You are on page 1of 4

130 4.

The Village and Its Catchment Area

Smith, C. E., Jr. West, R. C., N. P. Psuty, and B. G. Thom


1965 Flora, Tehuacan Valley. Fieldiana:Botony, 1969 The Tabasco lowlands of southeastern Mex-
31 (4):107-143. Field Museum of Natural His· ico. Baton Rouge: louisiana State University
tory, Chicago. Press.
1967 Plant remains. In The prehistory of the Woodburn, J.
Tehuacdn Volley. Vol. 1. Environment and 1968 An introduction to Hadza ecology. In Man
subsistence, edited by D. S. Byers. Austin: the hunter, edited by R. B. Lee and I. De-
University of Texas Press. Pp. 220-255. Vore. Chicago: Aldine. Pp.49-55.
Virri, T. J. Zohary, D. Chapter 5
1946 Maanjake-olojen Vaikutuksesta Maataloustuc· 1969 The progenitors of wheat and barley in rela-
tantous. Maatalous Agronomien Yhdistkysen tion to domestication and agricultural disper-
Julkaisa, pp. 6-11.
Vita-Finzi, C., and E. S. Higgs
sal in the Old World. In The domestication
and exploitation of plants and animals 1 edited
SAMPLING ON THE
1970 Prehistoric economy in the Mt. Carmel area
of Palestine: Site catchment analysis. Pro-
ceedings of the Prehistoric Society 36:1-37.
by P. J. Ucko and G. W. Dimbleby. london:
Gerald Duckworth & Co. Pp. 47-66. REGIONAL LEVEL

Introduction ley. Rather than "sampling," these archeologists


have surveyed the entire region-field by field,
In Chapter 3, we confronted the problem of meter by meter, often using large-scale aerial pho-
sampling whole communities. Now we must move tographs and multiple crews of five to six persons
to the supracommunity or regional level, in which each. They know they have an adequate sample
the relationships of various communities to each because they collect the entire site "universe" (in
other must be considered. In this chapter, we will statistical terms), setting aside for the moment the
deal with the problem, "How do I get an adequate obvious limitations of site preservation and human
sample of sites?"-surely one of the earliest to be error. This is the ideal. If every archeologist could
faced by an archeologist starting a new project. In do this, we would not even have bothered to dis-
later chapters, we will move to two more questions cuss sampling on the regional level in this chapter.
in which Mesoamericanists have expressed an inter- Unfortunately, most Mesoamerican archeologists
est: "What pattern is discernible in the distribution do not have budgets of $50,000 to $100,000; nor
of sites in my sample?" and "What was the prob- do they have the time or the manpower to survey
able relationship between these sites and between entire valleys. Most must content themselves with
the social units that occupied them?" finding some percentage of the sites in their
Many Mesoamerican archeologists-among them chosen region-all the major ceremonial centers,
W. T. Sanders, R. S. MacNeish, J. R. Parsons, and perhaps most of the larger villages, and some unde-
R.E. Blanton-have taken as their region an entire termined fraction of the tiny hamlets and seasonal
highland valley, or a definable subdivision of a val- camps. If you ask them whether or not they have

131
132 5. Sampling on the Regional Level 5. Sampling on the Regional Level
133
Data collecting Our survey of the valley was not complete in that W. J. Mayer-Oakes and R. J. Nash had reo
systems the sense of a rigorous, careful, on-foot explo- design." Fortunately, he believes that he saw

~
cently presented, which R.M.A. was fond of mis-

/
Purposive or
1
Controlled Complete
ration of all the 'bottom lands.... Such a stag-
gering task was beyond our strength. [Willey
et 01.1965:576]
quoting (at least, S.G.S. and I agree that he was
frequently misquoting).
The argument began in the Quinta Las Rosas, a
Binford subjected to the ultimate put-down. He
tells the story often. In fact, he tells it every time
his Skeptical Graduate Student brings up the
subject of sampling.
'hunchI sampling sampling survey In this chapter, we will make two basic sugges- now-defunct "nocturnal center" on the outskirts "It was at the 1964 meetings of the American
tions, which we hope sound reasonable. First: If of Veracruz. The Quinta was famous for its colo-
I you can survey your entire region meter by meter, nial tile and for the fact that the shower was in the
Anthropological Association, held in Detroit," he
says. "Everybody was talking about Binford's
Random Systematic do so in preference to sampling. Second: If you center of the patio; customers who arrived before article. Well, Bill Mayer-Oakes and Ronald Nash
designs designs
can't do that-if, for example, you can only survey 9:00 P.M. were treated to a parade of 12 nubile had tried out some of his techniques on Bill

