-
The texts of the various popers in this yolume were set
individuallyby tlpists under the supervision of each of the authors
concerned.
ISBN 90 6r9r 528 7O 1984 A.A.Balkema, P.O Box 1675, 3000 Bl{
I{otterdanr, NetherlandsDistributed in USA & Canada by: A.A
Balkema l'ublishers, P O.Box 230, Accord, MA 02018
Printed in the Netherlands
-
PROCEEDINGS OF THE SIXTH SYMPOSIUM OF THE INTERNATIONAL WORK
GROUPFOR PALAEOETHNOBOTANY / GRONINGEN / 30 MAY-3 JUNE 1983
Plants andAncient ManS tudie s in palae o ethnobo tanyEdited
byW.VAN ZEIST & V(A.CASPARIEBi o log isch-Archaeologisch
Instituut, St ate Universit y, Groningen
A.A.BALKEMA/ROTTERDAM/ BOSTON/ 1984
-
Interpretation of archaeological plant remains :The application
of ethnographic models from Turkey
CORDON HILLMANL'tritersitl, of London, UK
ABSTRACT: Each step of crop husbandry and grain processing has a
measurableeffect on the composition of crop products and
by-products. These effects havebeen studied in archaic agrarian
systems stil-l surviving in Turkey and are summa-rised in the form
of 'cause-and-effect' models. Patterns of variation in
thecomposition of remains of si.mil-ar crops recovered from
archaelogical sites arefound to closely resembfe those presented in
the models. The use of these modelsto interpret the composj.tion of
individual samples of remains of crops in termsof ancient agrarian
practice is straightforward. However, when large numbers ofsamples
are involved, a series of analytical steps is necessary. Each of
thesesteps is expJ.ained in turn using examples frorn a large
assernblage of crop remainsfrom an excavation in North Wales.
Rational-e
About thirty distinct operations areinvo1ved in growing a crop
and convert-ing it to food for buman consumption.Recent
ethnographic studies of archaicagrarian systems surviving in
thepresent-day indicate that each ofthese operations has a
measurabl-eeffect on the composition of each ofthe major crop
products and by-products.Tbe composition of these crop productsthus
embodies information on the waythe crop was managed in the field
andprocessed back in the settlement.
Sampl-es of charred remains of cropsfrom archaeological sites
commonly ex-hibit a composition closely similar tothat observed in
one or other ofthese present-day crop products. Byreference to
modern equivalents, there-fore, archaeoJ-ogical samples can
pro-vide valuable clues to the husbandrypractices of prehistory
(Dennell 1974;Hilfman, 1973, 19Bl; Jones l9B1). Andif archaelogical
samples of this sorthave been taken from each habitationfeature, it
is possible to study thehorizontal distribution of the variouscrop
processing activities representedby the different sampl-es and, in
somecases, to identify the past function ofthe excavated
structures.
6th Symposium Palaeoethnobotanv / Groningen / 1 983 1
mhi -
.++6mh|c f i rcf 'l r, +^lrrln PoIJgr qLLErlrPur
!ll-Lrjsumrnarise some ethnographic models ofarchaic agrarian
practice in present-day Turkey. It secondly outlines theanalytical
steps by which the majorclasses of crop product (as identifiedfrom
the ethnographic models) can berecognized in complex qssemblages
ofcharred rernains from archaeologicalsites. The paper thus
explores one ofthe methods by which the modern modelcan be used to
identify evidence ofspecific agrarian practices in archaeo-logical
remains.
Alternative (non-ethnographic) methodsof interpretation
It must briefly be ltressed tfrat, inworking from explicitly
defined ethno-graphic models, tbe interpretive methodoutl-ined bere
and in the parallel paperby Glynis Jones is fundamentally
dif-ferent from thab explored by RobinDennel-l- (1972, 1974, 1976).
InDennell's approach, variations in thecomposition
-
of plant remains wereinterpreted principally by reference
toassurned pddt functions of the sitecontexts with wbich the
remains wereassociated, and, in certain cases, bycomparison with
grain size distribu-
-
Zwo of theaP6taLedspikeleEs
ls seg6ent!ach
,'(44' (i)../),,,",^"v 'z " V/ b/ "spiterer \M
-
f.,
.:," --:Ez-"'- V {z sr*. a^,".
part of ear showlng
FruE-THRESIIING llHilT e.q. aREAD WHEATwhen Ehreshcd ille ear
imediately breaks up 6-ROWED, HUf,IED ilRlEY
lnto: -
l| lroo one
/ rachls node
rachls segs.
