Maney Publishing
Plow-Zone Experiments in Calabria, Italy
Author(s): Albert J. Ammerman
Source: Journal of Field Archaeology, Vol. 12, No. 1 (Spring, 1985), pp. 33-40
Published by: Maney Publishing
Stable URL: http://www.jstor.org/stable/529373 .
Accessed: 26/05/2014 10:26
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
.
Maney Publishing is collaborating with JSTOR to digitize, preserve and extend access to Journal of Field
Archaeology.
http://www.jstor.org
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
Plow-ZoneExpenmentsin Calabna,Italy
AlbertJ. Ammerman
ColgateUniversity
Hamilton,New York,and
Instituteof Ecology
Universityof Parma
Parma,Italy
The experimentsreportedhere were carriedout in the Acconiaarea of
southernItaly between1977 and 1983. Theyrepresentthefirst long-term
studyof the effectsof plowing on materialseen on the surfaceof a site. A
series of observationswas made at a site over a period of six years in order
to monitorthe dynamicsof patternsof surfacematerialplaced at the site by
the researchers.The specificquestionsexaminedincludethe ratio of surface
to plow-zonematerialand the extentof lateraldisplacementof materialresultingfrom plowing. The resultsof the experimentssupportthe proposition
that the site surfaceoperatesin effectas a samplingprocess with respectto
materialcirculatingin the plow-zone. Theyalso show how localfactors at a
site, such as slope, can affectpatternsof lateraldisplacement.A primefeature of the experimentaldesign is its economyin termsof the workeffortthat
it requiresin thefield. Thismeansthatplow-zoneexperimentsof this kind
can readilybe incorporatedinto most multi-yearsurveyprojects.
Introduction
This report on plow-zone experiments that were
startedin Calabria,Italy, in 1977 has two main aims.
The first is to describebrieflythe experimentaldesign
that was employed in the study, since it offers the advantageof requiringcomparativelylittleworkin the field
and, for this reason, can be put to good use in survey
projectsdone in otherpartsof the world. In fact, as we
shall see below, the design arrangesfor time itself to do
much of the work in the experiments.The second aim
is to presentthe main resultsof the experimentsand to
discuss some of the insights along more general lines
thatthey provideinto the natureof surfacematerialat a
site thatis plowed. Partof the motivationfor undertaking
the experimentswas to improve the interpretationof
surface collections made at Neolithic sites during the
course of survey work in Calabria.For this reason, it
was also importantto considerthe ways in which local
factors, such as surfaceconditionsand slope, enterinto
the actualresults of the experiments.The dual task of
trying to develop a general understandingof how archaeological materialcirculatesin the plow-zone and
makesits appearanceon a site's surface,andat the same
time trying to documentthe ways in which local conditions play their part in terms of what is observedin
the field, pointsup the need to have an efficientmethod
for conductingsuch experimentsas a regularpart of
surveyproJects.
The surfaceof a site is the one partof the archaeological recordthatwe have economicalaccess to andyet
archaeologistshave been remarkablyslow in turningto
its controlledstudy.In light of the increasingimportance
of surveys in archaeologicalresearchas well as the
extent of the earth's surfacethat now is, or once was,
undercultivation,it is surprisinghow few studieshave
been concernedwith the natureof plow-zonematerial.l
Several studies that were publishedin the early 1970s
lookedat the questionof the congruencebetweensurface
and subsurfacematerialas seen in termsof global pat1. For a recent review of the literatureon studies concernedwith
plow-zonematerial,see D. Lewarchand M. O'Brien,"TheExpanding Role of SurfaceAssemblagesin ArchaeologicalResearch,"in
M. Schiffer,ed., Advancesin ArchaeologicalMethodand Theor 4
(AcademicPress:New York 1981) 297-342. For a recentreview of
archaeologicalsurveys, see A. Ammerman,"Surveysand Archaeological Research,"AnnualReviewof Anthropology10 (1981) 63-88.
For previousarticlesand reportsin this journalrelatedto the topic,
see especially P. Tolstoy and S. K. Fish, "Surfaceand Subsurface
Evidencefor CommunitySize at Coapexco,Mexico,"JFA 2 (1975)
97-104; K. G. Hirth,"Problemsin DataRecoveryandMeasurement
in SettlementArchaeology,"JFA 5 (1978) 125-131; and Robert
J. Mallouf, "An Analysis of Plow-DamagedChert Artifacts:the
BrookeenCreekCache(41HI86),Hill County,Texas,"JFA 9 (1982)
79-98.
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
34 Plow-ZoneExperimentsin Calabria,ItalylAmmerman
terns at a site.2 It is only more recentlywith a second
generationof experimentalstudiesthatgreaterattention
is being paid to questions of process as such.3 But a
common limitationof the few plow-zone experiments
that have been reportedin the literatureto date is their
short duration:they usually involve only one or a few
episodes of plowing. In conductingthe experimentsin
Calabria,the region formingthe toe of southernItaly,
the plan has been to make observationsover a series of
years in order to monitor changes in the patternsof
surfacematerialover a longerterm.
