You are on page 1of 47

ried (

man)
lnde!
3. Methods for Archeological patte
of da
Settlement Study
occas
yond
RdJm D. Drennan, Teng MiniO'" a,ri.~ian E. P<W7OTl, GUimn ShdLuh , (Fi~

Gregory G. Indrmnn, 7llu Yanping, Katheryn M. Linduf(, Gun Zhizi"mg,


glOn,
and Manuel A. Roman- Uuay>
tory
relial
Reg10nal survey rneth:xlology, BS it has coml' to be routinely employed in many parts of
cern~

th~ world, has not been without its critics. Archeologists hRV", woTTif'f'l aoout the extent to
n=!'
which the helter-skelter c:o!lections of artifacts ofTen made on regional survey accurately repre
,pee'
sent subsurface remains or can sustain quantitative or other analyses adequate for such pUI'JXIS
mWl~

es as chronological identifieation or detennination of site function (e. g. Tolstoy and Fish


the \
1975 ; Parsons, Kintigh and Gregg 1983: 1 - 2 ; Wandsnider and Camilli 1992 ) . Geol ,un,
ogists and geoarcheologists have worried alxlUt the exlt:llt to which sedimentation and erosion
affect the surfaces to which survey is applied. effectively ~emovjng sites from the accessible inevi
archt;'Ological record or even creating new sites through redep.::lSition of artifacts transJXlrted ies (,
from their originallocatic::ru: by nuvial Proces'>eS (e. g. Waters and Field 1986; Brookes. ",n,
Levine, and DenneU 1982). ProJXlnents of fuUooverage survey have debated advocates of relia)
sampling programs as if the two were inoompatible alternatives (Fish and Kowal~w~ki,ed~. ",d
1990) instead of JXltentially complementary strategies. Concern that archeological" sites" hold
& traditionally defined cannot adef}llatf'ly represent the actual distribution of artifacts and an toke
thropag:enic features of interest has led to calls for" siteless survey" ( Dunnell and Dancey olog;
1983 ; Dunnell 1992 ) . One major branch of" landscape archeolcgy " seeking to replace an ulati,
outrrKxled "settlement pattern" approach springs from a similar concern (e. g. Rossignol and ",d
Wandsnider.eds. 1992), while another represents a reaction against the "new archeology" beeB
asscx::iatiun WId mdkrialist slant of "settlement pattern" studies (e. g, Ashmot and Knapp, the 1

"j" 1999) .
A number of these concerns have been expressed repeatedly in the regional settlement othe
pattern literature over the years. Nevertheless. much of the largest-scale regional sun.'ey ccl' "",h
. 122 .
ried out during the past 30 years has put little effort into methodologiml improvement, as
many of u"'" seem to have bceome increasingly accustomed to simply ignoring such concerns.
Indeed, the im:roducrion to one major reeent retrospective collecTion of papers on settlement
pattern studies ends with a very similar lament aoout "cursory attention to the fundamentals
of data eollection" (Bdlman 1999:5), While amcern aoout the quality of data mllection is

occasionally expressed in that volume, little suggestion is offcred toward improvement be


yond, for example, an admonition to have "experienced, motivated, curious field observers"
(Finsten and Kowalewski 1999: 35).
The parlicipants in the CICARP have taken regional settlement study in the Chifeng re
gion, not only as an opportunity to learn more aoout a very interesting developmental trajec
tory of complex society [rom aoout 6000 BeE onward, but also as a challenge to provide more
reliable regional-scale analysis of this sequence by addressing s:Jme of the long-5tanding am
Is of
cems aoout dara collection in regional survey. As is the case in all regional surveys, we have
It to

.
needed to develop ways of collecting and analyzing data that are practical and effective in the
'pre specific conditions the Chifeng region presents us with (d. Sanders 1999: 13), but since
many of those conditions can also be found in other regions elsewhere in China and the rest of
Fish the world, we hope the appTOllches we explore may contribute to an advancement of regional
ieol survey techniques in a broader context as well.
)Sian It seeIIl5 to us unarguable that efforts to collect comprehensive data on a large scale must
;sible inevitably sacrifice s:Jme of the detail and reliability that can be achieved in smaller scale stud
)rted ies (d. Kowalewski 1990). This is not unique 10 archoology-for precisely the same rea
lkes, son, the information collected aoout anyone household in a national census is less detailed and
es of reliable than could be collected by intensive observation of that household. The less detailed
,eds. and reliable data can, however, be collected systematically for a very large number of house
,ta holds. and it is this comprehensiveness that gives a census data set its special utility. We abo
nd an take it for granted that regional survey is not the only kind of data rollection needed in arche
aneey ology, just as cell.'ilJSeS are not the only kinds of data needed for studies of contemporary JXlP
<ean ulatioll.'i. Research at the regional scale is an important complement to research at other scales
)1 and and vice versa. This point is not often made explicit in methodological discussions, perhaps
llogy" because it seems so obvious as to be trite. We mention it nonetheless because at least some of
napp, the more JXllemic discussion of regional-scale studies in archeology overlooks it.
The survey methcx:iology applied by the CrCARP is comparable in many ways to that of

ement other large-scale systematic regional surveys conducted in various parts of the world. Tearn.s,

'y car- each composed of aoout four archeologisl5, systematically walk back and forth across the en
. 123
tire Ilinclsc.aPf'_. spaced about 50 TIl apm. Most sites are detected as surface scatters of arti tion i
f.Kts (usually ceramics). and surface ..;sihility of artifact scatters is generally quite gOCJd foum
throughout the region-comparable, for example. to that encountered in the Basin of Mexi ,Iron
00, the Valley of Oaxaca, or highland Peru and fulivia, though perhaps not quite as gco:.l as cally
in the dry coastal valleys of Peru or in Me.."Opotamia. Surface visibility in the Chifeng region
0"""
is certainly good f'nough thRt we have not found it necessary to consider techniques such as arern
the excavation of shovel tests which have often been applied to regional surveys in North bern"
America, the nonhern Andes, and other regions. Nowhere in the Chifeng region does vege site n
tation pose challenges to regional survey even reroot.ely oomparable to those offered by the ismt
forests of both lowland and highland humid tropical zones. The natural plant oommunity of <Nern

the Chifcng region has been complecely remoVf'n hy human activity over several millennia. dem'
Today, intensive cultivation complemented by livestock raising (especially sheep and pigs)
leaves the entire landscape very bare except for crops. [gnat
Survey has been carried out during the spring and summer, between April and August, ~arc

April and May, at the end of the dry winter, present especially favorable conditions: temper tiona
atun~s arc getting warm enough for comfortable rielclwnrk; wild vegetation is very sparse; site \
and plowing and cultivation are beginning. By June, crop growth in the flat valley floors, desig
where planting occurs first. is beginning to intdere with both surface visibility and mobili moo'
ty. By July and August, maize, sunflowers, wheat. miUet, and a wide variety of other ma Ch"
turing crops make survey III the flat valley floors impossible. The later planting schedule of 2000
the upland:;, together with generally less intensive cultivation. me~m thllt surface visibility re as te
mains good into June. July, and August. High temperatures in July and August, however, covel
dramatically reduce the productivity (not to mention the romfort) of survey teams. Surface tics (
visibility seriously obscured by vegetation. then. is not one of the particular methodological and]
problems we hf-vC faced in Chifeng, except in tenns of scheduling fieldwork to avoid den,<;f' delin
crop growth. for il
The discussion of survey methooology below springs directly from our experience in ed al
Chifeng and revolves around five principal issuE'.5 which we will take up in tum: archeological archl
site; as unit~ of obSrvation, cha:racteri7..3tion of sites, sites 3.<; units of ana1ysis, environmental
processe..s, and chronology. worr
~~

3.1 Sites as Units of Observation prim

l>.rru
While it is not always considered a melhcxio1ogical issue worthy of mnch explicit atten ~hi{ti

124
ti-
P
'I rion in rep.-'!ts of re,Rional ,;urwys t:On~emed with complex ~)Cie(ies, mose such research is
00 I foundf'd on ,he notioll or C1rchl:'Ologicru ;,ites-a notion who"e milllY fI/id vil!iJilY hi:\.,'; been
I strongly qllf>stioTleo (e. g. Dunnell and Duncey 1983; Chang 1992; Dunnell 1(192). Pracli
"'
as ,I cally all regional surveys hClve iuvo lvee! ~tirre; OJITlPlicated and alrrust always largely subjective
,on orarbitmry deciSIOll5 aoout whm to call a sit~ . 11l1:: criteria by which thEse decisions ha\lf.' h'PIl. TTlo'Jde
as are r7J't'ly It'forted iil1d , we SlJ.'Spet."1 , often not eVt'(l made explicit mrong s~ field crews. perhaps

"h bocause what a si tl' is seems :oJ.) obvious to !D many arrhrok~lsts 11u: dccisiOfl ilOCu[ what II) calt a
,e ::.:
site matrers \.:.ecau...'<:' onlinarily infonnatiou i" recorded s)'Stf.'m'ltirnlly ilOOut sites, whelm5 infcITTl'ltia:
the is not likely w l:e remded at all ab:Jut locale; that are 1'\.:t de;ignated a~ site> , If evidence relevant tIl
. of O\erall research objectivES cx:LUIS on the lam~ in locations that are- not rn.lled sites and this evi
,. dcncc , W [I U)~~ , is nol num:led ther: the ability to achieve n:sean:h obj6:tives is injured .
~~) Chang (1992). for examrle. is particularly cuncerued that settlement surveys only des
ignate a~ .sill'S those localeS' though! 10 have been ~'illages or OthlOr !Jt:'IllJaUent .settlements; re
1St.

~,-
,
.~ S'8fCh on The mobill' pm;toral ways of life that iurerest her clearly requires attention
Tional kinds of evidence, but there is a more effective solution than replacing the runcepl of
[0 addi

'se; site wirh the concept or "supcrartifact" &; ClHmg :::il..gge:sts. Duta collection should pnx-eerl hy
Irs. designating as a :sile any locale WiTh saDie trace of ancient human activity, It appears that
)iJi - most tegional surveys do follow this practice. so that e'"iclence of the kind that concerns
~
na- Chang i.s in [111:1 l/lJlikdy to go unrc(-JOrteJ where it exisfs (st-;(" Persons, Ha.snr.R~, and Matos
:;' of 1.; 11
2000 for a particularly reJcvan I l'xample). Widespread evidence of inler.sive agriculture (SUC.1

<e- ~' a'> [errace~, canal~, or raised ficld~) can he more difficult to work with ~ince such fearures UlfJ
:'.
I~r cover large arcas. and obSErv8tirms rnnceming them arl' not as e&;y to record as charactl'ris
bee tic~ of a series of "sites." Some different means of organizing rhe process of clata recording

tical and management would indeed ,;eem called for in this jn.stance--ull~ [lUI organized [lr0und the
, I'
,~
d"linf'!llion or .sites-buT &'i a t':Jmplement to u:;e of the site concept rather than a replacement
for it. since nuIDy other human activities, inclnJing habttatjun, mn.y wll be spatially restrict

" tn
ed and distinu ~) [hat their evidenc!:' Gill bt dficiently and dccnrately recorded in ff'rTl1.'l of
;:ical arcneological sites.
:ntal This 11JSt ooncem, toongh, bri~ us c1a;er [0 the natlill::' of our :e<ean+ in Chifeng and 10 other
wo:Tie~ aoout tht sitt.:: Luncept. EvidcnL~ thm is widely dis[X'r-:t'l'J rlCfU;S the landscape is mt efficiently
or ilccllrately Il:UlrUeJ in term~ cf Site;. Working In the Mis.<;i::-sippi Valley, where, like Chifeng, tile
prif1cj~ arr:heok~imJ eWJence (~rvable witmut excavation oof1.':iists u( surface ~tter-; of anif&b,

