.5"
80
56
35
76
55
34
79
79
52
33
76
74
52
35
80
79
56
37
77
79
54
37
71
76
50
34
CIND
1
2
3
4
1
2
3
4
79
73.5
2.8
77
74.2*
2.4
55
72.5
2.8
37
72.9*
2.7
~ _ _ _ BIAB
_ _ _ _ _ _ _
81
100.5
4.5
79
96.5
4.7
57
100.7
5.2
38
97.3
4.4
79
78
57
36
80
79
55
39
X
RPBA
153.7
1,50.7
154.2
148.4
RPLA
139.4
135.9
140.3
136.3
KALA
94.0
92.9
95.5
92.3
RNAA
182.4
174.2*
183.8
177.3"
RPSQ
49.3
47.5
49.0
46.7
RORRW
39.7
38.9
39.8
39.4
RNSA
16.0
14.9
15.8
14.8
RSBP
71.2
69.0
70.6
68.5
FIND
73.1*
72.2
70.1*
70.4
MAXL
183.8
178.2
184.7
178.7
SD
N
6.9
7.2
6.4
5.1
71
73
53
33
6.3
5.0
5.4
5.0
70
73
52
32
5.1
6.0
5.0
,
76
53
33
5.9
6.2
7.1
6.0
78
72
53
36
4.7
4.1
4.2
3.4
79
80
58
35
1.9
2.0
1.8
1.4
61
61
48
26
1.4
1.7
1.3
1.4
73
72
53
31
3.3
3.4
4.2
3.7
70
72
50
29
4.7
4.7
4.7
4.8
77
75
50
36
6.2
5.5
5.9
5.8
80
78
57
37
6.2
nt
X
RSRC
94.0
92.3*
93.7
89.9*
RSLC
140.5
138.0
140.8
135.8
RABC
134.3"
130.0
137.2*
131.6
ROAC
61.4
60.0
62.0
60.2
RSON
103.5*
99.9*
100.2*
96.2*
RMZL
53.3
50.1
52.3
49.9
RBML
50.9
48.7
51.8
49.3
RFlM
106.0
102.5
106.8
103.6
BIND
65.7
65.7
65.1
65.6
MAXB
134.8
132.2*
134.0
130.1*
SD
N
~.
5.0
5.3
4.8
4.3
6.0
7.1
5.7
5.1
71
75
54
35
___
70
73
52
34
5.1
4.3
4.9
4.3
78
79
54
34
2.7
2.9
2.5
2.8
78
75
53
36
6.7
6.2
6.4
4.8
80
80
57
38
3.1
2.9
3.6
3.7
73
74
57
35
2.4
2.5
2.6
2.1
69
67
54
37
X
RSRA
110.7
109.0*
111.1
105.9'
RSLA
173.2
169.4
172.8
167.4
RARA
164.2'
159.3
167.5*
160.4
ROAA
68.5
66.6
68.6
66.5
RNEL
50.7
49.2
51.2
49.4
RMZH
47.3
45.0
46.6
45.4
RBSS
39.3
37.5
39.2
37.8
_-SD
6.4
6.2
6.9
6.3
8.1
7.0
7.3
6.0
6.9
6.3
6.3
5.3
3.5
3.6
3.5
3.2
1.9
2.1
1.9
1.9
2.9
2.5
2.9
2.9
2.0
2.2
2.7
1.9
4.7
4.6
5.2
4.6
4.5
4.6
4.4
3.4
5.7
4.9
5.3
4.4
*Significantly different a t the level of P 10.05. between the same sex a t different sites.
1 = Negev males; 2 = Negev females; 3 = Sinai males; 4 = Sinai females. See 'Fable 1 for description of trait, abbreviations.
'
bregrna chord and the left asterion-bregma
arc, and Function I1 with maximum cranial
length and right nasion-asterion arc. Function I11 has low correlations with most variables (Table 5).