-
tions generated in laboratory sievingexperiments. Thus, in
working directlyfrom the ancient remains themselves,Dennell
approached interpretation fromprecisely the opposite direction to
theethnographic approach outfined belowand in the following paper
by clynisJones. It should similarly be stressed,therefore, that his
subsequent recon-structions of archaic Arain-processingsequences
(eg. in Dennell- I976) owedmuch to speculation (as he himself
hascl-r6cc6/l\ :nA ralr|-irralrr 'lilFl6 F^e thnograph ic
observation.
1 ETHNOGRAPHIC MODELS FROM TURKEYOnIy a brief outline of the
availablemodefs from Turkey will be offeredhere, as they have
already been pub-Iished eJsewhere (in HilIman, 1981 andforthcoming
a. )
I.1 Field nethodsThe first step in assembling the ethno-graphic
models was to focate villagesin remote, generally mountainous
areasof Turkey where archaic forms of indi-genous crops were still
grohrn bypeopLes whose agrarian technology owednothing to the 2oth
century andappeared, indeed, to have remainedunchanged from
technologies availablei.n the same areas three or morethousand
years ago. My studies ofarchaic agriculture began, in fact, inthe
village of Agvan in EasternAnatol-ia in 1969 as part of the
rAFvanPr:oject' of the British Institute- ofArchaeol-ogy at Ankara
(see French etal-. , 1973) .
Having established contact with thealways very hospitable
villages, thefirst step was to list the crops grownand to collect
infornation on the fullrange of systens of crop managementapplied
in the area. For each of thecrops under cultivation the
procedurewas then as follows:
a) Detailed records were made of thefull sequence of husbandry
andprocessing methods applied to each ofthe crops gro$rn. Details
of the looltypes used were also recorded. .r': ':b) SampLes of ca.
2 kg. were takenfrom every crop product and by-productj.n every
processing sequence - fromthreshing onwards. Ho$rever, sampl-ingwas
never straightforward: samples had
to be collected as and when the rele-vant operations were being
undertakenin the various households of the vil-lage or on tbe
various threshing yardsaround the village. In villages whereonly
brief visits were possible, san-pling was inevitably very
'piecemeal',c) Together with each sample, it was
necessary to record details of (i) howthat particular crop had
been manageoin the field (e9. frequency of irriga-tion, whether
weeded, cutting height,etc.), (ii) what processes it had
beenthrough prior to the point at which thesample was collected,
(iii) the classesof village context in which eachprocessing stage
occurred and in whichthe crop product (or by-product) wasstored,
fed to animafs, burned, ortossed straight onto a midden.
d) The samples were next sorted andthe components identified.
Tbisallowed the composition of each productto then be classified
(i) in terms ofthe relative abundance of each speciesof weed seed
and class of chaff/strawresidue, (ii) in terms of any
majordifferences in the frequency distribu-tions of grain and weed
seed sizes.
e) Data from samples representing theequivalent products of the
same cropwere then compared, as hrere data fromsamples representing
different stagesof processing and different systems ofhusbandry in
the field. In this way itlras possible to identify t+re nostobvious
of the effects of each of thedifferent operations (or systems
offield management) on the composition ofthe major crop products
and by-products.
I.2 The resultsIn the course of this work, it quicklybecame
clear that the major operationshad clearly discernible and
consistenteffects on the composition of cropproducts. These effects
are summarised- albeit only in qualitative terms - infigs. 2 &4
(free-threshing cereals)and 3 & 4 (91ume wheats). (FiS.
4outlines the later stages of grainprocessing which are the same
for bothglume wbeats and free-threshing cereals.Fig. 4 therefore
represents the conti-nuation of both figs. 2 and 3). Inthese flow
diagrams, the numbered stepsin the left-hand column represent
theprincipal stages of crop processing.Entries in the right-hand
column listthe principal components of the major
-
'.'