Random with
/ \ / Systematic
a 20% sample-for heaven's sake, do it in such a
way that you have some idea of the reliability of
young ladies scrubbing down behind a translucent
shower curtain, before their evening exercise.
Sanders' Teotihuacan survey area, and they
presented a critique ..." (Mayer-0akes and Nash
Random with lukewarm drinks were served in tile booths, under
purposive systematic unaligned your results. One way to do that is by using some 1964,1965).
stratification stratification designs kind of probability sampling. Admittedly, that ceiling fans that did little more than stir up the
At this point, the Skeptical Graduate Student
sounds as if it should take longer than the way you tropical moths. Everyone in the place used eye
always rolls his eyes straight up at the ceiling and
were planning to do it. Believe it or not-as makeup, but it was particularly disturbing on the
shakes his head in disbelief. The action was not
Strat~ied Stephen Plog's section of this chapter perhaps male waiter, who also used spray glitter on his
Nested or
systematic lost on R.M.A., but he was interrupted by the
hierarchical will demonstrate-it doesn't really take much more temples. There was a 10o-record Wurlitzer in one waiter, who had just brought three rum-and-cokes,
designs unaligned
designs time than you were already planning to spend. The corner, and, in the subdued light, the patio took Three young ladies followed the waiter, circling
Figure 5.1 Diagram showing the relationships of differ- real payoff, however, is that, if your survey is done on the look of a small basketball court redecorated our table with little attempt at SUbterfuge. One
ent kinds of sampling designs: from the ··-hundl" sampling that way, there are further statistical manipula- for fue senior prom. On one wall was a recessed was clearly trying to see if R.M.A.'s lap would
of the Real Mesoamerican Archeologist, through the con- tions to which you can legitimately subject your shrine with glowing altar candles and the hand- support her full body weight; I doubted it, but I've
trolled sampling of his Skeptical Graduate Student, to the data; if not, your analysis ends there. In later sec- lettered prayer: "Oh Thou who wert Conceived
complete surveys of Sanders, Parsons, and Blanton (see been wrong before.
tions, we will consider some of the further manip- without Sin, help us to Sin without Conceiving." "What Mayer-0akes and Nash did was to take
text). [Redrawn from Haggett 1965:Table 7.2.]
ulations possible with total or. systematically While the waiter filled our order, R.M.A. drew Bill Sanders' survey map of the Teotihuacan Val-
sampled survey data (see, for example, pp. 236-248). on a paper napkin the outline of the Rio San ley, showing the location of all 500-odd sites he
an adequate sample, either they say "I hope so," Jacinto drainage and the pattern of sites he had had found," R.M.A. went on. "To this, they ap-
or they shrug and say "I don't know." Both an- Probability Sampling and Regional Surveys found so far. Reaching the end of the paper, he plied the 'stratified random sampling program'
swers are correct. They do hope so, and they don't concluded, "and to the south, it looks as if there that Binford had recommended. First, the 750-sq.
know. The Real Mesoamerican Archeologist doesn't were no more Formative sites-just Early Classic, km valley was divided into seven 'strata' or envi-
One of the reasons they don't know is that their Ii ke probability sampling. He regards it as (7) a and some smail Post-Classicsites." ronmental zones: the Rio San Juan delta, the
reconnaissance amounts to what Peter Haggett waste of energy, (2) too time-eonsuming, (3) not Near his elbow, the Skeptical Graduate Student Patlachique Range, Cerro Gordo, the lower valley,
(1965) calls "purposive" or "hunch" sampling (see as reliable as his intuition, and (4) not applicable quickly added, "but we can't be sure of that middle valley, upper valley, and northern valley.
Figure 5.1). Often, such surveys are strongly deter- to complex societies. He even has reservations because our sample of sites is inadequate and our They then gridded the whole map with squares
mined by where the good roads or clearings are, about applying it to such "simple" political units survey so far has been very haphazard and .6 km on a side, and selected a 20% sample of
where the local guides happen to have been, and as Formative villages. He and I have had acrimoni- unsystematic. " those squares at random. The sample was allocated
which mounds are visible the longest distance un- ous debates on the subject, neither of us backed Now, short of calling attention to a whole pro- so that various 'strata' received squares in propor-
der certain lighting conditions. Mesoamericanists up by very much data or mathematical expertise, jectile point on his backdirt pile, there is probably tion to their area-more squares in the biggest
are well aware of the incompleteness of such sur- and each of us continually harrassed by the Skepti- no easier way to make a Reai Mesoamerican Ar- areas, and so on. Finally, they placed their grid
veys, and usually very honest about them. A par- cal Graduate Student, who claims to have both. I cheologist angry than by telling him that his sur- with its sample areas over the map of Sanders'
ticularly frank statement by the very man most really feel I should introduce this chapter by de- vey techniques are inadequate. In fact, R.M.A. is sites, to see how many they would have found."
often credited with inventing settlement pattern scribing one classic argument we all had during the still overheated from having read Binford's 1964 He smiled triumphantly. "And you know what
studies comes to mind: 1966 field season. It centered around a critique article "A consideration of archaeological research they found? Do you know?"