The Mahalanobis distances between
group centroids are significantly different
for all comparisons (P < 0.01). The group
centroid values are presented in Table 6 for
the three functions. For Function I the polar
groups are Sinai males and Negev females,
with values of 2.15 and - 1.78, respectively;
313
SEPARATING HUMAN POPULATIONS
TABLE 3. Stepwise variable selection: summary table
For sex and tribal affiliation
Entered
Wilks
traits
lambda
______________
,581
RARC
,459
RSON
.422
RNSA
,402
RSSF
.363
1x33
LOAC
,302
LNAC
,302
RBML
,276
,264
ROAC
.257
LALA
.ZSY
XfZA
230
RSLC
.214
RMZH
LAX
,188
LABA
,175
RNAL
,167
.158
LSSF
RIND
.I45
,141
RNAC
RMZL
.I31
.I33
RPLC
.127
RBSS
RORBW
.124
CIND
.120
LNAL
,117
.111
RABA
.106
LALC
RPLA
.lo3
LPLA
.098
.091
LORBH
L.NAA
,087
,081
RORBH
Males for tribal affiliation
Entered
Wilks
traits
lambda
-Females
- _ _ _ _ _ _for
_ _tribal
_ _ _ _affiliation
___~
LABC
RSON
RPZA
LOAC
MAXBR
RABA
LPZA
T,NSA
LPNAP
RBSS
LSKA
LBML
RSBC
LOAA
LABA
ROAA
LSBA
LSBC
LSLC
CIND
ISON
RMZL
LNAA
_ _ .
,756
,600
.455
,435
,416
,404
,390
,381
,365
,360
,346
.338
,328
.313
-307
,301
,293
,289
,280
.269
Entered
traits
Wilks
lambda
---_
RSSF
,901
,615
543
RNAA
RNAI,
LBSS
LNAC
MAXBR
LSBP
LALA
LOAC
,494
.443
,416
,390
.363
,343
RRSS
21 Q
LSBA
RABA
LBML
RABC
RMZH
1,ORBI-I
FIND
RORBW
RPST
,291
.266
,232
223
,213
,205
.195
,188
,181
.271
.263
,253
TABLE 4. Correlations of bilateral measures in order of intensity and fluctuating asymmetry level for
31 pairs of craniofacial traits
Measure
PBC
SON
PNAP
NAC
FIM
ORBH
PSQ
SLC
NEL
MZL
PLA
NAA
ALC
PLC
MZH
'Calculated as l-?
Correlation
coefficient r
,928
,891
,886
,885
382
379
,866
,861
FA level'
Measure
.139
SLA
SRP
ABC
ALA
ORBW
SSF
PLA
BSS
NSA
OAA
,207
215
,217
,222
,228
,250
,259
,859
,850
,262
,849
,842
,834
,832
,279
,291
,802
.278
.305
,308
.357
SRC
.OAC
PRA
SBA
ABA
BML
Correlation
coefficient r
,797
.788
.786
.774
,770
,764
,761
,755
.747
.740
,735
,707
.701
,667
.662
,639
F A level'
.365
.379
,381
.401
,407
.417
.421
.430
.442
,453
.460
.501
509
,555
,562
,592
314
1. HERSHKOVITZ ET AL.
TABLE 5. Pooled withira-group correlations between canonical discriminant functions and
discriminating variables
Function
TI
I
Variable
RARC
LABC
RABA
RNAA
RNAC
LABA
LPLC
RBML
RPLC
RSSF
LBML
MAXI,
LSSF
LPBA
BIAB
537
515
429
417
388
189
188
191
316
301
,259
.I17
,214
137
,385
,379
343
330
1%
n37
325
32 1
,275
,256
,236
303
320
,148
,178
,183
111
-
046
- 279
- 003
138
134
---.I87
.058
.I44
058
"2
- 051
- 027
.022
,065
-.004
TABLE 6. Group centroid values of four
Bedouin groups
Group
Function
I
Function
I1
Function
111
1 Negev males
2 Sinai males
3 Negev females
4 Sinai females
-0.35
2.15
-1.78
-0.13
1.72
-0.13
-0.43
-2.26
0.51
-0,42
-0.67
1.02
Function
I1
Variable
I
RPBA
RPLA
LPLA
RFIM
LPBC
LALC
LOAC
LFIM
LSRA
HPCC
RALC
RALA
LPSQ
ROAC
LOAA
236
229
072
030
123
112
- 054
-.049
212
141
146
110
099
.049
-.067
228
222
22 1
,216
,207
207
198
-
111
,028
.om
.I66
..