.F 6fk^ r4f4-41 at *ihdat)aff froE lachis of ear)
-
buii oi hdanaaed stEaw;-;,;-.;;;;;;-; -----+i srF.N slpRE |
;;;il;;":i::+ @arsF reeds
.- cdrser stEae rraos-_
_f;;cs-!;;i:-iffiF-l ioE fuer, iodde!+ sone racEses and arrs
L--_- v_ r'Jrv'* Ied coalse te@pe!+ coarse veeds
'p'rnur."."" Ehose points ar vhrchr. nARvEsrINGr a, t . ; .- ,'
I n-a** *3 **'.9 ro fale sd
! . . , I Ehence ro Ehe possiblliry oftv rapw bv @@Linci I
e'eservatron by charringlleals+sarae+reeds
edrs+strae+culrEses+reeds I El = rebiredinsieve
.---2. DRYING - rn Ehe fieLd, tn bdne o! ,!aEeI7)rn ovens/kirns
(F)
'..,r' ,.ri o,.r"
heads and seeds -/
',. t'^'. -o.s-ej) 5. I{TNNOWTNGS. (x2 _ x4) )/a5. WINNOWfNGS
(x2 - x4) ,/'Y sv er ftikke -
--f------ tiehc ctutr + Etu tonser /{chaEr) F
", :::il,',:;3 :";:,""::::.:::--* '"::1"::ZZ';":
reh.s!+lEaF-d*F'"o-ffl iii i!ii:.'"...'
tu|k-Eipa sratn '. I * mre .acEses'+ noscttBlf rcascedt \, I avn
fraqs.|rukrsh 'frlrg'i, -6. tst SteVItlC - wich nediu-coarse
rj.ddle
+ F- (ro leEove conldnanls coarser Ehan 9ra1n){rmediare I
-
consuption) gratn r occastona), racbs tragsJ R-sc EeDininq sLraw
nodesinc r|jn trass:,' yeed seeds |sl
.,_,,:11-y.1:_l\:?..::.!::r6n'cs-sroRE il :::"::.t -" FI rt
-fragrents(Eng.tcavings')l'it
?. 2no srEvrNG -
wlth ,wheat sleve,
I sy er ftikk. light cbtf + ttu longer'.
grara r EaW sE aH nodes + sona
f- ,.. r"r.". ".ntulnanrs rrner rhan plrne glain)I ----
ptire grain + reed seeds of "u* --.+
tell gftin + rcst veeat ror di@Ist as $!re gtain + Eate tacds I
rseeds snalter thn plioe-lrc,EAN!'lcsr SE II | (esp6c.
fowlst,ttawnts' LZl li grain + rcfr . snatl atso fdine f@d
| .achis segnents s aun frags. for h@s, FI or burned I
8. KILN-DRyING - in ueE areae - to avoid
"p"irae. (p)(soEeiEles applied earlier in sequence, qg, p!io!
rohwelrng of balleyt)
(STEP 9 ON$IARDS -
pr.para!1on of graln ploducls for food: exacEly as pe! sEeps Il
- 26
l'To rblt se cooplexrEy or lhe d!49!dr sepalale haryesting of
ears and sEra{ and rts effecEs on cooposrtron have not
beentnco4oraled, though rhey are dtscussed h the texE. 2. The heaw,
basatr rachis se@ents qre disproportronaccry Herr lepresenredrn lhe
P.ioary Ploducts (relarrve to de rlghrer.upp.r -*p.^rs). 3. hny of
rhe rrghEer, upp....9.n., o! broken lachisesale urnnoved our Hilh
rhe frne cnaff. 4. These Evo sels of creil1nga are ofren eargeated
(see EexEr.a Fenton 1978; Granc r96L
. 5' rf Plhe ploducls are slored in pics, chen amual clearslnq
of chese prcs by !!rin' Hiu charcny grain adeling Eo che srdes (see
aeynolds in lhllvolwe). 7..The sequence fo! balley (and oaEsl
drlfers sllghtty _ in thenulled loms - in thaE an erLla s
rone plror Eo siep 5. rcep (HUffiLTNG) rs(aPPried to remve the
reaainhg, basal palt of lheir awns. This rs gener.lry
Fiq. 2 .ir'e sequeEc. c: c!s.rrrons apprrd Eo pulsc clops
- e.9. horse-beans, fietd-peas or veEcns - rs rdenricar tn
-J5: lesPccts' crc-l:i Lhe srve Eesh srzes a! differen! and lhe
reninology ro..nirr rrrclrons rs nor sElrc.lycomgarabre)
.