~
,....
. :'. '::
_.'.:''1'';
5. Sampling on the Regional Level
5. Sampling on the Regional Level 135
134
"These Binford communicated to the assembled are only five 'Zacatenco phase hamlets', then such
remember, Maver-Oakes and Nash were rather
HI can't imagine." crowd." sites make up only 1% of the universe; the chances
temperate in their criticism, and even said, 'it
"They missed Teotihuacan. For God's sake, the "What a memory," I marveled. are that, in a 20% sample, you would recover only
seems clear that Binford's theoretical framework
largest Pre-Columbian city in the New World, 20 "Then Binford compared these with the totals one of them. In the case of 'Cuanalan phase large
and specific sampling techniques offer much of
sq km, an estimated 125,000 population, and they for each type of site found by Sanders. And do villages', there are only two in the whole universe;
interest and value to archeologists working
missed it. Now why, for God's sake, should I use a you know what?" small wonder the sample didn't recover any at all.
anywhere' (Mayer-Oakes and Nash 1965:21).
technique that won't even find Teotihuacan? I "I can already guess." Probability sampling isn't the best way to find
I wondered if Sonia, Rosa, and Yolanda were
could find it with my eyes shut and my hands tied "Virtually every type of site recovered by the sites-it's just the best way to get a representative
taking notes. In fact, I was wondering how they
behind my back." 20",£ stratified random sample-'rural Toltec sites', sample of sites, if you can't go for the whole
would score on a surprise quiz on sampling design.
"Yeah, it is hard to miss," I admitted. "It seems to me," S.GS. went on, "that you and 'rural Aztec sites', and so on-was recovered in ap- universe as Sanders did."
"Well they did it. And what's more, as Mayer- several others who heard that talk have a complete proximately the proportion it contributed to the "It's too complicated and it takes too long,"
Oakes ~nd Nash pointed out, the 20%. stratified misconception of what a 20",.6 random sample is whole site universe. If one type of site made up, R.M.A. replied. "And as Mayer-Oakes and Nash
random sample recovered none of Sanders' 'Proto- supposed to do. Somehow you seem to think that say, 40% of the total 500-odd sites, it also made pointed out, 'increasing the areal coverage to find
Classic urban sites', none of his 'Cuanalan phase its purpose is to find a lot of sites-more than up about 40",6 of the sites recovered by the sample. the rarer types of sites is a waste of time and
large villages', and only one of his 'Zacatenco Sanders could find in his total survey, or more As Binford put it: 'the results are an excellent con- resources'" (Mayer-Oakes and Nash 1%5:14).
phase hamlets'." than I could find in a comparable period by racing firmation of the value of stratified random "Why would it take any longer than your tech-
"Not too good, I guess." around the Teotihuacan Valley with a bag over my sampling'." niques?" asked S.GS. "We spend most of our time
R.M.A. adjusted his position slightly to accom- head, picking up sherds,
"Fantastic.' pushing the Jeep out of the mud anyway."
modate the ample young lady who now occupied "That isn't what it's supposed to do at all. "You see," S.G.S. went on, "what the critics Sitting in the slowly moving shadow of the
his left knee. "And do you know what Mayer- "And you, and many others, missed Binford's misunderstood was that probability sampling is ceiling fan, listening to Sonia prattle in one ear and
Oakes and Nash concluded?" most important comment, since itcame at the end not a discovery technique. It isn't a better way to S.GS. in the other, I realized that opinions would
"Lay it on me." of the conference session during a three-way con- find lots of sites. As Mayer-Oakes and Nash them- always differ on what had happened in Detroit.
"They said, and here I am going to quote them versation between Mayer-Qakes, Deetz, and selves said, 'we are not saying that Sanders has Some people had gone away feeling vindicated,
exactly to the best of my memory, 'given the sa~e done a better survey, because ... he has sampled pleased to hear that traditional survey techniques
Ascher."