164
:?f;
<
074
- ($13
- 119
- 114
lJLi
186
185
,178
,172
.I70
075
052
,169
,037
-.069
----,006
--.I52
.010
for Function 11, Sinai females and Negev
males are extreme and for Function I11 Sinai
and Negev
Classification of the skulls according to
sex and tribal affiliation (Table 7) shows all
groups to have approximately similar numbers of misclassifications, ranging from 7%
to 17%. The average percentage of correct
TABLE 7. Classification of four Bedouin groups into their tribal and sex affiliations'
_ _Actual
_ _ _ _group
_~~
1 Xegev i d e s
No. of
cases
_.
~-~
1
41
34
(s2.9%1,)
2 Sinai males
39
3 Negev females
38
4 Sinai females
21
2
(5.1%))
3
(7.9%)
0
(0%)
-.
Predicted
group
____..._
--membership
3
2 _____
2
3
(4.9%)
36
(92.3%)
0
(owl)
1
(4.8%))
(12.2%)
2
(2.6%)
33
._____.. -.
4
0
(0%)
0
(0%)
2
(5.3%)
18
(85.7%)
(86.8%)
2
(9.5%)
'Percent of grouped cases correctly classified = 87.058.
TABLE 8. Classification.of skulls of two Bedouin
groups into their tribal affiliation, males only'
Actual group
No. of
cases
1 Negev males
41
2 Sinai males
40
_
TABLE 9. Classification of skulls of two Bedouin
groups into their tribal affiliation,females only'
Predicted group
_membership
_ ~
____
1
2
Actual group
._______
38
(92.7%)
1
(2.5%)
'Percent of grouped cases correctly classified = 95.06%
3
1 Negev females
(73%)
2 Sinai females
39
(97.5%)
No.of
cases
47
Predicted group
membership
1
2
42
(89.4%)
22
I
(4.%)
'Percent of grouped cases correctly classified
91.03%
5
(t0.6'%)
21
(95.5%)
SEPARATING HUMAN POPULATIONS
classification is very high, being 87.05%. Of
the total 17.1% misclassified Negev male
skulls, most (12.2%)were erroneously attributed to the Negev female group, and only
4.9% were misclassified as Sinai males. The
Negev females present a similar picture,
with most of the misclassified skulls attributed to the Negev male group. As for the
Sinai males, approximately half of the 13.2%
misclassified were assigned to the Negev
males group (7.9%) and half to the Sinai
females (5.3%).Most of the 14.3% misclassified Sinai females were attributed to the
Simi males (9.5%> and the zest (4.8%) LO tile
Negev females.
The predicted tribal memberships for
male and female samples separately (Tables
8 and 9, respectively) indicate a very high
level of correct classification for both groups,
with that for males being somewhat higher
than for females. As can be seen in Table 3,
prediction of group affiliation, by sex, is generally based on different measures. Most
measures that highly discriminate between
the Negev and Sinai females pertain to the
base of the skull and the calvarium, whereas
those discriminating between the Negev and
Sinai male skulls are from all three regions.
When using a higher criterion for variable
selection in discriminant analysis (with minimum F ratio to enter into the function equal
to 41, correct classification of skulls according to sex and tribal affiliation (with only
four variables in the function) is greatly
reduced (53.1%).However, it is still greater
than that achieved by the five conventional
variables used by Arensburg et al. (1987) of
$74, When sex is determined first, successful classification into tribal origin (with only
four to six variables in the function) remains
relatively high: 82.4% for males and 76.5%
for females. This clearly indicates the discriminatory power of our sets of bilateral
measures.
In order to evaluate properly the validity
of the classification results obtained for a
given number of individuals (266) and given
number of descriptors (661, we determined
the percentage of correct classifications due
merely to chance.