My own interestin doing the experimentsstemmed
from work thatwe had previouslydone on the repeated
collection of site surfaces(e.g., counts of obsidianobserved in grid squareson differentdates).4One of the
ideas to emerge from this study was thatthe surfaceof
a site acts in effect as a samplingprocess with respect
to materialcirculatingin the plow-zone.In otherwords,
the pieces seen on the surfaceat any one time represent
a sub-set or sample of the set of pieces moving within
the plow-zone as a consequenceof periodicepisodesof
tillage. As the repeatedcollectionsthemselvesshowed,
this sampleis subjectto changefromone boutof plowing to the next. In orderto evaluatethis idea andto find
out whatkindof samplingprocesswas actuallyinvolved,
it was necessary to carry out more controlledexperimentsin whichknownquantitiesof materialwereplaced
in the groundat the startof the study.
There were four main questionsthat the experiments
2. C. Redman and P. Watson, "Systematic, Intensive Surface Collection," AtnAnt 35 (1970) 279-291; L. Binford, S. Binford,
R. Whallon, and M. Hardin, Archaeologyat HatcheryWest,Carlyle,
Illinois. SAAMem24 (1970) 1-91; A. Hesse, "Tentative Interpretation
of the Surface Distribution of Remains on the Upper Fort of Mirgissa
(Sudanese Nubia)," in F. Hodson, D. G. Kendall, and P. Tautu, eds.,
Mathematicsin the Archaeologicaland Historical Sciences (Edinburgh and Chicago 1971) 436-444. These studies focus on the extent
of the correlation between surface and subsurface patterns of material
at a site and not on the actual mechanisms that are involved as such.
There is also a failure to make a fundamental distinction when it
comes to the subsurface: namely, the plow-zone itself, where material
is in circulation as a consequence of tillage, and deeper parts of the
subsurface, where such plow-induced circulation does not occur.
3. D. Lewarch and M. O'Brien, "Effect of Short Term Tillage on
Aggregate Provenience Surface Pattern," in M. O'Brien and D. Lewarch, eds., Plowzone Archaeology:Contributionsto Theoryand
Technique.VanderbiltUniversityPublicationsin Anthropology27
(1981) 7-49; N. Trubowitz, "The Persistence of Settlement Pattern in
a Cultivated Field," in E. Engelbrecht and D. Grayson, eds., OccasionalPublicationsin NortheasternAnthropology(Department of Anthropology, Franklin Pierce College: Rindge, New Hampshire 1978)
41-66.
4. A. Ammerman and M. Feldman, "Replicated Collection of Site
Surfaces," AmAnt43 (1978) 734-740.
in Calabriawere designedto address.The first was to
providean estimateof the ratiobetweenthe materialon
the surface and that in the plow-zone and also to see
how this ratio behaved over time. Knowledgeof this
ratiowould be of considerableimmediatevalue in interpretingthe surfacecollections made at sites duringthe
course of our survey in the Acconia area of Calabria.5
The second questionconcernedthe natureof the sampling process as mentionedabove. It was of broadinterestto know, for example,if surfacematerialcouldbe
regardedas a realizationof a Poissonsamplingprocess,
the simplestmodel to workwith in mathematicalterms.
Under a Poisson model, each piece of a given size
circulatingin the plow-zonewouldbe consideredto have
a fixed low probabilityof makingits appearanceon the
surfaceand to comprisean independentsamplingtrial.6
Withoutgoing into a more formal accounthere, what
this would mean in archaeologicaltermsis that surface
material,as a sample,wouldforma "fair"representation
of the populationof remainscirculatingin the plowzone. Deviationsfroma Poissonmodelwouldmeanthat
the relationshipbetweensurfacematerialandplow-zone
materialwould be a more complicatedone, and the
interpretationof surface collections would accordingly
become a more complex matter.For the archaeologist,
it would be favorableif a Poisson model turnedout to
The thirdquesprovidea reasonablefirstapproximation.
tion (or set of questions)concernsthe lateraldisplacement of pieces as a result of plowing. How far on
average do pieces in the plow-zone move? Does displacementtendto increaseas a field is subjectedto more
bouts of plowing? Such considerationsare of obvious
interestto those who do surveys, since one would like
5. A. Ammerman and G. Shaffer, "Neolithic Settlement Patterns in
Calabria," CA 22 (1981) 430-432; A. Ammerman, The Acconia
Survey:Neolithic Settlementand the Obsidian Trade (Institute of
Archaeology: London in press). Support for the fieldwork in Calabria
was provided by grants from the National Science Foundation (BNS
76-15095 and BNS 79-06187). A preliminary version of this report
was given as a paper as part of the "Symposium on Plowzone Archaeology: Tillage Process and Archaeological Material" at the Annual Meeting of the Society for American Archaeology held at San
Diego in 1981.