amount of time, we believe that an archeolog~st "I don't remember a thing," said R.M.A. more than 20% of the area, and it is not as if we would recover more sites, that probability sam-
working by instinct (parenthesis) i.e., expertise "Mayer-Qakes had, in the interests of impartial can pit one approach against the other ... this is pling wouldn't find unique features like Cuanalan
(close parenthesis) could certainly locate a greater about what we would find with 20% areal cover-, phase large villages or the Pyramid of the Sun.
"Scholarship, provided Binford with a copy of his
number of sites' (Mayer.Qakes and Nash 1965:16). age" [Mayer-Oakes and Nash 1965:14). Surveying Others had gone away convinced that only prob-
results before the talk. It showed the following
Now, isn't that what I do every day? Hell, I found the entire area is always preferable to surveying ability sampling would produce reliable, statis-
recovery of sites by the 20% sample (Mayer.Qakes
33 sites last week without a table of random only 20% of it. But what you and most other tically valid samples whose confidence levels could
and Nash 1965:13):
numbers." people do is survey about 20% in a haphazard be defined in mathematical terms. There was no
The Real Mesoamerican Archeologist sat back in Aztec sites fashion. We can never know if you have recovered hope of rapprochement.
satisfaction while we finished our rum-and-cokes (i) Urban-4 each site type in the same frequency with which it "Let's go home," J said, and R.M.A. rose
and ordered a round for our newly arrived com- (ii) Rural-61 occurs in the total universe of sites. On the other unsteadily.
panions. We hadn't heard from the Skeptical Grad· Toltec sites hand, if you took a 20% sample according to "You go on," said S.G.S. "I'm staying to take
uate Student yet, which was unusual, but I figured (i) Urban-2 probability sampling techniques, you could Rosa up on a very interesting suggestion she just
he was too smothered under the weight of the (ii) Rural-3D multiply each type of site by 5 and have some made."
young lady in his lap to reply. I was wrong, of Classic Teotihuacan sites confidence-in fact, a mathematically definable "Her? You think she's the best in the place?"
course; he's never that out of breath. (i) Urban-1 confidence-that the results would approximate R.M.A. demanded incredulously. "Don't you
"I have never," said S.G.5., "heard such a gross (ii) Rural-23 the real site universe." think you should try a 20% stratified sample of
distortion of what went on at that session of the (iii) Traces of occupation-11 R.M.A. sighed impatiently. the whole universe of girls before you decide?"
meeting." Proto-elassic Hamlets-5 "A 20% random sample isn't designed to find "As an archeologist," smiled S.G.5., "my
"How would you know? You weren't even born Cuanalan hamlets-1 Teotihuacan," S.G.S. continued, "or any other instinct (parenthesis) i.e., expertise (close paren-
yet." Zacatenco hamlets-1 type of site that is unique or represented by only a thesis) tells me that she is unique at the 99%
"I was there," said S.GS. "That was back when Pre-Classic hamlets-1 few examples. If, in a universe of 500 sites, there confidence level."
I was a Skeptical Undergraduate Student. As I Preceramic sites-1
136 5. Sampling on the Regional Level Relative Efficiencies of Sampling Techniques
137
All of which, I think, goes a long way toward pher Peter Haggett-simple random, stratified, sys-
tematic, and stratified systematic unaligned. Plog already have been mentioned in Chapter 3 (this
explaining the applicability of probability sam-
•• • ~olume). All four are defined in simple terms and
pling. There are times and places where your then evaluates the "efficiency" of the various sam- •• •
knowledge is limited, and drawing a systematic pling techniques for predicting the total number of
• • • •• Illustrated graphically by geographer Peter Hag-
gett, w~ose diagram we have reproduced as Figure
random sample is the only justifiable procedure. In sites from a 10% sample. Without giving away the
results, we can advise the Real Mesoamerican • •• 5.2. HIS definitions are as follows (Haggett
other times and places, with some prior knowl-
edge, your instincts tell you when you already Archeologist to be of good cheer: The simplest •• • 1965:195-196):
have a really great site. and least sophisticated sampling techniques give