In anthropology, discriminant functions
are usually generated and used without reference to the level of chance classification;
random results are assigned the value of
50% correct classification (in the case of two
groups), and any results greater than 50%
are usually considered to be due to the infor-
315
mation contained within the descriptors
(measurements). However, for a given number of individuals, the probability of fortuitously obtaining 100%correct classification
increases as the number of features (d) increases from 1to the number of individuals
in the study (N) (Stouch and Jurs 1985a,b).
These classifications, although correct, are
due only to artifacts of mathematics governing the process of generating linear discriminant functions (LDF). They are not due to
any relationship between the individuals,
and the resulting LDF will have no predictive ability beyond random guessing (Stouch
and Jurs, 1985a,b).
Until recently, it was accepted that if the
number of descriptors is kept below one third
the number of individuals used, the probability of complete separation due to chance
could be kept low. The ratio of NIda.3 was
accepted as a minimum requirement (Stuper
and Jurs, 1976; Varmuza, 1980). More recent studies (Stouch and Jurs, 1985a,b),
however, have shown that, a t that ratio,
random classification results ranged around
90%. Even one descriptor for every ten individuals would yield random correct classification of about 75%. It was also found that
unequal group sizes serves to increase correct random classification and that for any
one value of the ratio N/d, the percentage
correctly classified is the same regardless of
the number of individuals used in the study.
The N/d ratio in our study was four prior t o
the stepwise procedure, after which the ratio
increased to six (in certain runs). Because a
posteriori reduction in the number Q€ variables (i.e.. using only those vtlriabies thet
work best) does not solve this particular
problem, the best way to show that our discriminant functions are reliable discriminant tools, despite the low N/d ratio, is by
showing that they will work equally on an
independent but otherwise similar sample.
Unfortunately, there are no such samples
available in our country on which to perform
complementary studies. Nevertheless, we
strongly believe, for the following reasons,
that the proposed method has great potential.
1. The theoretic background (relationship
between level and pattern of asymmetry and
population genetic structure) of the proposed
method is well established in the literature.
2. An important problem in pattern recognition is the selection of the features used to
316
I. HERSHKOVITZ ET AL.
characterize an object or, in this case, the
skull. In the first and most important step,
anthropologic and genetic knowledge must
be, and was, used to select those measurements that hopefully are related t o the classification problem. After the preliminary
preselection of features, we further selected,
using the mathematical method, for the best
features for a given classification problem
(see F = l , F=4 method). To verify that our
traits properly serve the goals for which they
were intended (i.e., discrimination between
closely related tribes), we used these same
traits and same samples in a discrimination
for which they were not designated and that
lacks the justification of any practical or
theoretic genetic background, i.e., discrimination between two arbitrary age groups.
Because we determined the age of each skull,
we split the total sample into two age groups
(below and above 30 years of age) and used
the same descriptors for age classification.
The stepwise procedure (F=4) found very
few variables (d=2) that could serve such a
purpose, and the average classification success was, of course, much lower than that
obtained in our study for tribal classification
under the same conditions (82.4%vs. 66.1%).
Moreover, when we used 60 asymmetry parameters (30 of fluctuating asymmetry and
30 of directional asymmetry), they still
showed a high level of success (77.9%) in
discriminating the skulls by tribes, although
the results were somewhat lower than obtained using metric traits.
3. Although both parametric and nonparametric methods are used for the developmeat of discriminants (asing different statistical parameters), the N/d requirement
value in the literature was obtained from a
nonparametric method; ours is parametric.
4. Skulls that were misclassified by the
functions were categorized generally within
their own ethnic group (314 cases; see
Table 7); i.e., of the 17.1%wrongly classified
Negev male skulls, 12.2% were classified
into the Negev females class, and only 4.5%
into the Sinai males group.
5. The minimum requirement of variableto-pattern ratio (N/d) is 3 (based on Varmuza’s 1980 textbook, on which all later studies
are based, and on the study of Stuper and
Jurs (1976),and our ratio is 4. Recent studies
(Stouch and Jurs, 1985a; Jurs, 1986)propose
to increase the ratio to 10. In physical anthropology studies, such ratios are rarely
achieved, as the skeletal samples are almost
always too small. Surely this does not mean
that we must stop using discriminate functions, but that we must be careful in evaluating the results.