Dunnell (J 9l12) and Du;mell iU1d l~n''l'Y ( 1983) have iu)..-u:atW eliminating the Sill' corK~pl ~iJld

ten- shifting tc the individual artifflc! as thi, unil of ob,'*,rv~tion and analysis. Wherher this is prnnical or
125
desirable ~ rn the natL1rl? of the arcltrokgical rerord and of the resean:h. D..mnell and Thmcey
provide an illustnltion of their rnethxblq!;)' in a lOlle where artifacts 00 app:m to be .....;ddy dis an an
persed. They fann 3JT1f.' OJIlcentrnnons, but t~ OJIlcentrnnCJ!"1S. might plausibly be defined in dif &nne
ferent ways, and a substantial p:rtion of the artifacts visible on the surface o::cur outside such <.UrlCfll been ,
tratioru;. To Sl.Jbjectively dfsignate &Ire plare; as sites and igmre everything outside tI1ale placB 100 n
would. indEHl., .92eITI to 00 L1ltrlE'l::'effi8IY viol.l"I1ce to the evidence there to be oI::sm>ed.. catiar
The ardxnlcgical rerord in many~, ~,ffirnply cb:s fDt have these charnct:ffiStiC5. reron
lt certainly 0Ces rot in QIifeng. Here evm a :relatively &nail site might have t:h:u;ands of artifact<; on "'d
its surface, and areas betweP.n sites are s:J a:.rnparntively devoid of artifacts that ~ lTI'WS can of sherd
ten v.al.k far many kilcrreters with:JUt mng a single artifact. Recording the location of each artifact found
on the surface individually would be wlulJy impractical for the Chifeng region and for many other t'e is ret)

gioru; when- u:mplex meties developed. It would be such a time-con..'UITIing task that it would rnt
be feasible to survey an area large en::JUgh to a:mprehend the :regional-salle pherrmena WE' wish to
study. Dunnell's 09(2) example regicn .....'<15 only ab::ut 0.08 krrt cum~ lD a projected survey
""""
reme
region of 3JT1f.' 1130 krrt in QIifeng. Of greate<f relfi.'ll<lfiCe, 00wever, artifoct<; and the activitie; they Unde
reflect are fDt widely di~ acn:6S the land-<c:ape in Qlifeng, but very strongly cuncmtrated into natio!
~le an:.heJ.l~cal sitES, and thi~ has been the ca<:(' for many millennia. Even making al one s
lowances for a tendency In undeITIlte the quantity of artifacts oumide OJl'"lm1trntiCJ!"1S., a~ ob<.:erva:i by defin
Wandsnider and C...am.illi 09(2), it remains dmr that "site" i:< wt an arbitrary ronstruct here, im high(
p.:l'Ed <JVtt a variable but true or l~ continl..lOUS I"gional distribution of surface artifacts, but mther
, "'"
a reas:mably accurnte and effective way to charncterire and manage infonnaticn al::x:ut a fairly dinm. 'goo'
tinlDUS Tt'g1C1la1 distribution of arcllEDI~cal materials. mud
In Chifeng, any locale with ancient architectural remains, dome.'itic features, graves, h",
1;llld.sca.pe mxiifiCBtion, or the like is designated as a site. Since most sites, however, are on wh",
ly surface scatters of artifacts, we face the same practical question often faced in regiona1sur cultj,
veys similar to ours: How substantial does a surface scatter of artifacts have to be in order for lng to
it to be worthwhile to designate it a ..,ite and record information alxlUt it? Most regional-scale the a
surveys concerned with complex societies, for example, seem not to have called a locality in place
which a .'lingle sherd was found a site, although few rePJrts are explicit on this PJint. Lack of tions
explicit attention to such decisions means that they are made in very different ways in differ
ent surveys. and often even in very different ways by different survey teams in a single pro very
ject. The fomler differenees do not neeessarily invalidate compari~n of the results of different such
projeet~, but they can interfere with it. The latter kind of difference can be II threat to com sent.
prehensive analysis of data from a single project, even without compari~n to others. ",d ,
. 126 .
~
'~

ocey In Chifeng, a~ ill llIoM reglons, the fimt indication that a survey tf'MI1 hR.~ encountered

"'
dif-
an archeological site i~ typically [hat one member of the te..'UT1 finds a sherd un the surface.
&JmetiIlles, of coumc, the tearn may find no other ~herds beyond that filSl one. Our rule has

""- been that, if no one finds any rrore sheIos after the team has eontinued on its course ior aoout
.!=S 100 r:l beyond thp /X);m where me fin>t sneed was found, then that sheed is discarded. the lo
cation is not eonsidered an archeological site, and, consequently, information aoout it is not
;tie>. recurded. If, however. a second sheed is found within 100 m, the team gathers in the area
150n and searches more in[f~:lsively for more ceramics cr orher artifacts. If none are found, the two
n of sherds are discarded, and no information is recorded. If, however, three or more sherds are
tifact found, the locarion is OJllSidered an archeologic.a1 site, artifacts are collected, and information
'I re- is recorded.
f fit (n the absence of much explicit discussion of this issue in regional survey reJXlrts, this

"10 seems to be a lower threshold for wnat constitutes a site rnan many regional surveys con
cerned with complex societies have applied, although perhaps comparable to that applied by
""'Y -\
they Underhill et a1. 0998:459 -460) in the Rizhao region in Shandong Province, where desig
I into nation as a site required at least one sherd that could be ickmified in the fieid as to period. In
ol one survey in the United States S:!uthwest, Plog (Plog,ed. 1978: 387) set the criterion (or
rl by defining a site at a minimum 3urf<:l.l:e Jd15ity of five artifacts per m2 . This is a substAntially
_'l
im higher surface density than OCCIl1'S at mcst of the sites we have defined in Chifeng. From such
am" a pen;peclive, the real WarTY seems not so much that imJXlrumt information may be l08t by
>Ln- ignoring a few sherds that occur singly or in pairs but rather that we may be making too
much of scatrers consisting of as few as three or four sherds by identifying them CUi sites. It
LV e5, has, for example, often worried archeologists working in China (and. other parts of the world
~ Dn- where cultivation is intensive), that smalllow-del1Sity surface scatters may be produced when
sur- cultivared fields are enriched with soil or composr I which may motain ancienr sherds pertain
,r for ing to the locatiol1S from which the fertilizer is brought. This would not, of course, increase
scale rhe amount of ancient pottery in rhe region, but some of the ancient pottery could be dis
ty in placed on the landscape, in effect creating "sites" that do not really represent ancient occupa
ck of tions in thQ!; locations.
iffet- We have recorded information on smail low-density scatters, because it does not occupy
pro- very much oj our tirr.e in the field and because systetnatic.a1ly recording information aoout
eren! such scatters is the only way to begin empirical in~tigation of what they do or do not repre
",ro- sent. In this way we hope to keep data recording separate from analysis and inteTjJretatiull,
and concentrate the attention of survey crews on recording and/or collecting what they can
127
observe rather than ~nterpreting it. In this instanc'e, Ihis Illeam recording for later analysis with
any omcentration of as many as three artifacts and maki~ a surface collection, rather than subsu
p:l5Sibly making a subjective decision in the field rhat the coucenrr;Jtion i~ not meaning{Jl and mure
thereby losing the PJS.<;ibility of studying it more fully later on'. As a m;ult we hope that in are. ,
terpretations will be more standardized acmss the survey region, irrespective of which survey tranSI
crew recorded a particular ~;te. Interpretations will also be more subject to revision based on
further analysis and additional infonnation from other SJurces. Since we have chosen to ,hen!
rerord as siteS surface scalten; of as few as three mifacts, we can pursue further analY-Sis in in th,
8ll effort to detennine to what extent they represent ancient activity locality or more recent tions
transport and redeposition of ancient artifacts. mod,
As a fit'St step in .~uch analysis, we can look at the frequency distrihutiou of the number than
of sherds per collection in a sample of 1691 separate collections. (As discus....ed more fully be cient
Jow, .. mllection units" and .. sites ~ are not the same thing, since each ceramic collection is ities
made in an area oat excee&og 1 ha. Sites larger than 1 ha, then, are represented by multiple thoUI
wJlecliun units. Our coucern at present is with single collection units yielding very few sherds pattE
surrounded by areas where no artifacts at all were found.) Some collections (58, or about mak,
3%) pnxluced no sherds at all (Figure 3. 1). &JlIle of the:JC: consist entirely of lithic arti the.<;(
fact~; others represent the locations of graves or other ancient features with no surface arti they
faets in &'l'lXiation; still others represent locatious in sites with visible arehitectura.l rema.in~ uf tion
such substance that more detailed mapping and surface collecting is warranted, but where nmru
surface artifacts are 00 sparse that we preferred uot to collect them now but to leave them iu d",,'
pla.ce: for sue:h future work. Another 55 collections yielded only oue or two sherds each. Like
rollections with no sherds, these coltectioos sometimes contain lithic artifacts or are as.."OCiated faun
with obvious ancient fetltur~, such as slab grave~. In other cases, they rep~ent an addi Iy 31
tional area ocrupied, if sparsely, adjaeent to the other collection uuits in a larger si tl', acW.
Figure 3.2 shows the locations of collections wlth no sherds, one ur two .,herd~, three (38
sherds, fOGT sherds, and five or more sherds. [f the spar~ sherd scatters represented by col ,,,-eo
lections with very few she~ds result from relatively recent transport of ancient materials, their lecti
distriblltion might well differ from that of ancient sites, since recent soil enrichment would site
not likely be distributed through the region in the same way as ancient sites. By and large, tion
huwever, the distributions of the small collections flcm~~ the region surveyed fire quire similar site
to those of thl:' larger collections that represent denser surface s::.atten;. One exception to this
mr;esjX.lndence in Figure 3.2 is produced by ~e\'erHI mllections with no sherds at all that c~n sher
be seen In the northeastem and 9Juth-central parts of the area surveyed, w~ere co!lectiom sher
, 128 '
lysis
than
f

"

with marc ~herds are uot particlllarly' abundant. These collectior.s represent hilltop sites with
5uhstc.rrtial architectural r"'maius where sherds were not collected iu the initial survey pending
and more detailed mapping and surface collecting in the future (as mentioned above). Anif!:\(;ts
t in-" are, indeed, quire sparse un tht: surface~ of these sites, but rhey c\PlIrly do not reflect recent
~'Y II
trllns{X)rt of aucient artifaets. since the !i.I"chitectural remains are unmistakable.

II

don Another discrepancy between collections with very few sherds and collections with mare
o to sherds is the scarcity of colleetions with betwet:n one and four sht:rd:s in a roughly circular area
is in in tht: nurth-central PllIt of the survey lOne, when: there are reasonable uumbers of collee
,
:cent tions with five sherds or more. This is a part of the survey <\Tea that does have a substantial
mcxfern occupation and where the practice of moving soil or compost seems no less common
obee

'be
Ii
than in other parts o( the region. This is wllsi:stent with the notion that the are.aS where an
cient she-rds are pre..<;f'llt but SCliIce represent small, sparse, or short-tenn, but aucient, activ
)n IS I
ities rather than sherds moved recently from elsewhere with !'Oil or compost. In general,
tiple though, the distributions of collections with fewer than five sherds da not shaw vt:ry different
Jenis pa.tterns from thJSe of lulle<:tiom; with five sheds or more. This means, of coun>e, that it
00"' maKes little difference to our general cnaracterizatiol15 of the distributions of sites whether
arti these small collections are includffi or not. Fer such purposes, then, the question of whether

I
they are "real" ~ites or not diminishes somewhat in imporranct:. Fur Qlh~r rca.som, this ques
arti
ns of tiun n:llIai:ls important. It is, for example. a potentially interesting clue to changes :n eco
,here nomic and political organizatiou if the uumber of small, ephemeral occupations increases or
min decreases sharply in a particular period.
LiKe There are also other ways to examine the impact of our decision that any placp where we
iated found three or mOTe an(~ient sherds withiu an area less than 1 ha would be called a site. in on
3.ddi ly 38 instances in the area surveyed up through 2000 does a collection with just thret: sherds
actually constitute an additional site, spatially separate {rom other indications of .st:ulement
trn.., (38 out of the total uf 836 site...., delineated, or less th8n 5%). These sites are wideiy scat
, 001 tered through the surveyed area, with a distribution not noticeably different from that of col
their lections with more sherd:s (Figures 3.2 and 3. 3). They represent only 0.19 km1 0f the total
muld site area of 20.27 km 1
( less than 1 %). 1'hrn these sites have little impact either on percep
ll"ge, tion uf p<:lltt:.rns of overall 3ite disrribution or on the assessment of total number or area of
milar site".
) this Altogether, 282 colleetions in the 1999 and 2000 survey areas contain fewer than five
t can sherds. This represents lb. 7% of the total of 1691 wllecttons made, but the 476 clAssified
;tiU1l5 sherds from these collections llre only I. 9 % of the tocal of 24,034 classified sherds collected .
. 129 .
I!
;'1
If the sparse sherd scatters were produced by recent transport of ancient artifacts, such trans
port might be expected to affect sherds of each period sjmilarly _ When the proportions of
~;
( ';
,,
sherds in srmtll wHeo:.:tion1i are calculated for each period, however, a fpw perioos rliffer from I

, ,
the value of 1.9% olrerall (Table 3. 1 and Figure 3.4). The biggest difierence (5.5% for
Xinglongwa) has very little statistical significance, since the Xinglongwa sample is ~ small
,I ,

(Figure 3. 4). For the same reason, the differences observed for Zhaobaogou and Xiaoheyan
are not worth much attention. At the other extreme, we can be highly ronfident of the lower
proportions of sherds in small collections for Hongshan and Lower Xiajiadian and of the high
II ,
,
er proportion of sherds in small collections ior Llao (Figure 3. 4). (Note here that ',.;hen we I
say "snall uAlection:;j'" , we mean collections with fewer than five sherd,; -'lltogether--not col ,

lections with small quantities of .'!herds from a particular period ~.gardless of the total number
of sherds collected. )
The relatively smalL, but highly significant differences for these threc periOOs are entirely
consistent with what we observe in larger collections as well. When we look at the numbers
It c

,"
of rollections with different quantities of sherds during each of these periods, for only collec
,

I
tions with five sherds or more (Figures 3.5 and 3. 6), we see somewhat differentiy ~haped
distributiOIlt;. Fur Liso there are more small collections (up to f100ut 35 sherds) cnd fewer
large collections----ronslstent with the high proportion of collections with fewer than five a
sherds ..In rontrast, for Hongshan and especially Lower Xiajiadian, there are more large col n

lectioru; and fewer small collections among the collections with fiye sherds or rnore-roIlSistent p

with the low projXJrtions of collections with fe~"er than five sherds. That is to say, periods
with unusually low prcportions of collections with fewer than five sherds are periods wilh un h
usually low proportion," of rolJections with slightly larger numbers of sherds as well (say, 5 to 1
20). These !lre clcnrly periods when the distributions of shf'rd quantities per rollection are
skE'Wed overall toward large collections (say, oyer 20 sherds). Similarly 1 the period with an ,I
unu<;ually high proportion of collections with fewer than five sherds also has an unusLJally high ,;
proportion of cojJectio~ with slightly large numbers of sherds. It is a period when the distri T

bution of sherd quantities per collection is skewed toward small collections (.say, fewer than d
20 sherds) generally. There is no reason to expect this pattern if small sparse sherd :;catters w

result from recent ttar.sport of a:tifacts. It is, however, perfectly amsistent with the notion l'
that such 5Catters mootly represent short-term and/or sp::lr~ orcllpations in situ. y'
The observation that an nnusually high projXJrtion of Liao sherds come:'i from collections T

with fewer than five sherds. rhen, i... consistent with the unusually largE' pmJ-Grtion of Liau m

cdJections that consist of not yery many more than five sherds. Similarly, the observation
130
"
ms th11t unusually low proportions of Hongshan and Lower Xiajiadian sherds come from rnllec

; of rions with fewer th11n five sherds is consistent with the unusually small prop::lrtion of rnllec
tions from th~ two periods that wnsist of not very many more than five shenfs. This mn
"'"
fa, sistency is difficult to undersmnd if small low-density sherd $CAtter9 were mostly the pmuct

nall of the displacement across the landscape of ancient ceramics by more recent agricultural prac

yon tices. (t does make sen:-e, however. if the pattern of small low-density sherd shatters reflects
a particular abundance of small, dispersed Liao occupations and a particular scarcity of such
w"
gh occupations during Hongshan and Lower Xiaiiadian.

w, Such differences in the densities of surfaa: artifacts are sometimes argued to result from

col differences in the lengths of the periods, with denser artifacts on sites of longer perieds. The
,be, observations we have ju:;t made about these three periods, however, cannot readily be at
tributed to this factor. It is true that Hongshan is probably the longest period in the se

rely quence. and this might result in fewer sparse occupations as artifacts had longer to accumulate
in particular places on the landscape. Lower Xiajiadian, however, which likewise shows a
""
1", scareity of sparse occupations, is among the shorter periods. and Liao, with its abundance of

perl sparse occupations, falls in betweeu.


Soil, and especially comJXlSt. used to enrich cultivated fields often originate in and
we'
five aroum.1 villages. Thus, if small, sparse sherd scatters are largely produced by recent move

col ment of ancient materials in such activities, we might expect small collections to have larger

tent propJrtions of Te:'ent ceramies than the larger collections that ~m indisputable indications of
,00, ancient settlement. The correlation between rollection size and proportion of recent ceramics,

un however. although it has some significance (p = O. 079) because the sample is so large (n =

SID 1633). is of negligible strength (r=0.043).


Yet another factor that might acrount lor difference> between periods if small, spane
'"
, on sherd Sl..:B.tter9 were largely produced by recent transp:Jrt of ancient artifacts would be a rela

ligh tionship between the distributions of settlements of certain periods and modem settlements.

itri- That is. if the in situ remains of settlements in II particular period tended to be especially

hon dose to modem l'Offimwlities. then the sherds of that period might more often be transp:med

t", with romp:JSt than those of other periods and, as a roru;equence. more small, sparse $CArters

tion pertaining to that peried would be created. This JXlSSibility was investigated with II GIS anal
ysis of the territory near modern communities. whose results are surmnarized in Table 3.2.

Ions This analysis considered only sites of each period whose rollections produced five sherds or
more (that is. the ones with surface material abundant enough to serve as plausible SOUIces of
~''''
tion transp:Jrted ancient artifacts).
131
If transport of ancient sherds with compo."t originating in or near mooern towns were a
major factor in pmdueing the high proportion of small, sparse scattel1l for the Lieo perioo,
then we might expect that an unusually large proportion of Liao site are.a with denser surface
I'
materials would be in or near modern towns, and this is, in fact, the care. On the same ba
I;
I,
sis, we would expect that an unusually smaJi proportion of Hongshan and Lower Xiajiadian
site area with denser surface materials would be near modern towns, since small, spar&' scat
tel1l are especially unusual for these periods. Hongshan does have a low prop:mion of site area
near modern tOW115. but Lower Xiajiadian has an even higher proportion than does Liao. Up
I ,

per Xiajiadian and Zhanguo-Han also had low proportions of small, sparse scatters, although If ,
these differences from the overall prop:mion had less significance than those already dis
cus...'led. In these twu cases, the corresp:::mdence with prop:rt"tion of site area near modem
towns is al:'O incunsistent: Upper Xiajiadian has a low prop'Jrtion of site area near modem
towns. but for Zhanguo-Han this proPJrtion is high. Precisely as these observations would
suggest, a rank-order correlation between the prop'Jrtion of sherds from small collections and
1 d



the proPJrtion of sites near modem towns has little strength or significance ('., = 0.103, P =
0.870). We do not, then, see the pattern<; we might expect to see if small sparse scattcrs
"
I,
were largely the result of moving ancient sherds around the landscape with COITlp::lSt.
"
The GIS analysis of archeologjcal sites near mooem towns w~ based on defining the ~

zone "near" modem towns with a 500 m buffer around modem town areas. Since this choice d
of distance is somewhat arbitrary, one wonders whether a different but al:'O reasonable choiC'e a
of distance might yield different results. In order to guard against this jXlSSibility, the same m

analysis was carried out twice more: once using a distance of 200 m from mooern tawns in
"
stead of 500 m, and yet again considering only site areas actually within the limits of modern
towns (Figure 3.7). The results of these two analyses are al:'O preslO'nted in Table 3.2.
'tf"
These two analyses al:'O produced. no consistent corresp:mdence between pro(XJrtions of small, m

sparse scatters and tendency of more substantial sites to be near mooern towns. The prop'Jr th
tions of sherds from small sparse collections by perioo do not correlate well with the prop'Jr Sil

tions of site.5 either in or within 200 m of mooern towru; (" = O. 205. P =0 O. 741; and r, = O.
410. P = O. 493, respectivEly). We have failed, then, to find consistent evidence of yet an
other set of plausible expectations derived from the idea that small, sparse sherd scatters were
fanned largely !hrongh modern transiXJrt of ancient materials in crnnpost.
In sum. examination of the data set itself snggEs!.." that at least :-orne of the very small , I

sherd scatters mnsisling of no more than a few sherds may well be il1 situ remains of ancient tft

activity. Their inclusion as small low-density sites does not have a major impact on the ont "
. 132
,

I wme of distributional analyses, since they are not distribnted very differently from larger

J, sites. or on the outcome of regional demographic analyses, since they repre.<;ent such small
amounts of occupation (a subject to which we will retnrn in the next chapter). Inclusion of

." such site5 in the analysis, however, does add detail to the picture that would be obtained if
only scatters producing larger numbers of sherds were ronsidered sites. As noted aoove, for
an
,t example, such small scatters are more common in some periods than in others. If some or all
of these scatter:s do represent" real" sites. then this suggests that the incidence of small
""p- ephemeral occupations or activity areas changed through time, and this seems an observation

,h worth pursuing farther with regard to changing ~ial, PJlitical, or economic organization in

,, the Chifeng region.

'm The same conclusion might or might not be reached for other regions where similar

'm doubts have arisen conceming small, low-density "sites". The methodological {:Oint is that

,ld we can do better than to jnst arbitrarily or snbjectively decide what to believe aoout such

nd small low-density scatters. We can investigate the issue empirically in any regional survey da

= ta set if only information on such small scatters is systematically recorded. To record such in

~rs
formation is not to make an a priori decisiun that such scatters ~ "rear sites; it can instead
to be a decision to systl'1Tl<ltically investigate such a possibility. Using the site concept in this

:he way as a unit of data recording has not impeded empirical consideration of the nature of the

ice distribution of archeological remains in the region flnd of what level of archeological evidence

ice can be considered low enough to be ignored without seriunsly interfering with the achieve

me ment of research objectives. Calling locations with as few as three artifacts "sites" and sys

m tematically recording them is a dfltfl collection threshold almost as low as that advocated by

,m proponents of "siteless survey". In order to acoomplish this, it is not necessary to abandon

2, the notion of site and shift entirely to the artifact as a unit of data collection. The principal u

,I, nit of data collection we have relied u{:On in Chifeng is actually intemlediate in scale between

oc the site and the anifact, and this brings us to the snbject of how data are collected within

0' sites.

0,
an 3,2 Characterization of Sites
ere
Data collection begins, then, in our survey, as in most large-scale regional survcys, unce

wi a location is identified as a site. It is customary to describe and sketch or roughly map archi
tectural remains and other anthroPJgenic features in sites. Information about setting and oth
en'
'U[ er environmental parameters is often rerorded as well, although it is increasingly practical and
133
effiC'ient to incorporate such iufonnation in map fonn through GIS analysis. Here. V1{' are be
concerned particularly with characteriz:ing sites in terms of four interrelated parameters: of
chrouolog)', spJltial extent, function (or ORtUff of utilization). and intensity of utilization. .1

Information relevant to these parameters, of COUY"!:le, can came from architectural remains and th,
other feature~ visible Ull the surface, but in many regions (including Chifeng) most sites lack "
tia:
sueh features. [n such cases, the single most important source of infonnatiou available for all i'"
four parameters is the artifacts available on the surface. Reliable information, then, alxllit the
""

int
nature and spatial distribution of artifacts within sites must be collected at a level of detail I",
corrunensurate with both the needs of regional-scale analysis and the practiC1lliti~ of compre
hensive regional-scale data collection. "
I
"t
;(s
The interrelations among the four parameters camplieate the task. ChronOlOgy is typical lee
ly assessed in regional surveys by as.."igning sites to one or more periods, and we do not pursue jI ",

other possible approaches here. But spatial extent, intensity of utilization. and function must !II ""

also be lI.s.'lt:'s<;erl in chronological terms, since all may vary through time. At a minimum. ac
then. this means a separate evaluatiou of each of these three parameters for each period repre am
sented at each ~ite. Similarly, spatial pxtent is related to intensity of utiliz.ation and function orr
since ooth may vary spatially across a site. And sites may, of course, contain multiple func an,
,
tions of varying int~l\5itie,;. it has often been telllpting for regional slirvey crews to make th,
I
overall subjective characterizations of such things as period of occupation, density of surface wi
artifacts, and site function. and to rePJrt these interprdations for a site rather than. for ex
ample, to make artifact collections in such a way as to provide more rigorous direct observa
I:
otl
tions tG base such cha.racterizations on. Subjective characterizations made in the field. howev tio
er. are likely to be largely unsupPJrtable 8S'>Crtions, and the possibilitie.s for later analysis and loe
reinterpretation aTe sharply limited. Making collections of artifacts is a:mducive to more accu di~

rate dassification and quantification because these tasks can be accomplished with greater tare
"'.

later in the laooratory. ly,


Although ~ysteTnfltic surface coll~tion methods have a considerable history in archeolo ,. '1 I"
gy, large-scale regional surveys still often follow the practice established in the Basin o[ Mexi mE

lD, where making systcmn.tic artifact collections or even helter-skelter collections from all loe
sites was deemed too time-consuming and impractical (Sanders, Parsons and Sandey 1979: th,
20 - 3U). The helter-skelter mlllXtiurl.'> made at only Xlme sites in the Ikc;in of Mexico have oq
been regarded as inadequate for quantitative analysis, 00 characterization of chronology, spa en
tial extent, intensity of utilization and functi,:m is based on subjective visual 8ssc..'I::!mcnts made
in the field. Greater reliability and preci:-;ion for all focr of the parameters considered here can col
134
e be achieved when artifac[ oollections are made that can sustain quantitative analysis at a scale
01 spatial resolution smalier than that of the individual site.
The methodology we have followed to accomplish this in the Chifeng sUlVey addresses
d the issue; of time and practicality in the fjeld. sample size, sampling bias, and spatial resolu
,k tion. Spatial resolution is based on collection units not exceeding 1 ha. For sites of 1 ha or
,ll smaller, this means that a single artifact collection is made. Sites larger than 1 ha are divided
into subunits {or oollection, Thus each IlrliIac! oollection represents a defined area of I ha or
"
ril less in the field. The ooundarips of this area are marked on photographic enlargements of
e- satellite images carried by each survey team in the field so that each area can be located and
its area measured with some precision. A site, then, ronsists of one or more rontiguous col
J lection units whose areal extent indudes the entire distribution of features and/or artifacts
visible on the surface_ This makes it PJSSible to calculate the area of each site by :mmming the
"'
,t areas of the collection units that make it up. More important, it make.,,; it PJSSible to calculate
,. a different occupied area for each slte during each period, since after the ceramics have been
e- analyzed, those rollection units that do not contain ceramics from a particular period can be
on omitted from the area calculation for that period. Similarly, different intensities of utilization
>C and different functions that it may be PJSSible to identify for different periods on the basis of
ke the artifacts present, are assigned not to entire sites but to individual areas of 1 ha or less
within sites.
""
x- Utilization of such collection units makes it p:J..'lSible to deal with the spatial aspects of the
<a other parameters in tenns oE areas measured to the nearest hectare. Our focus on these collec
~v tion units parallels practice in the Basin of Mexico as well, where surface distributions of arti
.nd facts finally were characterized "field by field" because some large areas of rontinllOlJS artifact
,lj distributions defied easy definition of site ooundaries in the field. Site limits were defined lat

"e er, subjectively, based on the infornLation recorded on air photographs. and results were ana
lyzed and reported in teIlIlS of the sites so defined. In Chifeng, we have carried this approach
,10 farther by standardizing the areas of rollection units at approxinLately 1 ha (which mayor
>0 may not correspond to fields or other mcx:lem feattl!e:5 of the landscape) and collecting arti
ill facts separately in all rollt.'Clion units. As disc~ more fully below, the roIJection unit, nol

/9 ; the site, beromes the important basic unit of analysis, and we can move directly from data
eve organized by rollection unit to definition of communities at various scales, largely without ref
p' erence to the site as a unit of data recording or analysis.
,de With spatial units of approximately 1 ha, then. the next issue is how to make artifact
collections in such a way that they can sustain quantitative analysis. There are two kinds of
"'"
135 .
quantitative information that it is particularly important to deriv~ from these collectiorus: pro
portions of artifacts of various kinds (e. g. proportions of sherds of a particular perioo, or
.
Ii
I

~I
L

.., ~

to
proportions of lithic t:eols of >llmlP p~rticlilar kind. or proportions of sherds of !'OUle particular to I
2
vessel form, etc.) and densities of surface artifact:s (i. e. the number of artifacts pt:r m on the
the surface across the area of the rollection unit). If a collection of artifacts is to be taken to ten
repre'lf:nt a collection unit in these regards. then both 5Mnple size and sampling bias are con
cerns to be addressed. Sample "representativeness' m;ght seem the relevc.nt issue, but this
l "'''
aeit
characteristic of samples cannot be assessed directly-uuly uy attempting to increa.-<lC Sllmple
!
'" ,
:size and to reduce sampling bias can we increase the probability that the samples of artifacts hoi
collected will aCCllrHt",ly repre;;ent the collection Llnits from which they come (Drennan fae
1996b,79-97) iI ,h,
Iu Chifeng we have set 20 sherds as the minimum sample size we always Sf'ek to achieve 1. IS E

~
in a collection Llnit. This number, like much in field archeology, is a compn:nnise betweeu
'Y'
theoretically dtrived desires and n:aJ-worid practicalities. A sample of 20 sherds, for exam pet
ple, makes it PJSSible to estimate the proportions uf sherJs of different ~riod..<; (or sherds of ; I~

different vessel forms, etc.) with error ranges of ~e 10% at about the 66% confidence
level (i, e_ onp stflnrlllrn error). If ceramics of a patticlllar periex:! are absent altogether from a
I
I, "'"
10k
sample of 20 merds, we can be similarly oonfident that they comprise le'iS than 5% of the m
rupulation of sherds these 20 are a sample from (Drer.nan 1996b:255 -259). Sample size
I
criteria are stated in terms of ceramics because these are the trrlSt common artifacts on the
i, pee
bi~
surface; all cl~"''"es of artifacts are, however. collected. A larger sample would. of course, fon
enable us to characterize the anifacts in the wllf:'ction uuit with grcotcr precision and/or high 1=1
er ronfidence, however. we often have no choice but to work with even smaller samples I call
whpn HS TT1f\ny as 20 sherds simply cannot be found or. the surface. and it is not infrequently pro
the case that, by the time we reach aoout 20, we are at the point of diminishing returns on suit
our effort to increase sample size. &nnetimes our samples are larger than this, and, in these
cases, they produce relults we can be somewhat more confident of. We do not, however, or 2U!
dinarily invest much energy in collecting very large samples even when the opportunity pre roy
sents itself, since the extra effort ur collf:'cting them (not to mention carrying them UQ1111rl all we'
day) does not seem warranted. Only rarely do we coHee! more than 50 sherds in one unit. roll
,~mpling bias is a much more complicated is-'me thau sample size beamse there are many roll
sources of bias operating at different scale.~. "Systematic" collecting has c'Orne to refer in hev
archeology 10 the practice of carefully collecting all arjfacts (or all artifact:< of certaiu classes plat
or with certain characteristics) within a dearly delineated area. Such i:I procedure reduces
H"
. 136 .
,ro sampling bi~ because survey crews exercise no judgement aoout what to oollecr and what not
tu collect, and becauSE arrifacts whc>s( characteristics make them inconspicuous are less likely
'"
,1M [0 be overlooked when the ground surface in a small area is l'xamined very carefully. Al
. on though :lUch techniques have been used in many oootexts. large-&:ale regional surveys are of
,to ten carried out without systematic mIJection because ehe practice is thought to be too time
on- consuming to be practical on this scale. In Chifeng, however, we have found making system
this atic collecdons eminently practical. Two members of a survey crew can mark out a circle 3 m
lple in diameter very quickly. (One stands still holding one end of a 1.5 m rope while the other
,," holds the other end and walks around in a circle making b::ot marks on the ground. ) All arti
nan facts within the circle are collected (it usually takes less than ten minutes). If fewer than 20
sherds are found. then additional adjacent circles are collected until the minimum sample size
<eve is achieved. and the total number of circles serves as a record of the area within which the

.
'cen
~.

"i
systematic oollection was made so that the average number of sherds (or artifacts of any kind)
per m2 can be calculated .
This kind of systematic collection procedure. however. introduces another potential
nee sampling problem because the artifacts found on the 5urfal.:e in a very small defined area are
m" caken to represent those in a much larger area (up to 1 ha in our case). Field crews can. for
th, example. subamsciously choose to place systematic oollections in spots where artifacts are e.s
Size peciaIly dense or where partkularly unlL'o;ual artifacts are noticed. We attempt to avoid such
th, biases by selecting the precise spot for a systematic oollection within a 1 ha collection unit be
-,;e, fore we get dose enough to be able to observe exactly what the surface distribution of artifacts
gh looks like there. While this can prevent subconscious selection of unusual (that is. systemati
pies cally unrepre.sentative) locations for systematic collections. there is still the concern that the
ntly proJXlrtions of artifacts may vary widely across the surface of a oollection unit causing the re
; on sults of a systematic oolleetion to depend on whl'Te, exactly, it is placed.

""~ This plSSibility was inve.stigated through more intensive suriaee oolJection of one site in
oc 2001 prior to small-scale stratignlphic tesTing. The surface distribution of artifacts at site 674
Jre mvers about 3.3 ha, so on regional survey u.sing the methooolc:gy described above. this sire
I all would be divided into at least three oolleetion units. For ilIu.stration, it is divided into five
t. collection units according to the topography (Figure 3.8). Prior to testing, 24 systematic
=y oollections were made at the si,e, so we can investigate just how systematic oollections might
, on have repre:o;ented their [X)(Jection units differently. depending on exactly where they were
placed. The ceramics of three peri<:Xb (Lower Xiajiadian. Upper Xiajiadian. and Zhanguo
""
"" Han) were pre.<;ent on the surface of the site in varying prop:m.ions in the 24 systematic col
. 137
lections. as illustrated in Figure 3. 9. The collections in the south-centrol collection unit m
(Figure 3.8) are shown on the first row in Figure 3.9. The proportions of Iypes from the
'"
different periods vary somt:what, but all collection5 from this area in the site are characterized
'"
by very high propJrtions of Upper Xiajiadian sherds and v{'ry low pro~rtions of sherds from tir
the other two perioos. Whichever location might have been chosen for II systematic collection
in regional survey would have represented the sherds on the surface in the collection unit rea tv

sonably well. Exactly the same can be said of the collections made in the north-central unit
'"
(those shown in the second row of Figure 3.9). All four systematic collections in the 5Outh co
eastern unit (the third row in Figure 3. 9) have much more similar proportions of Lower Xia ro

jiadian and Upper Xiajiadian sherds than do the colla:tions in the first two rows and no sic
Zhanguo-Han material. Sometimes Upper Xiajiadian sherds are more numerous than Lower Ie

Xiajiadian. and sometimes the reverse, but the propxtions for these two periods are alWllYs T,
much more nearly even than in the first two rows. In similar fashion, the systematic colJec V
tlUns from the northernmost unit (the fourth row in Figure 3. 9) have substantial amounts DE
Lower Xiajiadian and Upper Xiajiadian and very little Zhanguo-Han. And the collections
from the northwestern unit (the !:xJttom row in Figure J. 9) are predcnninandy Lower Xiajia
I.
r.
00

'"
'J'
dian. While the proPJrtions of i;lherds do vary across this site, dividing it into five collection I~

un~ts, each represented by anyone of the sy~tematic collections actu?!ly made within the unit sit
would provide a reaoonable approximation of the overall distributional patterns. 1m
Systematic collections cannot, however, be made in all instance~. In some sites surface 67
artifacts occur at such low density that there may not even be 20 sherds visible on the surface of
in an entire hectare. If it appears that the surface artifflct density is so low in a collection unit th
that 803m diameter cirele would not contain as many as five sherds, then crews do not at I"
tempt systematic collections, but make opPJrtunistic general collections instead. Crews col no

lect the first artifacts they see in a collection unit until the minimum sample size h.as been sur
"
passed; then rollecting stops. This procedure is an effort to eliminate judgements aoout co
which artifacts to collect and which ones to leave on the ground, so as to reduce the sampling
'"
bias in favor of more noticeable, more unusual, or more interesting artifacts that has often th
been noted for such opportunistic collecting (e. g. Drennan 1996b:88) , If crews range
""
Widely around the all''a of a collection unit making a general collection, then the worry that ,~

the single small loc..:J.rion from which a collection is made may not represent the entire area
very well (discussed aoove (or ~ystematic collections:' simply deJl:'s not arise. Although our Ie<
prE"ferred method of mllection if' ;oystematic , ~:te~ whert ~urface artifact demities are high on
enough to make it practical arc a distinct minority in the Chifeng region. Up through 2001 . d,
, 138 .
nit only 241 of 2168 collections made (about 11 %) were systematic, Despite comprising such a
%f
the ~mall fraction of the L'ollections, the sy,.;tematic collections have provided extremely valuable

red I infonnation on snrface artifact dem;i[ie~-especially so considering thaI they are the collec

om tions from the largest and densest sites.

,on Since stratigraphic tests were carried out at Site 674, it also provide:! lL'3 with an oppor

"' tunity to explore the relationship betwecn surface collections and the materials encountered in
lnlt subsurface 5trata. Test A was located near the bonndary between the two centrally located
th- collection units shown in Figure 3. 8 (those whose collections are represented in the top two
.Ia- rows of Figure 3.9). This test contained primarily Upper Xiajiadiall ceramics down to con
no siderable depth. Tests B and D were located in the southeastern roUection unit (whose sys
tematic collections are shown 111 the third row of Figure 3. 9). Some 92 % of the sherds from
""
,y, TesT B were Lower Xiajiadian, a somewhat higher proportion than encountered in nearby
eo- Collection 17, although Collection 17' s proportion of Lower Xiajiadian was the highe:'it of all
, of collections in this part of the site. Test 0, which was excavated down to sterile soil, had
ons substantial amounts of both Lower Xiajiadian and Upper Xiajiadian material, with Lower Xi
]la- ajiiadian being slightly more abnndam, much like nearby Collection 16. Test C was in the col
ion lection unit with Collections 19, 20, and 21, and. like these colleclions in general, showed
mit similar amounts of Lower and Upper Xiajiadian (with the latter slightly predominating) a
long with a small amann! of Zhanguo-Han ceramics. The systematic collections made at Site
'co 674 prior to stratigraphic testing, then. provided a view of the proportions and distributions
'co of ceramics of different periods broadly consistent with that which emerges from analysis of
mit the ceramics from our stratigraphic tests. Even at this site. where cultural dePJSits :teII.ch at
,,- Jea'il 5 to 6 m beiow the surface and three different perioos are represented, there is no pro
nl- nounced tendency for earlier materials to be underrepresented on the surface because the earli
;ur- er dep:JSits are overlain by more recent materials. Our results, then, agree with others about
fjut consistency betwecn surface lind subsurface remains (e. g.l))wnum and Brown 1998). In
ling some cases, surface remains may provide an even roore useful indication of ancient activities
-ten than subsurface remains, as in cases where no detectable stratigraphic evidence of features re
age mains bU[ the characterislic signatures of artifact distributions have not been erased and are
_hat recoverable with careful surface collecting (e. g. Hawkins 1998).
Aio we continne to conduct stratigraphic tests in Chifeng al more sites representing dif
"'"
om ferent periods, we will continue to systematically monitor the correspondence between surface
ligh and subsurface assemblages. So far, though, it ~ thal the surface-collecting procedu~
11 described above pnxluce samples of artifacts of sufficient size and small enough risk of serious
. 13'1
bias to enable us to arrive at quantitative ch8racteri7.ations of the chronological ;x-riods repre ,h,
sented. The same characteristic'> make it possible to analyze artifacts in functional terms 3.
(proportions of different vessel forms, for example) so as to reconstruct how patterns of ac'
tivities may have varied from one rollectiun unit to another during any particular period. The
'"
prr
spatial resolution of ooth chronological HOO functional analy~ is the 1 ha rollection unit. If ev,
future work suggests that, for example. eRrlier periods are systematically underrepresented in au
5urface rollections, then the same rompari~s of surface and subsurface assemblages that 05'
sugge,~t this fact can be used to establish "correction factors" based on the degree to which I",
early materials seem attenuated in surface rollectiotl'l. be
The fourth parameter to characterize for rollection units is intensity of utilization. One sib
of the principal clues to this parameter available to regiomtl 9urvey is surface artifact density. vis
The rollecting proceciuIl'S just discu~ provide information aoout surface sherd densities, de<
first in ten11.9 of two rough categories: high and low. Low density equates to tho..'Ie collection
units where general collections were made because densities were nul high enough for system
atic rollection circles to ~ practicable. This is around 3 or 4 sherds per circle, so low density
2
areas wht:rt: general rollections were made have Ie&: than aoout O. :; ..hf'rds/m on the surface.
r.' ed
tl~

are
2
High density consequently means more than aoout O. 5 sherds/m , and the systematic collec 001
tion'" provide for precise quantification o{ these densities since both numbers of sherds and the wil
2
specific area'! (in m ) from which they were collected are known. me
As might be expected, systematic collections tend to oontain more sherds than general Chi
rollections (Figures 3.10 and 3.11 ). The fact thaI 1465 of the 1691 colJa:tions made (87% ) fo'
are general collections indicates that the vast majority of site area recorded in Chifeng oonsists tin
of relatively low density surface artifact~ (roughly O. 5 ~herd~ per m" or leSl). Surface sherd =
densities, as measured by systematic collections, are mostly less than 5 sherds per m2 but to

range up to a maximum slightiy over 22 ~herd::; per ml(Figurc 3.12). de


This variation in density is, of murse, affectEd to J'DITle extent by surface conditions, R"
and the degree of surface visibility is also recorrlPd for each collection unit acmrdinp; to subjec 'u'
tive judgement as high, medium, or low. Since surface ronditions have already been de bl.
scribed as favorable for settlement survey in the Chifeng region generally, it will rome as no tio;
surprise that fully 78% of the units coliected through the 2000 seBs:::Jn were rated as having "'
high surface artifact visibility; another 20%, moderately good vi."ibility; and only 2%, low tlK
visibility. The existence of this variation does, however, make it worth investigating its im lng
pact on surface ceramic densities iI." recorded. 20<
Just as one might feat, mllection units with low surface vi~iiJijilY hay/: the 1o-...e.:>t mean
"'
140
re sherd density, and those with high visibility have the highest mean surface density (Figure
3.13). There IS some statistical sigmficance to these differences. as indicated by the error
""
". ranges in Figure 3. 13-principally because the numbers of collections are large enough 10

be pn:xluce quite small error ranges. The magnitude of the differences between the means, how
II ever, is quite small. Looked at another way, it is clear that the degree of surface visibility is
in correlated with surface ceramic deILSity, and this cor-relation has some significance (r = O.
...t 056, p=O.QZ3; r,=O.040. P=O.10B). The correlation, however, is extremely weak;
ich less than 1 % of the variation in surface ceramic density is explained by visibility (0.3%, [0

be exact). If this correlation had Mn stronger, it might have been useful to explore the PJ5
>ne sibility of correcting the den.sity figures for condition.s of surface visibility. As it is, surface
:yo visibility is shown to be such a trivial comp:ment of the variability observed in surface ceramic
~, densities, that we simply OJ rot need to 'WOrry al:u..tt it.
can Ccrnbining accurate asse&'itl1ents of areal extents and of intensity of utiliz.ation as indicat
m ed by den.sities of artifacts on the surface provides an especially JXlwerful tool for regional set
ity tlement studies. This is particularly so when it comes to demographic approximation.s, which
:e. are discussed further below. Suffke it to say here that the field procedures for making surfac'e
ec collections described aoove provide artifact samples whose size and relatively unbiased nature
:he will sn.stain quantitative chronologieal and functional analyses connected to area and den.sity
measurements at a spatial resolution of 1 ha. This is substantially finer than the site-scale
"j characterizations of these parameters provided by many regional settlement studies. We have
~ ) found these field procedlll1".5 efficient enough in Chifeng that they require very little additional
time in the field so that they do not noticeably slow down field work or result in reduced areal
'"
~rd coverage. They have the additional advantage of shifting more of the pl"ClC5S of interpretation
>tit to the post-fieldwork phase of analysis when it can be based on systematic quantification of
c1e.med artifacts in the laooratory rather than impressionistic field ~ments. The demo

". graphic analysis presented in the next chapter can serve as an example of the kinds of analysis
~c- such field procedures pemut. Clearly, even more accurate quantification of artifact as;em
le blages, their spatial extents, and their densities can be accomplished with extensive excava
no tions. Regional survey cannot be expected to provide the same level of detail; it can, howev
ng er. provide less detailed information for all the sites that can be found in areas of hundreda or
,w thousands of square kilometers. It is the comprehensiveness of such a data set (now consist
m ing of 2176 collection units in 1069 "sites" scattered through 765 krn 2 in Chifeng after the
2001 survey season) that, as in the case of census data, gives it its value. That said, the fin
er the spatial resolution that can be achieved in regional settlement study, the better-as long
"" 141
as sacrifices in comprehen."iveness are not required. We have found that focusinR on collection Fig
units provides finer spfltial resolution than focusing on .sites, with little or no sacrifiee in COTJ1 of
preherL'llveness. SIS

prr
3.3 Sites as Units of Analysis ~p

As discussed above. we have used the site as a unit of data recording in a minor way, tab
grouping collection units. and such thing5 as notes and sketch maps of features visible on the [eo
surface together into numbered :r;ites. It is a simple matter of convenience to group together am
things that are spatially contiguous and separated from other IU"cheological rIIlain:r; when the tril
distribution of archeologieH1 remains is not continuous (as is the case in Chireng). In the tio]
chapter just concluded, however, it became clear that. even though the subjeet was the char 3.
acteri7..ation of sites, the collection unit quiekly superseded the site as the prime unit of analy ed
:r;i:r; and that this provided for more accurate and precise chamcterization of large site.<;, sector
by sector.
'""
th,
As regional sett!IIlent study turns to more abstract analyses, a different aspect of u:r;ing rotC

:r;ites as units of analysis beromes important. It is customary practice in regiona! settlement Th


:r;tudies to carry out analyses that assume a one-for-one correspondence between sites and an th,
cient human communities. Examples of such analyses would Include the examination of his th,
tograms of site sizes to infer chamcteristics of settlement hierarchy or of rank-size graphs to
""
as."lf'$ regicna I centralization (e. g. Johnson J 977; Inomata and Aoyama 1996; Liu 1996;
'"
Savage 1997; Underhill et aI. 1998; Mudar 1999). In the cases of both these analy bul
ses , the assumption is that each archcol~ical site represents a human community that it is
meaningful to distinguish from others and to delineate in the same way rhat it was delineated
'"
wo
as an IU"cheol~ical site. The conventional way in which sites are defined in the field during
regional survey mayor may not provide justification for such assumptions. tat
Figure 3.14 shows the collection units from a small portion of the Chifeng survey area as ,h,
an example of the impact site definition can have on analysis. All collection units are indicated th,
on the map, with those that produced sherds dating to the Hongshan period colored solid be<
black. Sets of collection units shown as contiguous on the map were called individufll "sites" to'

in the field. Our decision aoout ..... hether to eH1l a surfflce scatter one site or two follows the d"
common pmctice of calling a spatially continuous distribution of surface artifacts a single site. to!'
Since we will uS{' onJy the Hongshan siles In this example. those thflt had no collection units tb,
with Hongshan ct:ramics will simply disappcqr from the analysis. As can be clearly seen in th,
]42
'n Figure 3.14, for most sites, the area covered by Hungshan occupation (the sum of the arc~

n or the collection units that produLcd Hongshan ceramics) which forms the basis of this analy
sis is smaller than the total area of the site (which would also include the collection units that
produced no Hongshan). Altogether. the area mapped in Figure 3. 14 contains 40 spatially
separate surface scatters of Hongshan ceramics (40 Nsites").
When the site is taken to be a meaningful unit of analysis, it becomes important to es
y, tablish some standard gOYeTIling when to call a gap in the apparent surface distribution of arti
he facts large enough to require defining the scatter as two sites rather than one. It is not un
common to use 100 m as the criterion. That is, if a gap of as much as 100 m exists in the dis
'"
he tribution, then two sites are defined: gaps of less than 100 m are ignored, and the distribu
he tion is treated as the oontinuous manifestation of a single site. ::Orne of the "sites" in Figure
,, 3.14, however, are not as much as 100 m apart. Fjgure 3.15 illustrates these sites surround
ly ed by 50 m buffers; where these buffers intersect, gaps between site:l are less than 100 m,

'0. and the two sites separated by such a gap might be called one. For cJUUTlple, the grouping in

n"
I' the north central section of Figure 3.15, originally called three sites in the field, would be
oome only two ~ites, as two of the original sites are separated by a gap of less than 100 m.
oot The grouping originally called five sites just slightly farther south. would become four sites as
m- the two closest together would beoome one since they, too, are 5eparated by a gap of les.s
lis- than LOO m. The total nwnber of sites shown in this map would become only 31 if all sites
to separated by gaps less than 100 m were combined. If, instead of 100 m, 200 m were taken
16; as the minimum gap in surface distribution u5ed to separate sites, then all thooe whose lOO m
ly buffers intersect in Figure 3. 15 would be oombined. The original three sites in the north cen
: is tral section of Figure 3. 15 would berome only one, and the original five sites just to the south
ted would also beoome only one. There would be only 22 sites in the area.
lng It is not at all surprising that the total number of sites changes if the criterion for sepa
rating surface scatters into different sites is changed. The changes noted in this area of Hong

," shan sites may seem fairly unimP='nant. but they can result in major qualitative changes in
ted the oonclusions reached from some kinds of analyses. It has, for example. as noted ab:we,
)!id berome OOTl1mon in regional settlement studies to examine histograms of site si~ in an effort
" to determine the number of levels in the settlement hierarchy and thus something aoout the
"the
degree of mmplexity and centralization of regional organization. Figure 3. 16 provides his
teo tograrru; of site sizes for the three versions of the Hongshan sites illustrated in Figure 3.15. In
uts the histogram of the sizes of sites as defined originally in the field (at the left in Figure 3. 16)
m the distribution has the general shape of the Poisson distribution one would expect of such da
. 143
ta in the event that only one kind of site was present. This histogram would, then, ordinari
'"'

ly be interpreted in regional settlement analysis as an indication that no site hier<m:hy was m


pre..sent. since a single level does not constitute a hierarchy. The histogram of site si~ when ,0<

sites separated by less than 100 mare ctJmbined (in the center in Figure 3. 16) shows a rather bit
different pattern. Here the three largest sites (from 5 to 8 ha) are separated from the rest in ,""

the kind of distribution usnally labeled bimodal in regional settlement analysis and taken to m"
indicate a two-level settlement hierarchy. crJITI~sed of three central places and a number of rna

subsidiary coITlmnnities. The histogram of site sire when sites separated by less than 200 m of
are combined (at the right in Fignre 3. 16) tells yet another story. The departure from uni h",
modality is much stronger; this histogrnm would surely be interpreted as indicating at least
""

two levels of regional settlement hierarchy, and quite possibly three, with three central places fro
between 8 and 10 ha. one 5eCOndary center between 6 and 7 ha. and a number of subsidiary log
communities. Examination of site size histograms for muhimodality would, then. result in th,
sharply different conclusions about the degn-e of complexity and centralization of regional or tiol
ganization for this set of survey results. depending entirely on minor change; in the criterion of
applied to decide whether to call a partieular surface distribution Df artifacts one site or two.
Much the same effect is seen in the rank-size analysis often carried out in regional settle
""

,h,
ment studie'>. In rank-size analysis, sites are rank-ordered according to sire, and the loga
rithms of site size; are plotted against the logarithms of the ranks. The "log--nonnal" pattern ing
produces a straight line, and a convex line is interpreted as an indicator of weak centrali7..ation tie'
or integration. while a concave line is taken to indicate th{' strong centralization of a regional hill
system with an extremely JXIwerful "primate" center (.r ohnson 1977; Savage 1997 ) . d"
Rank-size plots for the same Hongshan sites from Figure 3.15 are illustrated in Figure 3.17. I'T!
In the plot at the left, the sites as originally defined in the field show a ronvex pattern indi to ,

CBting decentralized regional organil.Btion. In the plot at the center, with sites separated by no'
less than 100 m combined, the line is even more ronvex, indicating an even more decentral
""

iw:! system. And in the plot at the right. with sites separated by less than ZOO m <umbined, inti
the convexity of the line has increased yet again, indicating a very strongly decentralized sys
tem. As the criterion for defining sites was modified. the ronclusion about the degree of cen
l

It

oth
thrl
tralization in regional organization once again ehanged :sharply. The direction of the change, ee,
moreover, runs oPJXlSite to that seen with the site size histograms. As slightly more distant t

,
M'
scatters are combined into single sites. the site :<ize histograms suggest greater complexity and , ,, site
centralization in regional organization while the ranksiZ-t' plots suggest les,: centralization. k""
!
It is an issue for concern that both these conunon tools of regional ~ttlement pattern sep.
144 .
>an- analysis prOOuce such confusing, indeed a:mtradictory, results. depending on minor variation
W~ in the decision aoom whether to call some surface scatters of artifacts one sire or two. The
,hen concern only increases when we recognize that these criteria are usually est.ablished largely ar
ther bitrarily before carrying out fieldwork, if, in fact, they are the subject of much explicit at
,[ in temion at all. One reaction to the recogni[ion of these probleII18 lXluld be an effort to deter
n to mine what is the "corrEct" way to define a site. There is, however, no a:mvineing way tD
~r of make this determination. It is much more proouctive to seek ways to avoid using the notion
Om of site as a unit of analysis. This re1)nates strongly with the "siteless survey" approach thar
um has been advocated by Dunnell (1992) and orhers. Thc reasons we have arrived at such a
least a:mclusion in the analysis of the Chifeng seulement data, however, are somewhat differt:ont
from, although not inconsistent with, Dunnell's argwnent. Unlike Dunnell, we find archeo
"'"
liary logical sites to be "real" phenomena in the Chifeng region, as diseussed aoove. Application of
It in the site notion provides a reasonably adequate characterization of the discontinnous distribu
lor tion of archeological materials on the surface, and has some utility in facilitating the recording
~non of some kinds of data.
WOo The difficulties arise when we shift from using thl' conCl'pt of archeological site as a
ttle :;()mewhat vaguely defined unit of data recording to using it 85 a nnit of analysis in a context
oga that assumes that each archeological site represents a single human community that is mean
.tem ingfully divided from nearby archeological sites which constitutE separate human communi
Ilion ties. The notion that there is such a perfect co.rt'S{:Ondt:once between archeological siles and
iDnal human communities, of murse, arises from the asswnption that groups of people living in
17) . close proximity to each other interact more intensively than those living farther apart. By and
17. large this notion.seems valid, but the distance-interaction scale is not always easy to divide in
ndi to c1earcm categories. The map of Hongshan sites in Figure 3.15 illustrates this {:Oint. In the
d by north central section of this map are Eight sites (as originally defined in the field). It seems
traI reasonable to think that each one represents the living area of a gronp of pcople more tightly
ned, interrelated with each other in at least some ways than they were with the residents of the
'ys other seven sitcs. At the samE timt:o, it seems plal..l5ible that those who lived in the northern
CEn three sites, on the one hand, and those who lived in thl' five sites farther south, on the oth
nge, er, were more dosely interrelated in some kinds of activities than were these two groups with
,lant each other. The more inclusive criteria for site definition that make the;e occupations two
. ""d sites rather than eight would reflect social reality at this slightly larger scale. By the same to

'. ken, these two larger sites are only slightly separated, comparl'd to the larger distances that
ttem separate them from their other neighoors. This suggests that this entire group of people may
145
have interacted more strongly with each other than with the residents of the other more dis for l

tant site.o;. At the next larger scale. then, these two larger sites could rea.~nably be combined
"
as II single human corrununity.
In sum, human communities are composed of nested sets of relationships that form ever
expanding interaction networks. This can be expected to be reflected archeologically in re I,
gional-scale ~ttlement data in the form of hierarchically nested sets of "site;". In different eith,
places and times, there are likely to be different scales at which it is meaningful to draw the
'em
dividing lines between human comrnutIities. This issue cannot be dealt with effectively
j, enti:
through arbi trary decisions prior to fieldwork al:xJut how far apart surface artifact scatters ,~u

must be to be called separate archeological sites which are then assumed to reflect ancient hu stud
mlln o:nnmunities one-far-one. The definition of meaningful human o:nnmunities at various =,
scales should, instead. be an objective of explicit and systematic analysis in and of itself.
,I stro:
Such analysis, like site size histograms and rank-size plots, involves demogrnphic a.."'-"ss gist.:
ments, fot this (5 what site area is really taken to mean in lx>th these grnphR. but
;1
In the next chapter we will consider approaches to making such assessments that take us Fo'
beyond simple use of site areas as a proxy measure for P=lpulation, s;) further consideration of ",ed
such analysis is deferred until then. Suffice it to say for now, however, that reJying on
j. (Fif
archeological sites as units of analysis ooITesJXlnding to ancient human oorrununities seems ~J
unpromising with regional settlement data [ike that from Chifeng, that we prefer to pursue pnn
other approaches that do not rely on this assumption. Although we use sites ill a very approx est ;
imate way as an aid to practical data recording, we do not focus analysis on sites. instead the
i', It h
mllection unit is the basic unit of analysis. making no assumptions whatever aoout any corre tura
sJXlndence between mlJection units and ancient !'OCial rcalities. In effect, the mllection unit disso
j
becomes analogous to the grid square familiar jf! site excavl'ltion. When excavating sites where moo
it does not seem practical or worthwhile to record the JXlSition of each individual artifact with rom
precise axmiinates in three dimensions, we are quite accustomed to establishing a grid and
l.I.Sing squares(of 1 by 1 m. 2 by 2 m, 5 by 5 m or some other sire) as spatial units within
"'''
gulli
which artifacts recovered are aggregated. In just the same way, mllection units can be used ,ern
as the grid squares of regional survey. Rather than attempt to record precise OXlrdinates for nun
each individual artifact, the surface is examined grid ,'\Quare by grid square (that is, mllection
unit by rollection unit). If more than two artifacts occur in a grid square. its location is vallI
recorded, artifflcts are mllel"ted, and it bemmes a "Rite" (or part of one), The~ grid new
&J.uares, a.~ seems appropriate to the regional scale of study, are much larger than the grid haVt
squa."'eS of a site excavation-approximately 100 by 100 m-flnd they are irregularly shaped te.~t(

146 .
---_._-~

:iis for mnvenience ratber than truly square, but the principle is the same.
ned
3.4 Environmental Processes
:ver
re Any regional survey faL-es data rtiXwery problems re.''lUlting from environmental processes
rent either contemporaneous with or :subsequent to the creation of the elements of the archeological
the record it attempts to record. One of the principal benefits of systematic examination of the
vely entire landscape in regional survey is the confidence with which one can say that areas on the
ters resulting map that are devoid of si tes are areas that were not occupied during the period being
hu- studied Having this eurnidenee depends on considering explicitly whether an absence or
scarcity of sites in some part of a regiou could result from environmental processes which de
"""
;elf. stroy, modify, or obscure the archeological record. This has often been rerognized by geolo
;ess- gists and geoarcheologists (Bn:okes, Levine and Dennell 1982; Waters and Field 1986),
but regional ~ttlement analysis has often been carried out with little attention to this issue.
~e us For Chifeng [he environmental processes of concern are rather different for the two zones cre
In of ated by the principal environmental dichotomy of the region-uplands and alluvial valley floor
, on (Figure 3. 18).

'''0 In the uplands, as already noted, surface visibility is goa:l at most times of year. The
IISue principal geological activity of concern for site identification is erosion. Sheet erosion is mod
,rex- est and seems largely to help pre,,-ent surface artifacts and features from beooming obscured.
j the It has not been severe enuugh to produce much site de..struction or movement of ancient cul
"re tural materials. Gully eJ'08ion, on the other hand is .';vere, producing an upland zone sharply
unit diSSL"Cled by gullies sometimes 20 to 30 m deep or more, with vertical walls. These can be al
rhere ll1.OOt imp:lSSible to crass for distances of a kilometer or more, representing a substantial in
with convenience to the movement of survey tearns in the uplands. Survey paths, cornequently,
I and are organized parallel to them to the extent p:lSSibJe .~ deep as they may be, however, these
ithin gullies are, for the most part narrow enongh that only the &nallest of sites could have been
u.oo removed entirely by this erosiun, and even the measured areas of larger sites would be onJy
~ for minimally affected.
1;tion The lone exception to this generalization is at the juncture between the uplands and the
on IS valley flooIS, where gully mouths broaden out. The very large gully mouths do not represent
grid newly fanned features during the time since the periods we are studying, but many of them
: grid have certainly broadened considerably, destroying cultural deposits in the process. This is at
taped tested to by the fact that sites on the bluffs. above the valley floors are sometimes visibly cut
147 .
by the gullies, from whose walls arttfacts and features may spill ouL At ~me gully mouth~, nug
the destructive effect of this natural efl)Sion on archeological sites is substantially exacerbated mo'
by the excavation of clay for brick-making. This is a particular problem for high density m-",
Lower Xiajiadian site.'i, which are often located on these bluffs adjacent to gully mouths, and toll ,
allowance may need to be made for the wmbined impact of gully erosion and brick-making in ley I
settlement analysis for this period especially. One prospect for determining the magnitude of
"'" I
the allowance that might need to be made is to search carefully in the natural <llJuvial fan de the
posits at the mouths oE the tributary streams for redep:>sited artifacts (Waters and Field tim{
1986). 1 much erosion and redep:lSition occurted, then rede(XlSited artifacts should appear at com
least occasionally on the surface of the fan deposits and in places where they have been ex
'","
~ through mo:::Iem excavation or erosion. Dating these ceramics in the same way that ce bed
ramics rollected iT! situ are dated would make possible the evaluation of the relative impact of rece
such processes on different pericds. Since these data have not been collected yet in Chifeng, how
this must remain for the moment a programmatic statement. And
There has been a potentially more seriou:o impact on settlement analysis from geological
action in the alluvial valley floors. Jt has long been known that archeological sites are scaree ru.v,
on these valley floors, and this represents a sharp contrast to the distribution of modern occu subs
pation, which is heavily concentrated on the valley floors adjaeent to the most fertile ~ils rem,
which are, tcxlay, the most intensively Olltiv<ltro zane.'i in the region. Substantial sections of bea
these valley floors are today subject to regular flcxxling. and considerable energy has been in pattI
vested in earthen dikes in an effort to contain the.<>e flood waten:. men
It is possible that ancient sites exist in the valley floor, but that they have been covered ncr"
by recent dep::lSition of sediment produced by gully erosion in the uplands. P.s noted aoove, cupa
this erosion goes on vigorously today, and the potential exists for its occUrtence over quite a othe
2
long time since the periods we are studying. Of the 550.5 km surveyed up through 2000, need
146.1 km2 (27%) rol1..5ist of lowland valley floors. The total area of all colleetion unit~ in (d.
2 2
these surveyed valley floors is 3.4 km (17% of the 20. 3 km total). The first systematic
survey of the valley floors, then. did ioc.ate substantial numbers of sites there. It is elear that
evidence of ancienc occupation has not been systematically and completely obliterated by more
recent sedimentation in the valley floors. A sub~tantiaJly smaller pereentage of the site area
does occur in the valley floor than would be expec.ted if sites were randomly distributed across 198;
nplands and valley floor, but there is, of course no rea~n to expect them to be randomly di,, guo
tributed with regMd to this imporcant environmental distinction. p,,,,,
If site." were buried by more recent sediments HS part of a long-term process, then one in ot
148 .
ths. might well expect that older sites would systematically be obscured to a greater degree than
ole<! more recent site:s. Thi:s could be reflected in a :steadily decreasing propxtion of recorded site
~HY area in the valley floor as one moved farther and farther back in time. Table 3,3 gives the to
and tal area of .'lites recorded up to now for each period, together with the area of !:iites in the val~
",m ley floor (and the percentage of the total site area for the period). 'Bl'08.dly speaking, we do
Ie of see lower percentages for earlier periods, but the figures in the table are not consistent with
l de the idea that the ability to detect ancient occupations decrea..-.es in proportion to the amount of
Id time that has elapsed for them to be covered over by sedimentation. In the first place, the
",. ,t complete absence of Xinglongwa, ZhaobaoKQu, and Xiaoheyan sites in the valley floor is not
l ex~ surprising. Given the very small total areas of sites for these periods. this observation must
t ce be disrounted. The laJl?:cst pro!XJrtion of site area in the valley floor is, indeed, for the most
.ct of recent period; and the second largest. for the serund most recent period. Lower Xiajiadian,
eng, however. has a much higher proportion of site area in the valley floor than Upper Xiajiadian.
And the pro!XJrtion for Hongshan, while low, is still substantial.
~ical These percentages certainly do not rule out the \U>Sibility that a number of sites may
ame" have been covered by more recent sediment, but they make it quite clear that there are also
x;cu substantial amounts of occupation, going far back in the sequence, that have not thus been
soils removed from easy detection. Although the occupation we see on the valley floors may well
,ns of be affected by such prucesses, we are also without question seeing changes in occupational
:n 10- patterns through time. A priority in plannEd research is a geological evaluation of the sedi
mentary history of the alluvial valley floors, in an effort to identify which sectors of the valley
vered floor.; are most at risk of enough ret'ent deposition to make it im\U>Sible to detect ancient oc
JOve, cupations through ordinary pedestrian survey. Depending on the results of this evaluation,
uite a other field methodologies (such as, for example, a sampling program of auger cores) may
!OOO, need to be devised to estimate the amounts of occupational evidence last under the alluviwn
its in (d., for example ling, Rapp, and Gao 1995,1997; ling and Rapp 1998).
matic
r that 3.5 Chronology
more
~ area The series of archoological culmres defined for the Chifeng region ( Gno 1987 ; Liu
1987;Liu and Xu 1981jXu 1989;Zhang 1991; Zhangetal. 1987; Zhong
""'"
y dis- gno 1974,1987,1988,1998) ha.s fonned the basis of the chronological scheme used in the
preceding pages. We have treated these cultures as a seqnence of periods-a praetiL" followed
~n one in other regional surveys as well. The definition of archeological cnltnres has similarities to
. 149
establishing a chronological scheme, but there are differences between the two as well, dcriv fie<!
ing frrnn their rather different aims, For Chifeng, three principal unresolved chronologir.al is-
Pen
sues arise frrnn these methodological differences. defi
The first of these, in chronolC@"icalorder, roncems Hongshan. which occupies the peri- rot
od betw~n 4500 and 3000 BeE. This archeological culture is well documented over quite a Jla[
large region in northeastern China. Most of the radiocarbon dates that identify its time span sue!
are for samples that rome from sites well outside the Chifeng region. however, and there is, the~
for this rea.~.:m, some suspicion that Hongshan might not begin in the Chifeng region quite as Lov
early as in rome other places. Thus its span in our study area might not really be as long as rot!
1500 yean;. grnl
Sea:md, as discussed in Chapter 2, the tran5ltion between Hongshan and Lmver Xiajia-
dian and their exact chronological relationship to the intennediate Xiaoheyan culture remain
inrompletelyunderstcod(Chang 1986b:375; Guo 1995a,1996b). In our survey SO far,
Xiaoheyan Cramics are extremely rare compared to those of preceding and succeeding peri-
oos. There are only 28 discrete sites with Xiaoheyan Cramics, compared to 129 for the pre- datE
ceding Hongshan culture and 299 for the succeeding Lower Xiajiadian. Only J74 sheros (of a ru,v.
total of 24, 034 classified) a11:' identified as Xiaoheyan, compared to 1 ,527 for Hongshan and alw:
7,288 for Lower Xilljiadian. All this suggests the po.ss.ibility that the Xiaoheyan ceramic com- oon,
plex, while making clear the local transition from Hongshan to Lower Xiajiadian. may nol Om
represent an entirely distinct period in its own right. II this is the case. then occupation dur- tru,1
ing this periocl wCJUld be underrepresented and preceding and/or subseque.nt occupation over- stue
represented in settlement analysis. met
Third, the archeological cultures defined for Chifeng leave a 600-year gap between the the
Lower and Upper Xiajiadian. This reflects the fact that there is a gap in radiocarbon dates at
this juncture in the sequence. The similarity of ceramic style between Lower and Upper Xia-
jiadian does suggest continuity (although S'Jme major social, political, and eronornic changes
occurred, 5ee below). In the demographic analysis that appears in Chapter 4 we have as-
sumed that thE' gap in the sequence of radiocaroon dates is an accident of archeological sam-
pling rather than a true hiatus in occupation. but as of now it is not possible to confinn this
or, if this notion is confinned, to locate clearly the dividing line between the Lower and Up-
per Xiajiadian.
Resolution of these three chronological issues is among our immediate goals for future
work in Chifeng so as 10 make settlement pattern analysis more precise. All three will require
stratigraphic test:s designl'd specifically to clarify chronology. Survey through 2001 has identi-
. 150 .
'nv- fled 160 Hongshan ~ite~, so there is no dearth of candidates for stratigraphic testing for this
I js- period. Clarifying the chronological position of Xiaoheyan is a task that overlaps with that of
defining beginning and ending dates for Hongshan. One reason this part of the sequence is
>en- oot cleu-er already. is that sites with a combination of Hongshan. Xiaoheyan, and Lower Xia-
te a jiadian ceramics are scarce (owing largely to the rarity of Xiaoheyan material). At least three
;pon such sites have been identified. however. among the sites recorded on survey to date, and
~ is, they appear to show some promise for stratigraphic testing. AB for the OOundary between
:e as Lower and Upper Xiajiadian, survey to date has documented 247 sites with occupations for
g", roth the5e cultures, so, once again there are abundant potential locations for relevant strati-
graphic tests.
aJla-
nain 3. 6 Conclusion
fa<
Jeri- In Ihis chapter we have attempted to consider and reflect Up:>n how regional settlement
pre- data are rollected and analyzed. The particular variant of regional survey methodology we
:of a have applied in the Chifeng region is tailored to the particular conditions found there. as must
and always be the case for archeological methodology. At the same time. in one way or another,
"m- conditions in Chifeng are not unlike those found in many other pans of the world as well.
not Our hope is that the discuS5ion ai:xNe will not only make clearer the nature of the evidence
d",- that stands behind our conclusions but also enrourage others involved in settlement pattern
Ner studies in other regions to ronsider more vigorously how we can rontinually improve upon the
methodology of such studies. Some of the iS5ues discussed arove will be PLmlued further in
I the the next chapter when we take up the task of making demographic reronstructions.
es at
Xia-

"""
~ as-
",,"-

this
Up-

Jture
qUIre

enti -
151 .
T.b
3.1 al>llll'J'll!. (1)'1' 5 It III Itll'Jll!. (<IlfJl[ ll!. '" lI'J ttl!!
(/999 !Il 2000 '"i1111t I< Jolll'JIUD
Table 3.1. Small rollectlOR'l as a prop:rttan or IlII coIlectloll'l, by perfuds

(datil from 1999 IIIld 2000 SllfVey area).

/JilLU j.''f 5 f\-rrg*_


All Q,nectic:Jru; Colleeri~ with <s "herd~

"'!lit
"'ltll:.
Numb.e., of aherds
"ltll:.
Number of .herd.

,
I %

Xinglongw.ll " 5.'

""jlJ
Zh."-w
263 6 2.3

IT" 1546 I' 12


-",",
IN\ili'Q

Xioohey>lll
178 2.2
.~li!ir~
1495 117 I 6
Luw"" Xiaj iadian
IBr.!L.t~
6732 115 1.7
Upper Xiaj;8ruan

Id!il-I'i.
298] 17
ZhMg~Hw;
"
;rit
5258 161 3.1
Lillo

.~lt
24510 476 1.9
Toull

. 176
r41 " 3. 2 I> otll!l1!i': l'IlIlIIlII >1M) '" '" , !\ tI.UIlt IIJ;fJll:ilillll'E1li'If , !H:U:'" IlIIJ;flll:
ililll'" IIJ 1t~(l999 I1l 2000 '!'iOJ .lHIIIIJ'Il*'l)
Tablr 3.2. AIl.'ali oC .si1ft; with n~ 1IheTd8 (8' r t l ( f t ftlWld within 500 m of IIlOlkm tOWJl\l all 8 proportion
,
, fJC lota.I area of sllei with nve sberd!i or IIlDn! for eath period (datil fram 1999 - 2000 s..uvey)

1 Jl;tJtll
3.ilIftMtIlJift:lll:.l!Iiel(r(JltiJ!
Proportion of Site Arnil near Modem Town

tl Penal liM) !I'


lOOm
200*
200 lrl
m.'"
Within
!I nlll
! """".... 41. 4% 13.9% 4.2%

IB-Ji5~~
57.1 % 21.5% 1.7%
Law.or XiajiadiBn
,;, IB-Ji5-.tm
., 37.5% 12.0% 1.0%
~ ,
Up~ Xiajilldian

~OO-N
50.0% HL4% 3.8%
'';: ZJ.ng~f<m

i
.,
iII\:
Uw
55.7% 2.1 2% 4.4%

:'
"3.3 l>il'IlI! "' l! 'l'1O.t IIJ;fJll: iii lRl'EIIi.;flll: iii III '" IIJ It ~ (1999 1IJ 2000 '!' iIiJ.IKJOIIIJ 'Il *'I)
Table 3.3. Are&'i of silei in the alluvial vlllicy nOOl" as a proprtion of letal ~ce area, by periods
.,
~.
(dara from 1999 - 2001 swvey)

,~jljjlR fiif;ij--fJ&

1 TOla! Valley Floor

1, <1M
Peria:i
-'fL:n%Jt
(kmJ )
.ljl-1i~.!I!
(k:n2 )
1
""it
XiT1@longwll
!J.tS 0,001 0.7

/,j","
0.37 0.000 0.0
Zhaoblq,ou

;rw 2.69 0.091 3.'


"""""'"
/.N'iirm-
0.36 0.007 19
XiBOheyan

I;i!:!i5-rm IS. 'I


8.15 L 293
Lower Xiajilldian
--
I:i':!lU." o 561
Upper Xilljilldilln
10.17
"
~~-N
~f<m
48' 0.64'1 13.5

iII\: 2.098 25.9


8 10
Lillll

177
. [ill .
UOlP;)II()() J;)d '>PJ;)lj~ JO J~wnN
&:Q~_Ul:'>'lf*.J.,!!f
0~1 001 O~
"0 . . . . . .
01
~
0
if"
" ]Il:
8~"
09 if~
g. i~
, !
I .
08
I>
lool
SU0i]';)[I0' 1lV'
"."'JlJB ,,..
i
t
r
I
I ,;
Jl
,


ti~-~JtIJ* .~>
Collections with no shtrds 1~ 2J-tlllilh-IYol*#1oA
Collections with 1 or 2 sherds

JJt~Jtrt.J*.A

COlleclions with 3 ,herds 4 g- '" It iJ "'H~LA


COlleCIKJUS ""ilh 4 sberds

~,;.

.-
..... .i
-:"- ,
J;
.... , ,.~ -',:'/'
F
,' .
.....
;}';':. "t
. _~:' , .~
',: .' ..
'. . ~ ..'f,;l..:_ .. :,.~
, <.

'..,..
.. 4""...... _.....
..
." . :r ,,~~. " ," .
~ ... .j...~. ,. 00 3.2 1999li12000'l'1fi']ll>1\1:1E;,
i.~ ...
, "'r jJ,*.~5t.1P
". _
.~

,. '. ' ' ' :.- ....."',-'..., .



~'.'I'
'~I:" f - -.
Figure 3.2, Spatisl distributions of l'l'Tla11
Collectjon.~ in 1999 Ilud 2000 SUTW'Y
5)1 ~'J r_Il1l;I)\-IJ*~.8
colleclIon, wilh _~ or more sherds

. 182 .
. IH .
I
,!'.
I
-~
~
Y-
J,-'
-(S=N '.~;I.-..JfI aaaz
pUl? 6661) SFWt:[s ;;wql ljll"" SUOllJ;3ll00
(8 1I1lU' li'IiMillI'" OOJZ '" 6<i6I)
't -(fCii :r 1lflj! CiI
r".,,, m'
!l!IlIl'*CiI-\ilO-\i [~lj")'l' 1'1
I ,~
, <
I
I'
li
15

I
99%
95%
80%

-~ 'fjg
10
1& E'~lRril]

I1
Confidem;e
.g::
J{j{ 8
oa ..
*~
, I

J[Juf-'------1
N ~ ~

g IT tj; 53 i: Ii: <It!! iL fa I
ri" ~Tc
I! oS<
r;
i'tJ.,
Ilr~ lor i':$:]~] fi1 ~ it:.:J
'"C if} Iii- J6 .-
';0' )6"'" I ::0
Q(l
.S" L
Iii X :><: r x r x ~ 53
.- ""
x N .... -.. .c
m o~ m ~ N

" ~

00 3.4 1999 1Il 2000 'l'llliJIIJ!l'1Il1ll '" lif1-IltJllleJ."F 5If iii If iJ<J*lIUWCIltJlII*Jl"U['p Jjf Ii IIHt
~IIJjf'lfeJ."F 5 If iii If iJ<J*_;H,ffll*_ 9' Jjf Ii IMIo~11.I1. 9% ,'FOOTIII "'!1I!'F Ifj dE FoIJI 1~,j,

*_iJ<J H;~"I~I1',!I WlIlIli':JliTffixftiJ<J,j,*_ii<J H;~IJil, JS -,Ii, 'If'll1& Ili iJ<JIJ:JtftX-)


Figwe 3.4. Pn::.p:mions of 1999 and 2000 collectiOf1.~ in each peria::! that cunsist of few-cr than five shcrds,
with error ranges for 80% , 95% , and 99% confidence. The overall prop:Jrtion of 1.9% is indicated by the
dashed line. The low proPJrtiOf1.~ of small collections in Hongshan and Lower Xiajiadian arc of especially
high statistical significance, as is the high proPJrlion for Liao.

".
! Fi
, d

. 184 .
'".
60 Tr---------------------
.
,
.,
..
.. = ~
'O.i 8
40

I
1
{,
"r- '0
~
,
z
20

o , 25 so
iIiV\il},-'-J:!lt
75 100

Number afLiao ~herd.

60 i I

! 40
.. M
li;l;8
"
:6 "::
~~ j
t o , 25 50 7S 100
I,
Ril:r.5T ~fliiIJai!l:
Number of Lower Xia)ladian .herds
~.
"'TI------------------,
'h,
ll,
~
.,. -
.~
.1(1, ~
~
40

,,~

. ~ 20

o
5 25 so 7S 100
~.Iw~Xfi\il:
NlIJtlber ufHongshan sherd,;

III J 5 1999 l!l 2000 'I'll i'll! iillO rE 1'01l1li'1' 5 y,-"'Iili1 5 It Ili/ ItI1'J!U! 1i'J!l!$?l1l>
VHUi! j>" IlllIi'J Ho lOLl,!,,, Ii'J 1l5L:f1'1 )
Figure 3. 5 Frequen~y distriburioru< o( collecrions fmm 1999 and 2000 sUIVey with five or !IDre sherrls far
the (hree periexb w/u;e propJrtiol1.S or small rulJectioru; differ significantly from the overall pror;urtlon fet
col.lectioru; of all ~riods.

. 185 '
, 9BT
Ekldlll.jS JO uos~= JOJ SOllllllIr.lJ.r.Jd
.....!+\![Tlllm::> 5'Il p<tlt=Jdal ~" "~1.:f UJUI} S1Jol,ng\Jl~!p '\JIr.lrlbald "9" ~\d
~PJ~4;; JD JJqw~N
.Q:tl'iill
Oll o
f - - - - L - -_ _L-_ _---+-o
,.,
., 0'
lIIIlpcllB!X J......[11
~.:lm:!?! n

0
III
in,-
" ~ '"~
0 '"
~
;0,
~
"
01
09 ;;-
01111
:trn ' ,, "',.
OJ
L81
,
"
,
. (h'3AJnF 100;: - 6661 UlOJj iiI ~ )
~!Shfl!(ffi SIQ JO) 'u OD!; pUB ill 00;: JO ~}Jnq tplHl ('l.JP.jq) ~l W'3fOW 'Lt: JJTlliti{
(1Il1f:>f ill r,:: li~ 0) \If q, I!J ) ~"" I!'i
fill'" 00' Ill'" 00' I'IIlIIltnm:tfili fill 'llllflfllil'!}'1l<!mj),J, 000' tll 6661 It SID [I{ n III
I
,
,!
."
}
. .,.
,i
891 .
'Ei:l!U'1 oo!1JlIfD OlU! LIJ!S!FlJlls dl'1r=d rUE .zaUlP> pIIJ!lJI? <O'il]JT'S JO Anlfu"'lIj (0 -V)
Ei:l~l :>!t-ldailJ!leJ\:> llUJ JO ~!1K'rJl (vZ - I) ~"UJ!palfD J!llllUfl&,>t> JO SWPl.O:lllf1\M I'L9 "1!S '8" ~H

_1lir;'HfI<ilCili fu' ~ W
"';;lli 01li!li'1'l' (0 - VI 117)) (~Wm",' m -l)JlI7))C;UliOlIl* !JiiH"L9 , 00
3WO~i' 3 woo!, 3. w O~ 3 W OO( 3WOSi: 3W DOZ
NWOSS
D
NUl 009
N w 0.9
NUl QOL
N w O,L
71
j'
,
1

L-L-I.L.L-.I.-.L..L
2 o 10 II 12 IJ 24

-'--L..L..L-.l.-'-~
) , 4 6 7 8 18

---- .. ---- ~
14 15 _. 16 17

IlL .... IIL


19 20 21

L~ 22 23

I'll 3.9 674 'o;lllJt"'-t-!\I!*JI!",::::-t-atlll!iiJ1tlIT Ii a<J Ho fI1I (-H,jr,J!j OO)l..HIJ<i


,HlJft:l>JU,mT Il! ,M:o<m .le Il! liJ otl>l'-N III fill It III tt fI1I. i n n1Il1l<'l'
;!J*JI!!IiI" .'" -IT III *JI!Jil T i'I--t-*JI>./i )
Figur~ 3.9. Ceramic:; of three perio:ls in the 24 sy:>tom-..atic c()Uections fmm Site 674. The bars for each
c'Oliectiou repr~nt the proportions ai Lower Xiajiadian, Upper Xiajiedian, and Zhangoo-Han sherds( left to
right) _ The roUections on ffiCh rem are tine thai v.uJd pertain to tt.... SEIl1"f cdIectim unit in ~ survey,

-,m".
Gener~l colleclioru;
IOOTI-----------------~

"IE"
, 80

'F
'ai 8
.. ~ 40
*]
120
o "
o 50 100 150
~1-*.p..fmwJtI!l:.
Number of sherds per collection

00 310 1999liJ 2000 'l'lliOlJ!l'iBOO r'l-Ri:tt*JI>IIl;i;:+lIllll'i',H, (*JI>./iIl<Ill;!J 1465)


Frequency di",ribution of size; of general collections from 1999 iIIld 2OC(l SIlJVP.y (N = 1465) .
Figure 3. 10.

. 189
061
',{'H...In~ OOOZ pUR 6661 WCU}
,UJ J:ld SpJ:l4S
-m- JiI-U l~fi+;:It- -,trW
Ol ~I 01 , o
o
01
L L O~;
SLlOII~;lt[O~ ~lltlUJ;lJS,(S
:~f*UY~'f
O(9ZZ=N) ,(aA.l11S OOOZ pU'l) 6661
UUI1~:I][U~ J:Id SpJ;llj~ JO J;l(jwnN
.]J: -U ~ f;.IJ :iii i'tfJc.,J,-'f
O~I OOi [)~
L -,--- ---l"
SLlO!l~J[['D :llltlWalS-,S
'!If?t .j,f .1U'K
161 .
~qs lilll.{~H p<o::lI'\)Xllrf >\JllJq pnOS palOJCO
aoolll e:.ue !i;/A.mS llu<lJ!4J ;;141 JO umucOlIl1UJS e U! papJ=J Sl!Un UOlj3<l!lCO]1V 'tI" ;;Imlll.,j
,
N-
I
o
J "b>0 o 0
.. 0
~ 0

, .
0 '" 0
0 ~ 0 <=
0
0
fle
~. o 0
'0
.< 0 -.
" 0 ~ '"
'0 o
o = .a'-. -
I
A:Jl[lqlS!A :lJIl,JJns I
lim fu~'ijIf[;J;!r1<l~
411nl
.'r
r-----------,00
~U:lPllllOJ
i
I .,-
r~J:fI4'J~ I
'-------------"80
, ...
:::
~.'
',' .
"', "'.,
~
Z61
UlJ( I 'i" 01
I
,
-N- tJ
I ~@
@
@

P @ ~
II. ~@l

,
I @ ~
@
~~
@I

,~@i@ l!i!J.
~
,
. 61
'(.IdMOl) fBU)qUUJ ill OOC: lIa\1
=-\ A9 e<>mJlrl>;: "'J:~ pm '(.'lIPM'J.l) p31,nqUn> IU 001 lmll ~I ilJ P"'.\lnlW ~11~ '(~) PPJ <lI.jl
m p;lUTJlP il~ se S31!",_~[ ' a.nUI!.':I U! SillJS Uf.'I..J~ alP JO] =:Sdl!~ JO ruJe..lfulq.} '91 , ~!..-l
Ill;!;'''' II1JWlI!l!l!lt Iill'li!! 1'1#" WIi!I!l OOZ Ill'iJll!llm' l-
*
,IlIJU"'II1JWll! iii !It 1iIl'li!! 1'1#" WIi!IiI 00[ I~l' iJll!ll", ' q,
*
,Jl/JU "'lJ1J]Jfi! Iil i!'y!q, lif lil!ii Hl " ' :r -III!{.\UlIJU1lfll!'l! "fjl U iii 'e " .(III 91 ( III
'4
*Y-kf{/W:1~
01 6 8 ~ 9 ~ to ( l I
, iz ..
"~. ~
'"
~. ~
01
'4 " '4
*U_:d<W~17t
:;.,fL"j,!U!O:l7t
8L9~tll 9 , [ 1
, z Z
01 g
[
""
01 "
~"
~ !J 01 ~
-a. '""
"
-
or or
l>61
, (J:W.D]) p"u~
tu OOZ u~.]1. >&11 "G P"'~ ><)~!10 pUll '( "'JPpllll) p;lU]qUDJ tu 00 I 1~' >Q[ .~q p;a.~ sa~!G '( L:rlctn )
PfllJ;:,yl 1J! fW!W f.:J1!IUlll!JO S8 o;;q!!,-~j 'f asmI1..'l ul sa.!!' ~ "'l{l- Xl} lOljd <JZ!S-JIiUHM 'Ll 'f dIfIitt_,!
llfill".ll111'O!Iii~Iii'i!l !l!!g[clf.!J-lllli<1ii '" OOC r~SlIflHI' i-
,lllill"Wll111li< fil ~ fil'i!l !l!! g[ clf.!J-lllilfil'" 001 I~ oJ' SlIflf~' q,
'llfilii"Wll111i!fil '!('t\q, l<1i\1JI1ll '\!' ,-llfill - liIl<filllli!1f rj1 "'q, ,n III U III
or 51 01
L --=:::.-:===J:OI
)(UIIl ~Ol
JiI~li'Hit-
Of Ol 0;1 01 ..
Or Of or 01
'>61
'pal&J!PUT ~ TOOl 1l~no.np6661 urn] pa!ldfl.,mS ~UO'l dlll
'IDE !lpfll5 all1. U! ~uaz (dlT'..IM.) JOOU !lalleA pm? (paPI'?t\~) f"Nidn 10 Sl.!0!1l?:XTJ 'SI ' aJffii~3

(I1I''iI'' ),Ihldl r,Jo< (-fililHIII!) !If'jlPilf rn iI EPl'I>J :Ifill!'" laoZ r~ 6661 '1 m
L

You might also like