(i) In simple random sampling a sample of N
results that do not differ significantly from those individuals is drawn fromthe areal population
~ In the section that follows, Stephen Plog attacks of more complex techniques. To the best of my at a series of random co*ordinates. The two
the problem of sampling a completely unknown knowledge, this is the first time these four tech- • • • • • • axes of the area are numbered and a location is
valley. He uses the same approach as Maver-Oakes niques have been evaluated, relative to each other, • • • chosen by a pair of random co-ordinates.For
• • • • • example, the random numbers ninety-eight and
and Nash: Three areas of the Valley of Oaxaca
which had previously been intensively surveyed are
for their efficiency in recovering reliable samples
of Mesoamericanarcheological sites. • • • • • • •• twenty-six ,would .give a location ninety-eight
• units north by twenty-six units east; or a grid
"blanked out," and various sampling techniques Of course, the 8-sq-km city of Monte Alban was
• • • • • • • • reference of 9826 in terms of a standard refer-
are applied to them. These include all four of the not found in any of Plog's samples. But then, it's c • ence system. (Fig. 5.2a) shows the location of
major sampling techniques described by geogra- unique. •
·d • • twenty-four points drawn by simple random
sampling for a hypothetical study area.
Figure 5.2 Alternative sampling designs. (a) Simple ran- (ii) In stratified sampling the study area is
dom; (b) stratified random; (c) systematic; (d) stratified divided into natural segments (such as cropland
s~stematic unaligned. '[Redrawn from Haggett 1965: and woodland) and the individuals in the sam-
Flgure.7.4.]
ple are drawn independently from each seg_
rnent, Within each segment the location of the
Relative Efficiencies of sampling Techniques found. The efficiency of each technique-relative points is determined by the same rendomtza-
tion procedure as in simple random sampling
for Archeological Surveys *
to alternative technlques-cls judged in terms of the (Fig. 5.2b) shows such a sample for twenty:
precision with which a 10% sample would make four points. In this case the number of indi-
possible a prediction of the total number of sites viduals has in each segment been made
STEPHEN PLOG
in each survey area. proportional to its area; sixteen points in the
left-hand segment [two-thlrds of the area) and
.In this section, I will describe the sampling tech-
eight points in the right-hand segment
niques to be used in the tests, and discuss their (one-third of the area).
Probability sampling is particularly useful in types of sampling strategies against "real" arche- nonstatistical and statistical advantages and dis- (iii) In systematic sampling a grid of equally
archeological settlement pattern studies in Meso- ological data. The three areas-located in the Etla, advantages, both in theory and in practice. spaced locations is defined with one individual
america, where archeologists rarely have the time, Zaachila, and Valdeflores regions-were walked Second, the data to be used for the tests will be at each location. (Fig. 5.2c) shows a simple
field by field, and all sites were recorded. This discussed. Third, a series of sampling tests will be case for twenty-five individuals. Here the grid is
money, Of desire to study the entire land area that
square and at righ t angles to the sides of the
is the subject of their investigation. However, survey was conducted by D_ M. Varner, R. D. run o.n the data. The effect of different sampling study area. The origin of the grid is decided by
many archeologists are reluctant to apply such Drennan, L. C. Kuttruff, and myself. The actual techniques on the estimation of a population the randomization of the original grid point.
techniques because there have been almost no experience of slogging through mud, wading rivers, parameter will be examined. Also, an attempt will (iv) A stratified systematic unaligned sample
objective tests of the efficiency of such sampling and climbing thorn-forested slopes provided an be made to determine the effects of settlement (Fig. 5.2d) is a composite design'derived from
programs (but see Maver-Oakes and Nash 1%5). appreciation for the practical cifficulties of Meso- pattern and site size on the estimation. the preceding sampling designs by Berry
(1962:7) with the theoretical advantages of
An intensive site survey of three blocks of land american surveys which, I hope, prevents this sec-
(a) randomization, and (b) stratification to-
in the Valley of Oaxaca during the summer of tion of the chapter from being an exercise in gether with the useful aspects of (c) systematic
1970 provided the opportunity to test four major "ivory tower" archeology. The Four Types of Sampling Designs samples. By avoiding alignment of the sample
Basically, what I have done is to "blank out" the points, it also avoids the possibility of error
*The research described here was made possible by an caused by periodicities in the phenomena. . . .
three survey areas, "sample" them by several dif- The four types of samples taken were Simple ran-
Undergraduate Research Participation Grant (GY-757 6) The study area ... is systematically divided
from the National Science Foundation to the University ferent strategies, and then compare the results of dom, stratified, systematic, and stratified system-
into a regular checkerboard of sub-areas. Be-
of Michigan (1970). the sampling with the actual pattern of sites we atic unaligned. Two of these sampling techniques ginning with the corner sub-area, a point, I, is

You might also like