In addition to all of the above, the “Monte
Carlo” techniques, which is a simulation
method employed by Stouch and Jurs
(1985a,b) to establish reliable criteria, were
used to simulate a study similar to that
presented here, involving 266 individuals
and 66 descriptors. This study was formulated to determine i f there vcrere substsntia!
changes in classifications provided by “real
data” as opposed to those of “idealized data.”
Six different runs were performed. Parametric discriminant functions were used under
all circumstances (F=l or 4; class = 2 or 4).
The results obtained by the simulation runs
were much lower compared with those with
“real data”: for two-group classifications
(when F=l), 95.1% vs. 70.9%; and for fourgroup classifications, 87.0% vs. 50.7%.
In sum, the 87.4% successful classifications in our study could well be real and
based on the actual differential asymmetry
level and not due merely to chance. Furthermore, the possibility of high random correct
classifications does not preclude the possibility of actual clustering of the skulls. The
difference of more than 20% between “real”
and idealized data provides evidence for a
true tribal-asymmetry relationship, which
may not be covered by appropriate random
variables that would contain no information
pertaining to biological asymmetry.
Any numerical technique has limitation3.
JVs believe that this should not discourage
us, but rather oblige us to use them carefully.
DISCUSSION
Relatively little research has been carried
out on the craniofacial characteristics of the
Negev and South Sinai Bedouins, most of it
being on the population level (Arensburg,
1973; Henke and Disi, 1981; Hershkovitz
et al., 1983; Arensburg et al., 1987; Brown
and Smith, 1988). The Bedouin skull has
been described as being generally dolichocranic in males and mesocranic in females;
facial height was usually low in males but
medium in females. All studies mention
marked sexual dimorphism in cranial size
and shape. Anthropological studies on skulls
as well as on living subjects revealed only
small morphological differences among the
tribes. Hence it was generally concluded that
317
SEPARATING H L W POPULATIONS
the Bedouins of southern Israel, Jordan, and
the Sinai, Egypt, are a relatively homogeneous population (Vallois, 1937; Arensburg,
1973; Milanesi and Petrucci, 1980; Hershkovitz et al., 1983). This explains why predictions of correct group affiliation based on
conventional morphological measures yield
poor results (Arensburg et al., 1987) compared with those obtained for human populations that are remote in time and space
(Howells, 1973; Falk and Corruccini, 1981).
The paired craniometric traits used in the
present study enhanced our ability to identify Bedouin skulls by tribe-sex aftiliation,
almost to the level attained by Howells
(1973) for different ethnic populations
(93.2% correctly classified males and 91.9%
females). Arensburg et al., (19871, using several conventional measures (e.g., nasal
height, cranial length, minimum frontal
breadth, nasal breadth, cranial breadth) in
discriminant analysis of the same Bedouin
populations, could successfully identify by
tribe and sex only 47.4%of skulls, as opposed
to 87% in the present study. The isolation of
and the ongoing inbreeding process within
the Bedouin tribes, accompanied from time
to time by a sharp decline in population size
(as a result of epidemics, drought, etc.), produce different outcomes of the genetic-environmental interaction effect. As a result, the
pattern and level of asymmetry will differ
from tribe to tribe.
The importance of asymmetry traits in
discriminating skulls of closely related populations can be appreciated from the following findings: 1) the small contribution of
facia! measures to the discrhinatinn hetween Bedouin skulls, with paired facial
measures generally displaying the closest
associations (probably because of their high
heritability) and hence manifesting the least
asymmetry (as already shown by Woo, 1931);
2) the reduced contribution of paired bilateral measures toward the discrimination between female Bedouin skulls of different
tribes-only one pair of bilateral measures
was selected here, as opposed to four for male
skull discrimination and nine for the prediction of both sex and tribal affiliation; assumedly, females are better buffered against
environmental stress (Hall and Macknair,
1972; Relethford et al., 1980; Stinson, 1985;
Rudan et al., 1986) and therefore tend to
manifest less asymmetry than males (Stinson, 1985); 3) the large number of paired
measures selected by the stepwise procedure
when both sex and tribal affiliation are predicted.