(reproducd frm RiIIM LgBl)
-
I. HARVESTINGItv lephq bv @@!ing
tlears + strp + Heeds ears + straH + culh Eses r ,eeds
--:y----'2. !3ll!l tr" the field, in barns or -rarety- rn
ovens/ktlns)
'l(-,t^
"/6u .@ts , @tr + dta-be aR .lR" rolbfca
-
ILl. 41" SIEVTNG - wi6 ft trar s-eve s scep 12
Prlne graln + qe& seedsot sae 0 as prine qraln
II
I14. HAND SORTINGsane at leEthlng heavg chatt fra9s. e.g.f!lther
olwe bses lall Dassed .hraLoh s)pwp)qtwe Eses (a11 passed chtauqh
s)eve)t-
Lrr* "^t^
eeeds or s&e I as prtne qra'n + Fanq of che
4 coarse ltbuqh l)qhE) concenants b.oash. IEo surtae bs
aqiaacron (E^s.:'984t ) iSheEl3ni: 'fdd' o! 'aloss'r )
scooFd olt surl.cet- tn vlnnortng Esret. (lipped over edge.
F
b. trrthe. sadfl ueed seeds, .- -
t,atn .
leEtn\Dg lraqs. ot tuavg chati le.s. splkelec j
I Ll.4\rNs JrvE l!9s2 l
cJeil ptine qEain
I
16. WINNOHINC oT 17. OUNXINGI
(htMeled q'a)n ||ith f.eshlg dorstened husls/I5. DE- ING OF BU
NS (i e. hulLed foFs of baltey sd oats o'ly)
- rrEh loosely-ser lolary querns, saddle & .rough
qlelns,rheel 'sclaper-quelns' (Nlkrsh 'seten')' at by poundlnq rn
large w@den(e-q. Ehe SheEland'knockrng scone ; Ttrrkish'dtrek';
cernan '41ac*s.anpf' , eEc- )
-
ROAST
F -
u* *'* ,-* (see rles,z s 3 )(b sepa!.te the freed huEks by Hind
or floatatlon lespecuvely) ,f FN @,ff Srctr{- ,/-,---lubbd
ttillied) qralh
-
h,'*' (ktua-s + paJ-ea')///- , _ - i!"_a5r_ _a_qr_;
!ubbd ltil!ied) qra!hI p!a.J' brleq, etc., sone am ftaqentsP!a.J
Drle9, etc.)tXnockic' 6n L )
! f ddfI
_ cROATS PMPASTION
frB. curu norl:nc+
19. GUIN DRYINCt
20. BN relOVI - sen. stth'scrlpe!-querns'I t"r soDetihes by
Fundins)I - see srep. 15.Y
?errca!ps 21. wINNOWING- (:o separaEe DlanJ| --'\
pee rco qralh -t\ r,.t --
6-=;:flI
23 . GRIN CUCKING
N & TOASTED GROATS ProPAMTION
24. GMIN RoASTING (on hot stone or @eEar - ilP,f+-6;)I -'\\
trtadsLed gratns --- r*brrneo grarns
\ tscots 'ornrsEels' r) -,o) or""", plcked ou! and discalded
(e.g. as Ehe Tulkrsh 'lavu.nag'or che A!$rc
'9a!i9e' )
'by ioose que!nrr,g or Dy poJndrnq -qcn. vith
-....1.''.'.-......r.^.lnlt
21. SIFTING - e!!h 'gloars sieve+ --\
rracxed griln \crusaea grain, ,
ttto!.-t,ee)A (ttouEevt =
'9agla'soup(r ;*1 llur.) or drled (as !.ud balls for saklnq
-
7 ALOUR-MIILING SUQUENcE (j SIFTING) - see Fencon I9?B; Horllz
1957.-
(*- r-.* ia !i+ .) ,/ \lffil r:o"r (+ bEil)-
-
-
-lmmrul
-
28. WTING SEQUENCE - fo! soBe of rhe tladitronal segurces, see
Fenlon 19r8; salzoan 1921.
r
PINAL 5TA
jiEe_li4ql__aijlglae_9ssen!1allv the same for both qlme whears
3nd free-thresh,ri9 cereais,irouoh thev C[ifer ln hulled oralqE
][Lu]l.pd lerleE alE qjtt-q)E lns ?heact
rve ani Lhe .qaked
Fig.4.
lrro HillIH l98l)
-
by-products. The entries inserted bet-ween the numbered
processing stages(left hand column) l-ist the principalcomponents
of the prime products.l"lost of these prine products are
tran-sitory: they are quickly fed into thenext stage of processing.
As a result,the majority of them are unlikely to bepreserved on
archaeo]ogical sites andare of only limited archaelogical
rele-vance. Items in boxes are storage pro-ducts.
It should perhaps be added that thesemodels - one for glume
wheats (figs.3 a4); one for free-threshing cereals andpulses (figs.