6. For an introduction to the Poisson distribution, see R. Sokal and
R. Rohlf, Biometry(San Francisco 1981) 81-93; for a brief account,
see also J. Doran and F. Hodson, Mathematicsand Computersin
Archaeology(Cambridge, Massachusetts 1975) 47-48. The tiles visible on the surface at any one time can be thought of as the positive
realizations of sampling trials in which each tile has a given probability
p of appearing on the surface as an event. The chance of a tile
remaining in the plow-zone (the event of not appearing on the surface), using conventional notation for probabilities, would be q, where
p + q = 1. An estimate of the value of p, at least as a first approximation, is given by the surface-to-plow-zone ratio which is discussed
below in the text.
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
tournalof Field Archaeology/Vol.12, 1985 35
to use what is seen on the surfaceto make inferences
aboutsuch thingsas the size of sites andspatialpatterns
withinthem. The fourthquestionis relatedto the breakage of pieces and their edge-damageas a consequence
of tillage. Ideally, it would be useful to know the rates
of breakagethathold for a given class of materialsuch
as potteryunder a given set of conditions.If rates of
breakageare relativelyhigh, the populationof material
in the plow-zone is progressivelybeing increasedin
numberand at the same time being transforrnedinto
smallerpieces. If breakageratesare quitelow, the population tends to remainmuch the same over time and
the archaeologistis in a more favorablesituation.
An importantpart of the experimentaldesign is the
use of what are knownas "replications,"
or the systematic repetitionof test rows. They are a commonfeature
of the design of experimentsin agricultureand biology.
By holdingotherfactorsthe samebetweenthe test rows,
variabilityinherentin the process of interest can be
examined.In the design to be describedbelow, the four
rows of tiles with a N-S orientationmake it possible to
documentthe level of stochasticvariationin the surfaceto-plow-zonesamplingprocess. Anothermajorfeature
of the design is the use of observationsmade over a
series of differentpointsin time. Such sequentialobservations make it possible to monitorover time the dynamic behavior of patternsand processes. This is of
particularimportancein the case of questionsin archaeology that pertainto the plow-zone, since plowing is
often a regularlyrepeatedpracticeandone thathas been
engagedin for many years in most partsof the world.
Design of the Experiments
It is best to begin with the choice of the materialto
be placedin the ground.The decisionwas madethatthe
pieces to be used shouldall have the same size, shape,
andcolor. Fromearlierworkon the replicatedcollection
of surfacematerial,it was knownthatlargerpieces seem
to have a betterchanceof occurringon the surfacethan
smallerones.7 This fact stems in partfrom the way in
whicha piece is recognizedas beingon the surface;i.e.,
wheneverany partof it is visibleon the surface.It would
7. Ammerman and Feldman, op. cit. (in note 4) 736-739. For a
review of the literatureon the size question, see Lewarch and O'Brien,
op. cit. (in note 1) 307. Some of the best evidence currently available
on the size factor comes from the short-term experiment conducted
by Lewarch and O'Brien, op. cit. (in note 3) 8-17. Their data-can,
in fact, be reworked to show the following. After three equipment
passes and considering only Pattern l (the richest one in material),
the ratios of surface to plow-zone material for their three size classes
are respectively: 0.111 for pieces larger than 1 inch, 0.079 for pieces
falling between 2 and 1 inch, and 0.050 for pieces less than 2 inch.
A progressive decline with size is seen in the ratio here.
be more correctin terms of what one actuallysees to
describematerialas being "in"the surface.8Thereis a
greaterchanceon averagefor some partof a largepiece
to breakthe surfacethan a small one. Ratherthandeal
with the complexrelationshipsbetweensize andsurface
visibilityin the small-scaleexperimentsthatwere being
planned, we decided that size would be held constant
and attentionwould be paid to the four questionsmentioned above. Small ceramictiles that measure2.5 cm
x 2.5 cm x 0.5 cm were eventuallyselected. They are
somewhat smaller in size than the average Neolithic
sherd recoveredfrom the land surface at Acconia and
yet largerthan most of the pieces of obsidianfound at
sites in the area. The yellowish grey color of the tiles
correspondswell with thatof the soil incrustations(when
seen in a dry state) that form on the surfaceof artifacts
in the dunesoils at Acconia. The tiles also have a glassy
andbrittlecharacterin line with whatis requiredfor the
studiesof breakageand edge damage.The tiles appear
to be morebrittlethanpieces of obsidianandtheiredges
also damagemore easily in the ground.In the design,
one groupof tiles wouldbe collectedandset asideduring
the second year of the experiment(1979) and a second
group during the fourth year (1981), so that rates of
edge-damageover time could be examined.One disadvantage of the tiles is that numberswritten on them
(whetherin ink or in paint)did not always surviveafter
a numberof yearsin the soil. This meantthatthe movementof tiles had to be studiedin termsof theirdisplacementfromthe lines in whichthey wereoriginallyplaced.
In all, 1,000 tiles were used in the experiments.
The area chosen for conductingthe experimentswas
part of a known Neolithic site in the dune area at Acconia.9The olive trees grown at the site today, which
are plantedon a regulargrid system, provideda convenient and permanentguide to set out rows of tiles.
Figure 1 shows the layout of the six lines that were
employed:note the patternof alternationof the N-S and
E-W lines. For each of the fourN-S rows (lines 3 through
6), 125 tiles were placed in the ground.For each of the
two E-W rows (lines 1 and 2), 250 tiles were used. The
tiles were spaced20 cm apartalong a line andplacedat
a depthof ca. 2 cm below the modernland surface.By
settingout the two groupsof "replications"
at rightangles to one another,it would be possible to measure
8. To give an example, the following was observedon the fourth
date:one-thirdor less of a tile was visible for 41% of the tiles; onethirdto two-thirdsfor 11%of the tiles; and two-thirdsor more for
48% of the tiles.
9. The site is number68 in the Acconia Survey;it is the same one
where replicatedcollections were also previouslycarriedout; see
Ammermanand Feldman,op. cit. (in note 4) 737-738.
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
X
X
O*-***X
X
X
X
X**
.
*X***-*X
.......
X
36 Plow-ZoneExperimentsin Calabria,ItalylAmmerman
N
X
X
X
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
06
02
05
04
X
ol
03
Figurel . The layoutof the six lines of tiles. The x's representolive
trees,whicharespacedca. 14 m apart.
displacementboth in an E-W direction(using the rows
with 125 tiles) and in a N-S direction(using the rows
with 250 tiles). One of the reasonsfor choosingthe site
is thatit has a gradualslope fromnorthto southof about
1 in 10. The relief is essentiallyflat in an E-W direction.
This situationmade it possible to monitorthe effects of
slope on displacement.
The parentmaterialof the soil in the site area is a
fairly coarse sand of aeolian origin. Soil profilesshow
well developedsoil formationto a depthof morethana
meter.Accordingto the classificationsystemof the Soil
Survey Staff, the soil would be describedas an entic
haploxerall.l0The plowingdone at the site, whichmight
betterbe describedas harrowing,is quiteshallow,ranging 15-20 cm in depth. It is done in the form of a
continuousloop for the olive grove as a whole two or
three times each year: in the early spring, so that soil
moistureis not lost throughtranspiration
of groundvegetation;in the late spring or early summer,in orderto
avoid fireswhenthe groundcover is dry;andin autumn,
in preparationfor harvestingthe olive trees.The plowing
does not seem to be done in any consistentdirection,
althoughit is constrainedby the gridof olive treesto be
done in one of four possible directions(i.e., N-S, S-N,
E-W, or W-E). There were at least three and perhapsas
many as six episodes of plowing priorto the firstset of
observationsin 1979." Since we were not in Calabria
throughouteach year, we were not able to make direct
10. Soil Survey Staff, Soil Taxonomy. A Basic System of Soil ClassiJVcation
for Making and Interpreting Soil Surveys (Washington
D.C.
1975). The soil descriptionsat Acconia were made by A. Remmelzwaal of the Instituteof PhysicalGeographyand Soil Science, Universityof Amsterdam.
l l. One strategythat has been adoptedin plow-zoneexperimentsis
for the episodes of plowing to be controlledby those doing the
experiment;see Lewarchand O'Brien, op. cit. (in note 3) 8-12.
While this control can be accomplishedin an experimentof short
duration,it is not feasible for a multi-yearstudy.We were interested
in seeing the patternsthat would result after numerousepisodes of
"natural"
plowing at the site. One distinctadvantageof havingmany
plowing episodes is that the patternseventuallyobserved are less
likely to be contingentupon the way in which the pieces are set out
(i.e., on the surfaceor insertedin the plow-zone)at the startof the
experiment.
observationson all episodesof plowingover a periodof
four years. But from the orientationsof furrowsthat
were observed on several dates, it would appearthat
plowingwas done primarilyin the directionsof E-W and
W-E from 1977 through1980. It was done at least once
in a N-S directionin 1981, andthis plowingepisodehad
a majoreffect on the displacernentof tiles, as we shall
see below.
The procedureused for making observationsin the
field was to have two or three people systematically
coverthe areasurroundinga given line. Eachtile visible
on the surface was flagged and left in position. The
followinginformationwas recordedfor each flaggedtile:
(a) its displacementfromthe line whereit started(measured at right angles to the line); (b) its position along
the line; (c) whetherit was whole or broken;and(d) the
proportionof the tile actually visible on the surface.
Exceptfor the two dateswhencertaintiles wereremoved
for the edge-damageexperiments,the tiles, afterbeing
examinedfor breakage,were placedback in the ground
in theirencounteredpositions.Notes were also takenon
the conditionsof the surfaceas partof each visit.
Results of the Experiments
Observationswere madeat the site on fourmaindates:
(1) March21, 1979, (2) July 1, 1979, (3) April7, 1980,
and (4) June4, 1980. All six rows were recordedon the
firstand thirdvisits, while only the two E-W rows were
examinedon the othertwo dates. The conditionsof the
surface were best duringthe first visit and reasonably
good for the fourth date. They were only fair for the
second and third dates. In addition, a fifth visit was
madeon August 1, 1981, in orderto collect tiles for the
edge-damagestudy and a sixth visit was made on June
3, 1983. While groundcover at the site preventedsystematiccoverageon the lattertwo dates, it was possible
to make valuableobservationsof a morequalitativenature on displacement.Few brokentiles were observed
on the surfaceduringany of the visits. In fact, broken
tiles were encounteredonly on three dates and in each
case only a single tile was involved. On the first date,
therewas one brokentile out of 69 visibleon the surface.
On the thirddate, there was one out of 39 tiles. Even
by the fifthdatewhenthe areahadprobablybeenplowed
at least 10 times, only one tile out of the 29 observed
on the surfaceturnedout to be broken.This is a heartening result which suggests that the populationof tiles
in the plow-zone is not being substantiallytransformed
at least over the shortrun.
The resultson the surface-to-plow-zone
ratioare presentedin Table 1.12 It variesto some extentbetweenthe
12. For the estimationof a ratio from samples, see W. Cochran,
SamplingTechniques(New York 1963) 29-33. The best estimatorof
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
Journal of Field ArchaeologylVol. 12, 1985
Table 1. Estimation of the ratio of surface to plow-zone
material for four dates.
Date
First
Second
Third
Fourth
Combined
37
Table2. Fluctuationsin the surfacecountsof the four N-S
rows for two dates.
Surface
Plow-Zone
Ratio
1:X
Date
03
04
05
06
m
V
Vlm
69
20
39
27
155
931
441
892
434
2698
0.074
0.045
0.044
0.062
0.057
1:14
1:22
1:23
1:16
1:18
First
Third
6
6
5
5
7
3
12
6
7.5
5.O
7.25
1.50
0.97
O.30
visits, rangingfrom 1:14for the firstdateto 1:23for the
third date. The values of the ratio seem to reflect the
differences in surface conditions at the times of the
respective visits. The counts of surface tiles are high
enoughfor the firstandthirddatesso thatthe differences
in the ratioshere are not likely to be the resultssimply
of stochasticsamplingeffects. Thereis no clearevidence
in Table 1 for time-dependenttrendseitherof increase
or decrease in the ratio. The values for the first and
fourthdates, allowingfor slightlyless favorablesurface
conditionson the latterdate, are much the same. Apparently,the episodes of plowing carriedout between
the startof the experimentin the springof 1977 andthe
first date of observationtwo years later (i.e., at least
threeepisodes of plowing and perhapsas many as six)
were enough so that equilibriumis alreadybeing approachedwith regardto this ratio.'3As a centraltendency or averagevalue for the firstfour dates, the ratio
would be 1:18. This figure representsthe relationship
betweensurfaceand plow-zonematerialunderordinary
conditionsfor fields withoutgroundcover at the site. It
shouldbe stressedthatit appliesonly to a class of given
size. Archaeologicalpieces smallerin size thanthe tiles
wouldbe expectedto have lowervaluesfor the ratioand
largerpieces the opposite.14 It is evidentthateven for a
fairly large-sizedclass such as the tiles, only a small
fractionof the materialin the plow-zone is visible on
the surfaceat any one time. Conversely,a ratioof 1:18
would mean that if one finds six tiles on the surfacein
one partof the site, it would be reasonableto inferthat
this ratio for any one visit is supplied by the total of the surface tiles
for the lines examined on that date divided by the total number of
tiles in the ground for those lines (i.e., the tiles originally placed in
the ground along the line less any previously taken out of circulation,
such as those for the edge-damage studies, minus those now visible
on the surface).
13. In the literature on agricultural engineering, equilibrium is considered to be reached after 10-15 equipment passes; Lewarch and
O'Brien, op. cit. (in note 3) 12.
14. For a mathematical formulation in terms of multinomial sampling, see Ammerman and Feldman, op. cit. (in note 4) 738-739; see
also note 7.
on the orderof 100 tiles occur in the plow-zonethere.
In such a case, light scattersof surfacematerialwarrant
our closer attention.
The countsof tiles seen on the surfaceof each of the
four N-S rows are presentedin Table2. Eachreplication
beginswith the samenumberof tiles laidout in the same
configuration.Througha given date, each is subjected
to the same episodes of plowing. Observationson a
given date are also being made under essentially the
same surfaceconditions,so that the variationsbetween
the replicationsshould representfluctuationsthat are
inherentin the mechanicsof tile circulationin the plowzone. The counts range between S and 12 for the first
date, and between 3 and 6 for the thirddate. On both
dates, the largestcount is abouttwice thatof the smallest. Fluctuationsof the kind that we would expect to
encounterif the surfaceoperatesas a samplingprocess
are observed.In orderto see whetheror not the counts
conformto a Poissonmodel, the variance/mean
ratiohas
also been calculatedin Table2 for the two dates.'5 The
value of the ratiofor the firstdate seems to be in accord
with a Poisson model;the ratiofor the thirddate looks
too small. While the resultscannotbe regardedas conclusive, they do not indicate and this is perhapsthe
most importantthing at the presenttime thata Poisson
model has to be rejected.
A summaryof the resultson displacementfor the first
four dates is given in Table 3 for the N-S lines and in
Table4 for the E-W lines. The thirdcolumnin Table 3
indicatesthe percentageof tiles thathave moved to the
west of their startinglines. There is an overallbalance
betweenthe two directions(i.e., east andwest) for both
the firstand thirddates. The fourthand fifthcolumnsin
Table 3 give the rangeof values observedor the maximumdistancesmoved to the west andeast respectively.
For the first date, this distanceusually falls between2
m and 4 m in each direction.Valuesfor the rangehave
in most cases become largerby the thirddate. Thereare
two differentways by which we can calculatea mean
of the displacementfrom a line. The first is to look at
the distancesfrom the line, where tiles on one side of
the line have positive values (the west side here) and
thoseon the otherside have negativevalues. The second
15. Sokal and Rohlf, op. cit. (in note 6) 88-90.
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
38 Plow-ZoneExperimentsin Calabria,ItalylAmmerman
Table3. Displacementvalues for
the N-S lines on two dates (distances
are in m).
First Date
Line
03
04
05
06
Comb
No.
%w
6
5
7
12
30
33
80
43
42
46
Max. w
2.83
2.44
3.75
3.50
3.75
Max.
E
2.04
1.02
2.29
2.54
2.54
m
S.D.
Im|
|S.D.|
-0.14
+0.77
+0.53
+0.38
+0.38
1.74
1.28
1.80
1.65
1.68
1.35
1.18
1.74
1.23
1.36
0.94
0.81
1.21
1.11
1.03
+0.97
+0.14
-2.43
+3.16
+0.91
1.95
1.76
1.77
2.61
2.70
1.95
1.17
2.43
3.16
2.19
0.62
1.18
1.77
2.61
1.76
Third Date
03
04
05
06
Comb
Table4. Displacementvalues for
the E-W lines on four dates
(distancesare in m).
6
5
3
6
20
67
40
0
100
60
2.58
3.14
6.65
6.65
1.55
1.36
3.39
3.39
First Date
Line
01
02
Comb
No.
24
15
39
%s
96
67
84
Max. s
2.47
3.09
3.09
Max.
N
0.70
5.58
5.58
m
S.D.
Im|
|S.D.|
+0.77
+0.06
+0.49
0.72
0.97
1.36
0.86
1.33
1.04
0.60
1.41
1.00
+0.91
+0.40
+0.63
0.52
1.09
0.91
0.91
0.91
0.91
0.52
0 66
0.59
+ 1.41
+0.77
+ 1.12
2.72
2.20
2.56
1.66
1.93
1.74
2.56
1.05
2.17
+0.82
+0.72
+0.74
0.78
1.16
1.00
0.94
0.88
0.91
0.62
1.03
0.84
Second Date
Comb
9
11
20
100
73
85
1.80
2.00
2.00
01
02
13
6
Comb
19
77
83
79
9.80
2.43
9.80
01
02
13
14
27
77
71
74
2.31
4.05
4.05
01
02
1.98
1.98
Third Date
1.35
3.47
3.47
Fourth Date
Comb
approachis to look at the distancesmoved irrespective
of sign or in absoluteterrns.The two means are given
respectivelyin columns six and eight of Table 3 and
theirassociatedstandarddeviationsin adjacentcolumns.
For the first date, the values of m show a slight shift to
the west with the combinedvaluefor the fourrowsbeing
less than 0.5 m. The mean distancethat an individual
tile moved in an E-W direction(i.e., |m|) variesbetween
1.18 m and 1.74 m among the replicationsand has a
combinedvalue of 1.36 m. Whena comparisonis made
with the thirddate, the mean values calculatedin both
ways show a more or less consistentpatternof increase
over time. The averagedistancethat a tile moved in an
E-W directionhas now reached2.19 m. While displacement as observed here cannot be described as pronounced, and few tiles have moved as much as 5 m
0.52
0.55
0.55
away from their startinglines, manytiles are no longer
within2 m of theirstartingpositions.
The samebasic formatis used in presentingthe results
on displacementfor the two E-W lines in Table 4. A
clear tendencyfor tiles to move to the south or downslope is seen. The percentagesfor the two earliestdates
are particularlyhigh. But counter to what we might
expect, they decline somewhaton the fourthdate. An
explanationhere mightbe an episodeof plowingin a sN direction, or upslope, at some time just before the
fourth date. The range values show a considerable
amountof variationfrom one date to the next. Thereis
one case where a tile moved 9.80 m from its starting
line. But againin overallterms,few tiles movedas much
as S m in a N-S direction.The values for the mearvm
indicatea progressivedownslopeshift throughthe first
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
Journalof Field ArchaeologylVol.12, 1985 39
threedates. The values of |m| are more or less comparable on the firstand thirddates with those observedfor
the N-S lines in Table3 16Againthe valuesfor the fourth
date are smallerthan those for the thirddate, perhaps
for the reasonmentionedabove. In general,thereis the
impressionthat slope has some effect on displacement
but not a major one. The shift downslope after three
years and perhapsas many as nine episodesof plowing
is only on the orderof 1 m. Observationsmade on the
fifth and sixth visits to the site, however, revealed a
differentpicture.Out of a total of 29 tiles visible on the
surfacein the summerof 1981, nine had moved at least
15 m south of their startinglines and another15 tiles
fell in the rangeof 5-15 m to the south.A majorchange
in patternsof displacementhadapparentlyoccurredsince
the fourthdate. Furrowsat the time of the fifthvisit had
a N-S orientationand presumablyone or more episodes
of plowing had recentlybeen done in a downslopedirection. It is worth recalling that previously furrows
usuallyhadan E-W orientationin the area.Duringa brief
sixth visit to the site, all seven of the tiles visible on the
surfacewere at least 15 m to the south of the 01 line.
The tiles observedon the latterdate were not the same
ones thathad been seen on the fifth date, since the tiles
on that visit had been removedfor the study of edgedamage. In retrospect,it is likely that no episodes of
downslopeplowing were done throughthe time of the
fourthvisit, and the effects of slope being monitoredup
to this pointwere "passive"ones in the sense thatplowing was being done at right angles to the slope and
perhapseven up it. Whenplowing was eventuallydone
downslopeafterthe fourthdate,the more"active"effects
of slope on displacementwere unleashed.
Discussion
One of the prime aspects of the experimentsis the
small amountsof work effort that they requiredin the
field. Using a crew of threepeople, work at the site on
a given visit was usually completedin less than a day.
At the time thatthe experimentaldesign was developed,
we were in the midstof a busy field season in Calabria
which, as is often the case in archaeology,offeredlittle
time for undertakingnew kindsof workwhose potential
was as yet unknown. It was only possible to conduct
experimentson a modest scale, and yet they were able
16. The distancethata tile has moved from its startingline may be
viewed as a vector representingone dimensionof its displacement.
There are two vectors to be consideredand they are measuredrespectivelyby the two groupsof replicationrows. If one is interested
in the full distance that a tile is on average being moved in two
dimensions, one can think of the two vectors acting togetherto
producewhat is knownas a resultant.
to providethe kindsof informationthatwe needin order
to develop a more generalunderstandingof the nature
of surfacematerialas well as to be in a betterposition
for makinginferencesaboutactualsurfacecollectionsin
the contextof local conditionsin an area. The point to
be made here is that such experimentscan easily be
incorporatedinto most multi-yearsurvey projects. If
experimentsof this kind are carriedout in a range of
differentareasover the next decade, it will be possible
to put surveyarchaeologyon a much sounderfooting.
In termsof our generalview of surfacematerial,the
experimentsreveal in an even clearer.way thanthe previous work on repeatedcollectionsthat the surfaceoperates as a sampling process with respect to material
circulatingin the plow-zone. The exact natureof the
samplingprocess, however,remainsuncertain.Withregardto whatwouldbe the mostfavorablecase, a Poisson
model, the results as seen in Table 2 are mixed. In
retrospect,especially given the small amountsof additional time and effort that would have been requiredin
the field, it is unfortunatethat more replicationrows
(e.g., six or eight N-S lines insteadof only four) were
not employed in the experiments,since they probably
would alreadyhave permittedthis fundamentalquestion
to be answered.It shouldnot be long, however,before
we can specify the essential characterof the sampling
process.
It is evident, as shown by the resultsin Table 1, that
the surface-to-plow-zoneratio varies as surfaceconditions at the site change from one visit to the next. This
changeis somethingthatall thosewho havedone survey
workfor any lengthof time arefamiliarwith intuitively:
more pieces will be seen on the surfaceof a field after
it has rained.17The experimentsillustratein quantitative
terms the order of differencethat can be encountered
even in fields withoutgroundcover or crops. The value
underthe best conditions(1:14) representsalmosttwice
thatobservedunderthe worstconditions(1:23). Perhaps
even more importantis the realizationof how small a
proportionof the material in the plow-zone actually
makes its appearanceon the surface at any one time
under even the best conditions. For every tile on the
surfaceof the site, there are some 15 to 20 other tiles
circulatingwithin the plow-zone. What we are dealing
with on the surfaceis only the tip of an iceberg, and a
stochasticone at that.Wherestochasticeffectswill come
most actively into play is among those categoriesof
artifactsthatare rareor less abundantin the plow-zone.
Due caution is thus called for when it comes to the
of suchcategoriesof material,
analysisandinterpretation
whichoftencomprisethe mostindicativeones in cultural
17. See, e.g., the discussionof surfaceconditionsin Hirth,op. cit.
(in note 1).
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions
40 Plow-ZoneExperimentsin Calabria,ItalylAmmerman
terms. One of the local factorsthat entersinto the surface-to-plow-zoneratiois, of course,the depthof plowing. It is worth commentingthat plowing done at the
site in Calabriais relativelyshallowin comparisonwith
that in many other places in the world. When plowing
is done to a greaterdepth, it is reasonableto expect
correspondinglylower values for the ratio. It will be
possible to state this expectationin quantitativeterms
when a wider corpus of results from experimentsconductedin otherpartsof the world is available.
So far the discussion has concernedonly material
belongingto a single-size class. The real complication
in the study of actual surfaceassemblagesstems from
the fact thatdifferentsize classes of materialwill at any
one time have different surface-to-plow-zoneratios.
Pieces that are largerin size are representedin higher
proportionson the surfaceof a site than smallerones.
Withoutgoing into the mathematicsthat are involved
here, a multinomialformulationprovides a means of
characterizingthe assemblage of pieces belonging to
differentsize classes that is recoveredfrom the surface
of a site.18 In orderto beginto "calibrate"
the size factor,
it will be useful in future experimentsto introducea
second or thirdsize class.19
Theresultswithregardto lateraldisplacementareboth
hearteningin some senses and disturbingin others.The
averagedistancethat a tile moved throughthe firstfour
visits is only a few meters, althoughit does seem to be
graduallyincreasingover time. The point needs to be
madeherethateven such modestdisplacementdistances
have implicationswhen it comes to the estimationof site
sizes (especiallysmallerones) andthe analysisof spatial
patternswithin sites on the basis of surface material.
Observationsmade on the fifth and sixth dates suggest
thatslope as a local factorat a site can influencepatterns
of displacementin a major way. In fact, the spatial
distributionof the tiles observedon the surfaceduring
the fifth and sixth visits is substantiallydifferentfrom
the configurationof the tiles at the startof the experiment. One of the things that is clearly pointedup here
is the need to conductlong-termexperiments.It is likely
that displacementdistancesare substantiallyunderestimatedin short-termplow-zoneexperiments.
In terms of our generaloutlook on surveys, the ex-
perimentsin Calabriasuggest that we need to replace
the ratherstatic view of surfacematerialthat many of
us have traditionallyheld with a much more dynamic
one. The surfaceof a plowed site as seen over time is a
restless place. There has been an increasingawareness
over the last 10 yearsthatwe need to returnto basicsin
surveyarchaeology.20
Whenthis is done, at least in our
own case, we are encouragedto adopt a much more
complex view of surveywork.
20. Ammerman,op. cit. (in note l) 81-83.
AlbertJ. Ammermanreceivedhis Ph.D. from the
Universityof Londonin 1972 and was recentlya
VisitingProfessorin the Instituteof Ecology at the
Universityof Parma, Italy. In additionto directing
seven seasons of the Calabriasurveyhe has directed
excavationsat several sites in Italy includingMonte
Leoni, Piana di Curinga,and Muteradi Oderzo.He is
at present an AdjunctProfessorin the Departmentof
Anthropology,Colgate University,Hamilton,NY
13346.
18. See note 14.
19. See the reworkingof the datafromLewarchandO'Brienin note
7 as an exampleof the kinds of resultsthat mightbe obtained.The
value of the surface-to-plow-zone
ratiofor theirlargestclass (0.111)
is morethantwice thatfor the smallestclass (0.050). Sinceonly three
episodes of tillage were used in their experimentand conditionsof
equilibriummay not have been reached,these resultsshouldbe regardedonly as a roughfirstapproximation
illustratingthe size factor.
This content downloaded from 149.43.104.9 on Mon, 26 May 2014 10:26:45 AM
All use subject to JSTOR Terms and Conditions