Undoubtedly, the success of the measures
used here should be partly ascribed t o the
fact that half of these measures do not span
more than one bone, which is highly advantageous from both a genetic and a functional
standpoint. Kraus and coworkers (19591, in
an article entitled “Heredity and the Craniofacial Complex,’’ claimed that “. . . the singlest type of traits-morphologic aspect of
single bones-is the best indicator of the
control of hereditary factors in the craniofacia1 complex.” hisewhere, 1Moss and Young
(1960)have argued convincingly that each of
the cranial bones responds to different functional demands, and therefore measurements like cranial length and width include
several functionally independent parts of the
skull. More recently, Key and Jantz (1981)
maintained that the cranium consists of a
number of independently varying units,
which may respond in unique ways to a
variety of external forces. From all of the
above, we may conclude that generally the
most commonly used conventional measures
(e.g., cranial length and breadth, head circumference, facial breadth) are sums of several inde endent variables and should
therefore ehave as any other com osite
trait; that is, their variance shou d be
greater than that of each of their components
separately (for a detailed explanation, see
Kobyliansky and Micle, 1986).Moreover, because the variation in size of each bone depends on intrinsic (genetic) and extrinsic
(environmental) factors, biological isolation
and the ensuing ~ ~ n ~ a nin ~Redouip
i ~ i t ~
tribes may also contribute to the variability
of conventional composite traits. Hence, not
surprisingly, the discriminatory power of
most conventional measures that span several bones decreases when the populations
are closely related. Falk and Corruccini
(1981) claimed that the overall external
shape of the skull is more important in evolutionary terms than its constituent parts.
Their findings, however, do not necessarily
contradict our own because their measures
on isolated bones were confined to a small
part of the skull (basioccipital area) and
because they compared human populations
from very distantly related racial stocks.
The face, generally considered to be the
most discriminatory part of the skull-on an
individual, ethnic, or racial level-is relatively similar among the studied Bedouin
g
P
318
I. HERSHKOVITZ ET AL
groups (Arensburg, 1973; Henke and Disi,
1981). This similarity may derive from the
fact that many facial traits manifest a high
heritability level (see, for example, Sharrna,
1987). Consequently, they will also be less
prone to developmental disturbances and as
a result will manifest a lower level of asymmetry. It is this characteristic that renders
craniofacial traits less useful in discriminating Bedouin populations.
phy-Part 11. Tel Aviv, Israel: Eretz, Ministry of Defence, pp. 855-859.
Bailey C (1980) The Negev in the nineteenth century:
Reconstructin history from Bedouin oral traditions.
Asian African%tud 14:35-80.
Ben-David J (1978) The Bedouin Tribes in Southern
Sinai (in Hebrew). Jerusalem: Keshet.
Brown T and Smith P (1988) Craniofacial morphology of
two skeletal populations from Israel. Hum. Biol.
60:55-68.
Cheverud JM (1982) Phenotypic, genetic and environmental morphological integration in the cranium. Evolution 36:499-516.
SUMMARY AND CONCLUSIONS
Falk D and Corruccini R (1981) Efficiency, of Franial
versus dzntal mea
nto h i - s ~ p a r a ~ i nnuman
g
!a SUZ, S E C C ~ Ssh!!
S F ~disjiriininaiiitii vf
populations. Am. J . Phys. Anthropol. 57:123-127.
closely related human groups cannot be per- Hall
RL and Macknair PL (1972) Multivariate analysis
formed by gross observation alone, but it
of anthropometric data and classification of British
Columhian natives. Am. J. Phys. Anthropol. 37:401necessitates more sophisticated statistical
410.
tools, through which variation in the size
W and Disi AM (1981) The craniology of recent
and shape of individual bones of the calvar- Henke
Jordanian Bedouins. JMAA 1.217-248.
ium and base of the skull, as well as in the Hershkovitz I (1985) The Effect of Mating Type on
level of asymmetry, can be detected.
Growth and Development of Children in South Sinai
Bedouin Isolates. Ph.D. Thesis. Tel Aviv, Israel: Tel
Paired measures have great advantage
Aviv University.
over conventional ones in discriminating
I, Kohyliansky E, and Arenshurg B (1983)
crania of closely related populations, basi- Hershkovitz
Migration and biological isolation of human populacally because they are better measures of the
tions in influencing range and variation of metric and
genetic-environmental interaction effect.
non-metric traits of the skull. Antropos 10:344-356.
Besides information on size and shape of Howells WW (1969) The use of multivariate techniques
in the study of skeletal populations. Am. J. Phys.
skulls of the different Bedouin tribes, they
Anthropol. 31:311-315.
also provide data on an additional selection Howells
WW (1973) Cranial variation in man. Papers of
criterion, “developmental noise,” which has
Peahody Museum 67:l-269.
never been used previously.
Jurs PC (1986) Pattern recognition used to investigate
multivariate data in analytical chemistry. Science
Measures taken on the calvarium and base
235: 1219-1224.
of the skull are more effective in skull disP and Jantz RL (1981) A multivariate analysis of
crimination than are those of the face. Dis- Key
temporal changes in Arikara craniometrics: A methodcriminant functions, which contain mainly
ological approach. Am. J. Phys. Anthropol. 55:247conventional measures of the face, will a
259.
IOeser YA, Groeneveld HT, and Preston CB (29%) Flucpriori manifest low discriminatory power.
tuating dental asymmetry as a ineasure of odontogrnic
Because discriminant functions must be
canalization in man. Am. J. Phys. Anthropol. 71:437developed independently for each popula444.
tion, we suggest that the measures used Kohyliansky E and Livshits G (1983) Relationship beshould be bilateral in nature and should
tween level of biochemical and morphologicalvariability in human populations. Ann. Hum. Genet. 47:215cover all three anatomic areas of the cra223.
nium, with preference to the calvarium and
Kohyliansky E and Livshits G (1985) Differential fertilbase of the skull.
ity and morphological constitution of spouses. Z. MorIt is the relationship between population
phol. Anthropol. 76:95-105.
structure and correlations observed between Kohyliansky E and Micle S (1986) Finger pattern characteristics as determinants of total ridge count varifluctuating asymmetry and heterozygosity
ability. Intern. J . Anthropol. 1t49-58.
that probably enables the successful disBS, Wise WJ, and Frei RH (1959) Heredity and
crimination of Bedouins into their tribal af- Kraus
the craniofacial complex. Am. J. Orthodont. 45:172filiations.
217.
Leamy L (1986)Directional selection and developmental
LITERATURE CITED
stability: Evidence from fluctuating asymmetry of
morphometric characters in rats. Growth 49:8-18.
Arenshurg B (1973)The People in the Land of Israel from
the Epipaleolithic to Present Times, Ph.D. Thesis, Tel Leary RF, Allendorf FW, and Knudson RL (1984) SupeAviv, Israel: Tel Aviv University.
rior developmental stability of heterozygotes of enzyme loci in Salmonid fishes. Am. Nat. 124:540-541.
Arenshurg B, Kohyliansky E, and Hershkovitz I (1987)
The effect of isolation and endogamy on mor hological Livshits G and Kohyliansky E (1984) Comparative analand metrical traits amon the Bedouin orsouthern
ysis of morphological traits in biochemically homozySinai. In G Gvirtzman, A Shmueli, Y Gradus, I Beitgous and heterozygous individuals from a single popArieh, and M Har-El (eds.): Sinai. Human Geograulation. J . Hum. Evol. 13,161-171.
SEPARATING H L W POPULATIONS
Livshits G and Kobyliansky E (1987) Dermatoglyphic
traits as possible markers of developmental processes
in humans. Am. J . Med. Genet. 26~111-122.
Livshits G and Kobyliansky E (1989) Study of genetic
variance in the fluctuating asymmetry of anthropometrical traits. Ann. Hum. Biol. 16:121-129.
Marx E (1974) Bedouin of the Negev (in Hebrew). Tel
Aviv, Israel: Reshafim.
Martin R and Saller K (1957) Lehrbuch der Anthropologie. Stuttgart: Gustav Fisher Verlag.
Milanesi Q and Petrucci S (1980)Aspetto fisico e misure
antro ometriche di un gruppo di Beduini nell'ambiente lesertico della Transgiordania Meridionale. Archivio per YAntropologia 110~267-276.
Moss ML and Young RW (1960)A functional approach to
craniology. Am. J . Phys. Anthropol. 18~281-292.
Muhsam HV (1966)Bedouin or the Negev: Eight Demographic Studies. Jerusalem: Academic Press.
Nakata M, Yu P, Davis B, and Nance WE (1974) Genetic
determinants of craniofacial morphology: A twin
study. Ann. Hum. Genet. 37~431-443.
Nie NH, Hull CH, Jenkins JG, Steinbrenner K and Bent
DH (1975) SPSS-Statistical Package for the Social
Sciences. New York: McGraw-Hill.
Palmer AR and Strobeck C (1986) Fluctuating asymmetry: Measurements analysis patterns. Annu. Rev.
Ecol. Syst. 17:391-421.
Relethford JH (1988) Effects of English admixture and
geographic distance on anthropometric variation and
genetic structure in 18th-century Ireland. Am. J.
Phys. Anthropol. 76~111-124.
Relethford JH, Lee FC, and Crawford MH (1980) Population structure and anthropometric variation in rural
western Ireland: mi ation and biological differentiation. Ann. Hum. Bior7:411-428.
Rudan P, Roberts DF, Janicijevic B, SmolejN, Szirovicza
L, and Kastelan A (1986) Anthropometry and the
biological structure of the Hvar population. Am. J.
Phys. Anthropol. 70~231-240.
319
Sharma K (1987) Familial resemblance for craniofacial
traits in a Punjabi population. Z. Morphol. Anthropol.
77~157-165.
Sharon M (1977) The Bedouin and Eretz-Israel Under
Islamic government. In E Sohar (ed.): The Desert, Tel
Aviv, Israel: Reshafim, pp. 199-211.
Siege1 MI and Smookler HH (1973) Fluctuating dental
asymmetry and audiogenic stress. Growth 37:35-39.
Soule ME (1979) Heterozygosity and developmental stability: Another look. Evolution 33~396-401,
Stinson S (19851Sex differences in environmental sensitivity during growth and development. Yrbk. Phys.
Anthropol. 28:123-147.
Stouch TR and Jurs PC (1985a)Monte Carlo studies of
the classifications made by nonparametric linear discriminant functions. J. Chem. Inf. Comput. Sci. 25:45en
c)v.
Stouch TR and Jurs PC (1985b) Monte Carlo studies of
the classifications made by nonparametric linear discriminant functions. 2. Effects of nonideal data. J .
Chem. Inf. Comput. Sci. 25~92-98.
Stuper AJ and Jurs PC (1976) A computer system for
automated data analysis using pattern recognition
techniques. J. Chem. Inf. Comput. Sci. 4:99-105.
Susanne C (1977)Heritability of anthropological characters. Hum. Biol. 49:573-580.
Vallois HV (1937) Les Bedouins Taamre du Desert de
Judee. Etude anthropologique. L'Anthropologie
63:62-92.
Varmuza K (1980) Pattern Recognition in Chemistry.
New York Springer-Verlag.
Vig P and Hewitt AB (19741 Is craniofacial asymmetry
and adaptation for masticatory function an evolutionary process? Nature 248~165.
Woo TL (1931) On the asymmetry of the human skull.
Biometrika 22:324-351.
Young R (1957) Postnatal growth of the frontal and
parietal bones in white males. Am. J . Phys. Anthropol.
15:367-386.
Документ
Категория
Без категории
Просмотров
0
Размер файла
1 442 Кб
Теги
measurements, population, efficiency, separating, cranial, bilateral, human
1/--страниц
Пожаловаться на содержимое документа