2 & 4) - i^,ere distitledfrom a coflection of different
flowdiagrams, each of which summarisedseparately the pattern of
changes inproduct conposition which occurred
-
a) in different geographic areas(principally 2 types: vret- and
dry-sunmer areas);
b) with different methods ofharvesting
- uprooting
- reaping ears onl-y
- reaping ears and strawtogether - etc.,(each of these
harvesting nethods canresult in certain extra components suchas
culm bases being present or absentin certain products);
c) when sheaf burning hras used toeliminate al1 the early
processing
rigures 2, 3 and 4 therefore repre-sent an attenpt to summarise
all thesedifferent sequences in just 2 flowdiagrams. It is for this
reason thatthey appear somewhat cluttered. (Theoriginals from whicb
they were sumna-rised are inevitably less complex).
1.3 RepeatabilityFor any one crop and any one processingsequence
type, there vJas remarkableconsistency in the composition of
theproducts and by-products produced atany one stage of processing.
This wastrue not only when conparing productsfrom different
households in the sameviIlage, but also when comparing pro-ducts
from different villages. Thisapparent consistency is now made
evenmore convincing tbrough Glynis Joneslethnographic studies in
the Aegean(cl-ynis Jones, 1981 and this vo1.): herresul-ts reveal
closely sinilar patternsof variation in product composition
andindicate just the same patterns of
cause and effect. It must neverthelessbe stressed tbat there is
some flexibi-lity in the points in the processingsequence at which
certain operations(e.9. extra rounds of sieving) arecarried out.
These differences seem tobe related principally to the stage
atwhich the grain or spikelets are putinto bulk storage (see
below).
I.4 LIMITS OF GEOGRAPHICAI RELEVANCE OFMODELS
There are, alas, no ethnographic stu-dies of primitive
agriculture in N or WEurope which have analysed the relation-ship
between observed agrarian activi-ties on the one hand and
composition ofproducts on the other. Is it feasible,therefore, to
use these ethnographicmodel-s from Turkey and Greece to inter-pret
crop remains from central or nor-thern European sites in terms of
pastagrarian practice?
a) Crop-type differences between mod-ern Turkey and ancient
Europe?Al-most all- the najor cereals and pulsecrops grown in
Europe in the recent ordistant past are still to be foundunder
cultivation in the Near East,especially in eastern Turkey. For
thepurpose of the ethnograpbic studiesoutlined here, particular
attention wasgiven to the cultivation and processingof Emmer
(T.dicoccum) because it wasthe glume wheat that dominated mucb
ofEurope and S.W. Asia for several mille-nnia. However,
free-threshing cerealsand pulses were more widely cultivatedthan
Eruner; there was, for exanple, noEmmer in the Agvan area. The
resultsfor free-threshing crops are consequen-tly more complete
than they are forErnmer for which the results must stillbe regarded
as somewhat provisional,despite the passage of l0 years sincetheir
collection. (Further vrork onTurkish Emmer cuftivators is now
inhand).
b) Weed flora differences betweenTurkey and Europe?A more
obvious objection is that theweed floras of Turkey are
differentfrom (and certainly richer in speciesthan) those of
northern Europe. In themodels outfined here and by GlynisJones
(this voI. ) , this problem hasbeen pre-empted: as indicated in
Hillrnan(198I
- figs. 5, 6 & 7) the different
crop products and by-products are not
-
distinguished on the basis of particu-1ar weed species being
present or ab-sent, but purely on the sizes anddensities of the
weed seeds present andthe height of the weeds when growing inthe
fiel-d. The actual species repre-sented in each size or height
class islargely irrelevant, therefore.
c) Differences in agrarian technologyDespite their never having
recordedthe composition of crop products, ethno-agricultural
studies in Europe do offera wealth of information on
traditionalagrarian practices and tool assemblagesin recent tines.
(Exanples includeIvlaurizio (L9271 , Leser (1931), and aweal-th of
articles in journals such asTools and Tillage, Agartorteneti
Szemle(Hungary) , and Beal-oideas: Journal ofFolklore of Ireland
Society). Fromthese and other accounts, it is clearthat the
techniques applied in recenttimes in central and northern Europe
tothe cultivation and processing of anyone crop species differed
remarkably1ittle in principle from the techniquesused today in
parts of Turkey. Thereason for this uniformity is simpleenough: in
the absence of modern tech-nology there are very few ways of
doingany one of the jobs involved in growingand processing any
particular crop.For example, to de-husk grains of hul--Ied barley,
most of the recorded groupsfrom the ShetLands to the Khyber
Passseem to use some form of pestif andmortar. (For the Shetlands
exarnple,see Fenton, 1978). Adnittedly, occa-sional Anatolian
households use eithera widely-set rotary quern or even aheavy
'seten' for the same job, butthis practi-ce seems unconmon.
Effect-ive alternatives are clearly rare.Striking differences do,
neverthelessexist in threshing methods, in theimplements used for
winnowing, and inthe size and shape of pestils and mor-tars used
for dehusking grain or break-ing-up spikelets. (Details are givenin
Hillman, forthcoming a ). Thresh-ing, for exanple, is effected by
usinganything from small beaters and flailsto threshing sledges and
trampling ofhooves. These differences appear to bebroadly
correlated with the wetness ofsummer: in dry areas, all- the
dustyjobs such as threshing, wi-nnowing andpounding can be done on
a large scal-e(often communally) out-of-doors; in awet area they
cannot, though cul-tural
intrusion can clearly create anomol-ieshere. For tbe purpose of
gatheringinformation for these models, a conspi-cuous advantage of
working in Turkeywas that one of its major areas ofEmmer
cultivation in which agrariantechnology retains its archaic form
alsohas hret surilners. This area enbracesthe eastern end of the
Pontus Mountainsoverlooking the Black Sea where annualrainfall
exceeds 75Omm. Within Turkey,therefore, it was possble to study
bothwet- and dry-summer adaptations of tra-ditional- crop
busbandry.
Hor.vever, the dif f erent processingmethods of wet and dry
areas have onlyminimal effects on the composition ofnrndrrcfs- This
IaSt faCt is bOth anadvantage and disadvantage: it allowsthe one
model- to be used in both wetand dry areas, but it generally
prec-ludes the possibility of differences incomposition being used
to identifywhether it was the dry-type or the wet-type system that
was used.
The one major difference in proces-sing sequences that seems to
distin-guish areas v{ith wet and dry summers(within Turkey, at
l-east) is the pointin the sequence at which the grain ofglume
wheats is put into bulk storage:in dry areas, the Emmer is
threshed,winnowed, pounded (i.e. dehusked), re-winnowedf and the
freed grain partiallycleaned - all in bu1k, out-of-doorsduring the
summer. It is therefore thepartially cleaned grain vlhich is
putinto bulk storage. In areas with wetsummers, however, tbe crop
is processedin bulk only up to the stage of spike-let cleaning, and
it is the spikeletswhich are put into storage (see fig.3).Indeed,
in some households most of thecrop is bulk-stored as sheaves as,
forexample, in parts of Scandinavia. A11the remaining steps of
processing arethen completed on a smal-l scale, day-by-day, as and
when grain is needed toprepare food for immediate consumption.
Despite these differences, the over-all sequences of operations
applied inboth wet and dry areas is basically thesame, and the
compositon of the cropproducts in either area also appear tobe no
different. The principal con-trast between grain processing in
wetand dry areas is therefore to be foundin the contents of grain
stores and inthe tools, which are sornetines of quitediffprent fvnc
(:s i.n the case of thre-shing equipnent) or much smafl-er insize
(as in the case of the pestils and
-
mortars used for pounding and dehusk-ing).
Thus, v/henever charred remains ofgrain stores are recovered, it
is imme-diately possible to identify which ofthe grain storage
patterns was in use,and this, in turn could perhaps suggestwhich
of, sdyr the threshing tool- trad-itions might have been
represented.Far from limiting the application ofthe mode1,
therefore, these differencesin storage practice inprove its
resolu-tion. (Archaeological examples aboundFnr hrrlk qf^ra^6 nF
hnl-h nr:in :ndspikelets and will not be listed here).
However, it must be stressed that thepattern of storage
practices of vret-and dry-summer areas is distorted by atleast two
factors:(i) the first factor is cultural intru-sion. It seems
probable that at leastsone farmers migrating fron wet areasinto dry
areas temporariJ-y retainedtheir former, indoor-adapted
processingnr^.Fi^ac
^h^ra^l-6r'i ca/l hrr fh6 hrr'l lz-storage of spikelets.
Examination ofthe grain stores of sites representingsoutherly
penetrations of, sdy, earlyIndo-Europeans into the
Mediterraneanzone night therefore al-fo\d a crudemeasure of either
conservatism or flex-ibility in prebistoric agrarian
techno-rogy.(ii) There is a second factor. Thedevelopment of large
farm buildingssuch as barns would have alfowed bul-k-processing of
crops to be undertakenindoors. In even the wettest areas ofoceanic
Europe, therefore, the rise ofmanorial- farming under Roman rule
pres-umably allowed glurne wheats from themanorial lands to be
bulk-processed'indnarq rich|- rrn f^ l.ho cfada aF!
uqYvgrain-dehusking and cleaning, ready forbulk storage and
trade.
Nevertheless, it seems unlikely thatthis novel system of
bulk-processingwould have been easily imposed on nativefarmers
lacking large buildings. Theagrarian technology of native
farm-steads and Roman manors may, therefore,have differed
dranatically, especiallyin the processing of glume wheats.This
hypothesis tvi1l, however, remainuntestabl-e until excavators start
sho\,r-ing more interest in recovering cropremains from native
farmsteads of theRoman period. Hitherto, most have pre-ferred to
unearth yet more garrisonsand manors and let the life style ofthe
population's najority remain amatter for conjecture.
1.5 LIMITS OF TEMPORAL RELEVANCE OFMODELS
Ethnographic evidence presented in thisvolume by David Harris
suggests thatseveral of the operations involved inharvesting and
processing grain cropsare not unique to agrarian societies:they
were (and in some cases, stillare) an integral part of the
technologyof non-agrarian societies in areaswhere borrowings from
intrusive agrari-an groups appear improbable. It there-fore seems
reasonabLe to suppose thatsuch techniques were incorporated
intoagrarian practice fron its inception.
In principle, tberefore, it seemsthat certain components of our
modelscan contribute to the interpretation ofplant remains from
even the earliestagrarian sites. Whether the same couldbe claimed
for the conpfete crop proce-ssing sequences is open to question:
soFer T hFiro norqnnallrz fall_ ralrr.fanf!g! tto advocate the use
of our models onsites earlier than the Iate ceramicNeolithic. Such
reservations may be un-founded: not only are there few effic-ient
v,ays of effecting any one stage ofprocessng, but there are few
possibili-ties of altering the sequence of theseoperations: for
example, you cannotsieve before you winnow, because thelight chaff
and straw woul-d immediatelyclog the sieves.
Arguments for continuity of agrariantechnology during the past
few millen-nia are further supported by tbe everincreasing number
of samples of plantremains whose composition has proved tobe
rernarkably similar to that ofpresent-day products of the sane
crops.A recent example comes from 3rd centuryAD Wilderspoof in
England (Hillman, inpress). Here, a huge cache of charredchaff of
glume wheats (mainly spelt)exhibited ratios between the
majorcomponents which were precisely thesame as those found today
in the rvrastefraction from the fine-sieving of de-husked grain of
glume wheats indicatedin stage 12 of the flow diagram
- fig.3. (The major components includedspikelet forks and glume
bases, rachisinternodes, tail grains, prime grain,smal-l weed
seeds, and awn fragments).Composition of this type is
entirelydifferent from that of any other cropproduct or by-product
found in thepresent-day, and the implication mustbe that the spelt
crop at Wilderspoolwas harvested and processed by methodsclosely
similar to those outlined in
-
characteristic cotrlPonents
TABB f. COMPOSITION OF THOSE CROP PIIODTJCI'S I'IKELY TO
SURVIVg IN CHARRED RDMAINS FROM ARCHAEOLOGICAL S]TES_ GLUM
WHEA'],S oNl,Y
o
q
-F
P{+i-E-!63 i-!q oboH
fr!:Xa D
oq qi g8i q.c
t
7o
E-t
;nb
d
- = very lare
)O( = s(re)m( = lotsItN in brackets 'l l' = Present onlY
if croP was lEruested bY uPr@ti'ng'smll scircled crosses '@) ' =
its
-
released only as a result of fragrenbtion of otler itre (such as
uholespikelets) n!!gl the latter fEve beencharred '
the slrnbot r=' = 'tie eguivalst (of )"I = 'srmof ..''
(t ?,i\':
2
8.9a
!Io
;p
6opFo
:lBg.e
q!
ER
Tb
I2q
a
o
E
\
-
the glume wheat model in fig.3.Remains of agrarian tools from
sites
such as Neolithic Egolzwil in Switzer-land tell a similar story
(see Wyss,1969, I97I). Here the beautifully pre-served array of
implements include mostof those needed to effect the majorsteps of
processing as practised todayand outlind in figs-3 and 4. Even
agrain sieve appears to be represented(see photograph of item
- catalogue
No.44416 -
described as a 'Tasche' inWyss 197I). The one tool which seemsto
be nissing is a winnowing fork orshovel: perhaps they used baskets
orfans instead, though it should be not-ed that sieves, too, can
serve as tbas-ketsr for small scale winnowing (Hi11-man,
fortbcoming a). Such a suite ofperfectly preserved processing
toolsare, of course, unique in the EuropeanNeofithic. However,
their rarity inremains recovered fron Neolithic wet-land sites
where wood can theoreticallybe preserved by waterlogging is
clearlynot evidence that sucb tools were rare-1y used at that time-
After all, woodentilling implements are equalfy rare,and yet they,
too, must have been ingeneral use.
An interesting exception to temporalcontinuity in grain storage
systems isprovided by the charred grain fromAssiros Toumba and
Iofkos reported byGlynis Jones (198I, 1982 and forth-coming). Her
work reveaLs that theglume wheats at these two Greek sites(where
the summers were presumably dry)were bulk-stored as spikelets.
Thisis the pattern that today is more typi-cal of areas with wet
sununers. Atthese sites, therefore, we must con-clude either (i)
that storage practiceswere of a tradition intrusive fromareas with
wetter summers, or (ii) thatthis particular feature of crop
proces-sing practice has changed during thepast few miflennia.
Neither possibili-ty would be suprising.
2. REDUCING THE COMPLEXITY OF THEMODELS
2.1 Deletion of superfluous variablesSuperficially, the models
summarised infigs. 2, 3 and 4 may seen somewhat com-plex. However,
on most of those arch-aeological sites where plant materialis
preserved merely as a result ofhaving been charred, only a very few
ofthe crop products and by-products are
represented. For the purpose of inter-preting tbe plant remains
from theaverage site, therefore, much of thedetail can be
eliminated. This simpli-fication involves two steps:
a) Remove aII of those products thatare unlikely ever to be
exposed to fireand thereby preserved by charring. Infigures 2, 3
and 4, the points in theprocessing sequences at v,rhich the
pro-ducts are exposed to fire are indicatedby 'F'. We can therefore
eliminate aIIproducts not marked with an rFr, inc-luding all the
transitory prime pro-ducts which exist for only a shortperiod
before being fed into the nextprocessj.ng stage.
b) Within the products that remain,we need consider only those
ccrnponentswhich, when exposed to fire, are small-enough and dense
enough to drop intothe ashes and be charred rather thanbeing burned
to ash themselves. Onthis basis, hre can eliminate alf
strawinternodes, most of the lighter strawnodes, Ieaf fragments,
all- the lightchaff (i.e lemmas, pafeas, glume tops,the lighter
rachis segment.s and most ofthe awns) together with most of
theIighter weed seeds. The sort of $reedseeds eliminated in the
course of burn-ing are those from genera such as FiIa-9a, Sa1ix,
Calamagrostis and Imperatawhich, because of their attachment to
afeathery pappus (Filago) or to floretswith rachil-la hairs
(Imperata) , areunlikely to be able to drop into theashes but
instead renain high in thefire and get burnt away. However
suchseeds rarely get onto a fire anyhow.(Note: Among the chaff
fractions whichcorunonly survive [generally as compo-nents of
coarse or fine cleanings] aredense segments of oat ar^rn and the
densebasal parts of the rachises of barleyand naked wheats).
The two sets of eliminations (a and babove) now leave us with a
maxinun ofeight products which we are likely toencounter on
archaeological sites. TheeJ-ininations have also greatly reducedthe
range of components in each pro-duct. This reduced range of
productsand their characteristic components aretabulated in table
l. Of the eightproducts listed, the first two (sheavesand straw
waste respectively) are rare-ly represented on the average site.The
same is true of rcleanings fromhand-sortingr (the 6th product
listedin table I), The range of charredproducts like1y to be
encountered onmost sites is therefore verv restricted.
11
-
2.2 Problem ofcharred rernains
fragmentation of
Tab1e I is complicated by the fact thatcertain items such as
charred spikeletsand weed-heads break into fragments,and in doing
so they generate new com-ponents. Spikelets, for example, canbreak
up to give six different cl-assesof component. (In table I these
'secon-dary components' are represented bysmaller crosses
-
txxxr ) . But even inmixed samples, recognition of
these'secondary components' is generally noproblem: firstly,
barring the effectsof differential preservation, ratiosbetween the
numbers of each of the moredurable classes of component producedby
the fragmentation of spikelets arecl-oseIy similar to the
equivalent ratiosin intact spikelets and readily recog-nised as
such. Secondly, seeds andother components fiberated by the
frag-mentation of weed heads (capitulae ofDipsacaceae and
Conpositae, capsules ofPapaveraceae, