Professional Documents
Culture Documents
Department of Sociology, MSC05 3080, University of New Mexico, Albuquerque, NM 87131, USA
School of Anthropology, University of Arizona, PO Box 210030, Tucson, AZ 85721, USA
c
Archaeology Southwest, 300 North Ash Alley, Tucson, AZ 85701, USA
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 3 September 2011
Received in revised form
14 December 2011
Accepted 15 December 2011
Artifact assemblages from long-inhabited sites may include ceramic types and wares from multiple time
periods, making temporal comparisons between sites difcult. This is especially problematic in macroregional data sets compiled from multiple sources with varying degrees of chronological control. We
present a method for chronological apportioning of ceramic assemblages that considers site occupation
dates, ceramic production dates, and popularity distribution curves. The chronological apportioning can
also be adjusted to take into account different population sizes during the site occupation span. Our
method is illustrated with ceramic data from late prehispanic sites in the San Pedro Valley and Tonto
Basin, Arizona, U.S.A., compiled as part of the Southwest Social Networks Project. The accuracy of the
apportioning method is evaluated by comparing apportioned assemblages with those from nearby
contemporaneous single component sites.
2011 Elsevier Ltd. All rights reserved.
Keywords:
Multi-component sites
Period-specic analysis
Ceramic wares
Popularity distributions
U.S. Southwest
1. Introduction
When artifact assemblagesdfrequency distributions of recovered
objects grouped into classes (e.g., ceramic types and wares)dare not
chronologically differentiated, assemblages associated with a longinhabited site may be attributable to multiple components. Such
mixture problems (Kohler and Blinman, 1987) are ubiquitous in
archaeology, as often there is insufcient stratigraphic control to
associate particular objects with particular periods. The problem
becomes particularly acute in macro-regional data sets compiled
from multiple sources with differing degrees of chronological
precision, and will become increasingly important as archaeologists
compile data from multiple projects to address research questions at
larger spatial scales.
We confronted this problem in our own work on the Southwest
Social Networks Project, which has compiled ceramic, obsidian, and
architectural data from hundreds of sources across the Western U.S.
dating between A.D. 1200 and 1550. Our ultimate goal is to use
these data to examine changing interaction networks. Without
some means of apportioning artifacts by time, however, assemblage mixing precludes comparisons of different sites contemporaneous occupations. Previous research on the quantitative analysis
of assemblages and time has focused on estimating site occupation
spans (Carlson, 1983; Ortman, 2003; Steponaitis and Kintigh, 1993).
* Corresponding author. Tel.: 1 505 277 3940; fax: 1 505 277 8805.
E-mail address: jmrob@unm.edu (J.M. Roberts).
0305-4403/$ e see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jas.2011.12.022
2. Related methods
The approach to estimating site occupational duration developed by Carlson (1983) and Steponaitis and Kintigh (1993) relies on
1514
1515
will differ if some wares lifespans do not cover the sites entire
inhabitation.
The at/uniform distributions assumption of constant popularity over a wares history may be less realistic than a trajectory of
rising and falling popularity. That could be represented by the
familiar curve of the standard (mean 0, variance 1) normal distribution. Carlson (1983) also used the normal distribution curve.
Note that the curves tails extend to innity in each direction, while
ware history is nite and the popularity distribution is restricted to
the years in this nite lifespan. This requires truncation of the
distribution, but the choice of truncation point affects apportioning
by changing the wares hypothesized relative popularity early and
late in its lifespan (Carlson, 1983). In applications we have followed
Carlson by truncating at 2 and 2 standard deviations from the
mean 0. Of course there are other possible distributions; Carlson
(1983) considered the chi-square distribution, Steponaitis and
Kintigh (1993) the gamma, and in our work we have examined
the beta family. As noted above, different distributions may be
more or less appropriate in different settings. But absent any
further theoretical guidance, the uniform and normal are reasonable starting points and, for our purposes here, helpful as examples
illustrating our method.
1516
Sherd
count
Ware start
date
Ware
end date
4
333
11
550
800
1200
1325
1400
1450
11
1150
1300
13
1605
252
9
48
1275
200
1275
1200
750
1400
1450
1450
1450
1300
1517
Table 2
Apportioning of ceramic assemblage at Bayless Ranch Ruin under at/uniform and truncated (2, 2) standard normal distributions, prior to demographic adjustment.
Ware
Period
1200e1250
1250e1300
1300e1350
Flat/Uniform
Cibola White Ware
Corrugated Brown Ware
Late Middle Gila Red Ware (plain & red-slipped)
Late Tucson Basin Red-on-brown Ware
Maverick Mountain Series
Plain Brown Ware
Roosevelt Red Ware
San Carlos Brown Ware
Tucson Basin Brown Ware
1.600
111.000
3.667
5.500
e
535.000
e
3.000
24.000
1.600
111.000
3.667
5.500
4.333
535.000
84.000
3.000
24.000
0.800
111.000
3.667
e
8.667
535.000
168.000
3.000
e
Total
683.767
772.100
830.133
2.115
152.690
1.604
7.513
e
658.990
e
1.313
30.870
1.397
109.718
3.992
3.487
1.896
530.189
37.146
3.266
17.130
0.488
70.592
5.404
e
11.104
415.821
214.854
4.421
e
Total
855.095
708.222
722.683
5. Evaluation
We selected several test cases to evaluate the apportioning
methods effectiveness. The logic of the evaluation is as follows.
From our work with San Pedro Valley and Tonto Basin data, we
identied pairs of sites in which the two sites, A and B, were
geographically close and likely culturally similar, and in which As
habitation history was shorter than (ideally only one 50-year
Table 3
Apportioning of ceramic assemblage at Bayless Ranch Ruin under at/uniform and
truncated (2, 2) standard normal distributions, with demographic adjustment.
Ware
Period
1200e1250
1250e1300
1300e1350
1.504
108.689
3.590
1.888
136.399
4.506
0.608
87.912
2.904
Flat/Uniform
Cibola White Ware
Corrugated Brown Ware
Late Middle Gila Red
Ware (plain & red-slipped)
Late Tucson Basin
Red-on-brown Ware
Maverick Mountain Series
Plain Brown Ware
Roosevelt Red Ware
San Carlos Brown Ware
Tucson Basin Brown Ware
e
523.864
e
2.938
21.287
5.679
657.417
110.090
3.686
26.713
7.321
423.719
141.910
2.376
e
Total
666.750
952.500
666.750
4.878
6.122
Table 4
Pairs of sites for evaluating the apportioning method.
e
512.667
e
0.972
24.529
2.600
711.258
50.896
4.171
23.471
10.400
381.075
201.104
3.857
e
Total
666.750
952.500
666.750
1.659
119.625
1.188
6.110
1.890
148.226
5.098
4.890
0.451
65.150
4.714
e
1518
Dik
.
X
Table 5
Dissimilarity Indices D from comparisons of apportioned assemblages with
observed test site assemblages.
Sites
Decorated Wares
Wright vs. Ash Terrace
Buzan vs. Lost Mound
Cline Terrace vs. U:3:128
Schoolhouse Point Complex
vs. Schoolhouse Point Mound
U:8:515 vs. Armer Gulch Ruin
U:8:515 vs. U:8:589
Plain/red/corrugated Wares
Wright vs. Ash Terrace
Big Pot vs. Lost Mound
Buzan vs. Lost Mound
Cline Terrace vs. Casa Bandolero
Cline Terrace vs. U:3:128
Schoolhouse Point Complex
vs. Schoolhouse Point Mound
U:8:515 vs. Armer Gulch Ruin
U:8:515 vs. U:8:589
Flat/Unif.
Normal
(2, 2)
Normal
(3, 3)
0.101
0.193
0.040
0.581
0.041
0.168
0.126
0.334
0.052
0.257
0.149
0.207
0.063
0.420
0.165
0.215
0.240
0.394
0.129
0.369
0.039
0.266
0.057
0.184
0.169
0.323
0.253
0.064
0.053
0.155
0.169
0.323
0.253
0.064
0.053
0.155
0.249
0.122
0.159
0.079
0.076
0.189
0.307
0.031
0.137
0.092
0.098
0.224
0.214
0.439
0.214
0.439
0.263
0.414
0.317
0.392
1519
vj0 max s0 ; wj0 ; and vj1 min s1 ; wj1 :
Then the relevant conditional distribution F* is
F*x Fx F vj0
F vj1 F vj0 ; for vj0 x vj1 :
In obtaining the apportioning probability pjt for any period t, we
allow the wares start and/or end dates to not be multiples of d. (For
convenience, the site dates are.) With
qjt0 max s0 t 1 d; wj0 ; and
qjt1 min s0 t d; wj1 ;
pjt F* qjt1 F* qjt0
i.
h
F vj1 F vj0 :
F qjt1 F qjt0
A. 2. Uniform distribution
In this case F(x) (x wj0)/(wj1 wj0), for values (years)
wj0 x wj1, so that
pjt
qjt1 qjt0
.
vj1 vj0 :
The general expressions can be modied for use with the truncated standard normal. Positive value z* indicates the (symmetric)
truncation points: z* 2 for truncation at values 2 and 2. The
midpoint and length of the wares lifespan are mj (wj0 wj1)/2,
and gj (wj1 wj0). Transforming values above leads to convenient
expressions involving V(x), the standard normals cumulative
distribution function. We transform vj0 and vj1 as
zj0 vj0 mj
gj =2z* ; and zj1 vj1 mj
gj =2z* :
A. 1. Formal description
Note that zj0 z* if vj0 wj0, and zj1 z* if vj1 wj1. qjt0 and qjt1
are similarly transformed into Z values:
zjt0
zjt1
.
gj =2z* ; and
.
gj =2z* :
qjt1 mj
qjt0 mj
F zj1 F zj0 :
pjt F zjt1 F zjt0
Table A.1 shows the necessary values and subsequent calculations for the example of Corrugated Brown Ware with the truncated
standard normal distribution that is discussed in the main text.
1520
Table A.1
Apportioning of 333 sherds of Corrugated Brown Ware (j 2) at Bayless Ranch Ruin using truncated (z* 2) standard normal curve.
Values
Period
1200e1250 (t 1)
1250e1300 (t 2)
1300e1350 (t 3)
q2t0, q2t1
z2t0, z2t1
V(z2t1) V(z2t0)
1200, 1250
0.667, 1.000
V(1.000) V(0.667)
(0.8414 0.7475) 0.0939
0.4587
333 0.4587 152.747
1250, 1300
1.000, 1.333
V(1.333) V(1.000)
(0.9088 0.8414) 0.0674
0.3293
333 0.3293 109.657
1300, 1350
1.333, 1.667
V(1.333) V(1.000) (0.9522 0.9088) 0.0434
p2t
N2 p2t
0.2120
333 0.2120 70.596
Also: s0 1200, s1 1350, d 50, T 3, w20 800, w21 1400; v20 1200, v21 1350, m2 1100, g2 600.
z20 (1200 1100)/(600/4) 0.667, z21 (1350 1100)/(600/4) 1.667.
V(z21) V(z20) V(1.667) V(0.667) (0.9522 0.7475) 0.2047.
z2t0 (q2t0 m2)/(g2/2z*), z2t1 (q2t1 m2)/(g2/2z*), p2t [V(z2t1) V(z2t0)]/[V(z21) V(z20)].
References
Brainerd, G.W., 1951. The place of chronological ordering in archaeological analysis.
American Antiquity 16, 301e313.
Carlson, D.L., 1983. Computer analysis of dated ceramics: estimating dates and
occupational ranges. Southeastern Archaeology 2, 8e20.
Clark, J.J., Lyons, P.D. (Eds.), 2012. Migrants and Mounds: Classic Period Archaeology
of the Lower San Pedro Valley. Anthropological Papers No. 45. Center for Desert
Archaeology, Tucson.
Cowgill, G., 1990. Why Pearsons r is not a good similarity coefcient for comparing
collections. American Antiquity 55, 512e521.
Deming, W.E., Stephan, F.F., 1940. On a least squares adjustment of a sampled
frequency table when the expected marginal totals are known. Annals of
Mathematical Statistics 11, 427e444.
Goetze, C.E., Mills, B.J., 1993. Ceramic chronometry. In: Mills, B.J., Goetze, C.E.,
Zedeo, M.N. (Eds.), Across the Colorado Plateau: Anthropological Studies Along
the Transwestern Pipeline Route. Ceramic Studies, Maxwell Museum of
Anthropology and Ofce of Contract Archeology, vol. XVI. University of New
Mexico, Albuquerque, pp. 87e150.
Hill, J.B., Clark, J.J., Doelle, W.H., Lyons, P.D., 2004. Prehistoric demography in the
Southwest: migration, coalescence and Hohokam population decline. American
Antiquity 69, 689e716.
Kohler, T.A., Blinman, E., 1987. Solving mixture problems in archaeology: analysis of
ceramic materials for dating and demographic reconstruction. Journal of
Anthropological Archaeology 6, 1e28.
Mills, B.J., 1989. Integrating analyses of vessels and sherds through models of
ceramic assemblage formation. World Archaeology 21, 133e147.
Ortman, S.G., 2003. Artifacts. In: Kuckelman, K.A. (Ed.), The Archaeology of Yellow
Jacket Pueblo (Site 5MT5): Excavations at a Large Community Center in
Southwestern Colorado. http://www.crowcanyon.org/yellowjacket.
Ortman, S.G., Varien, M.D., Gripp, T.L., 2007. Empirical Bayesian methods for
archaeological survey data: an application from the Mesa Verde region.
American Antiquity 72, 241e272.
Pauketat, T.R., 1989. Monitoring Mississippian homestead occupation span and
economy using ceramic refuse. American Antiquity 54, 288e310.
Ringrose, T.J., 1992. Bootstrapping and correspondence analysis in archaeology.
Journal of Archaeological Science 19, 615e629.
Robinson, W.S., 1951. A method for chronologically ordering archaeological
deposits. American Antiquity 16, 293e301.
Steponaitis, V.P., Kintigh, K.W., 1993. Estimating site occupation spans from dated
artifact types: some new approaches. In: Stoltman, J. (Ed.), Archaeology of
Eastern North America: Papers in Honor of Stephen Williams. Archaeological
Report No. 25. Mississippi Department of Archives and History, Jackson,
pp. 349e361.
Sullivan III, A.P., 2008. Ethnoarchaeological and archaeological perspectives on
ceramic vessels and annual accumulation rates of sherds. American Antiquity
73, 121e135.
Varien, M.D., Mills, B.J., 1997. Accumulations research: problems and prospects for
estimating site occupation span. Journal of Archaeological Theory and Method
4, 141e191.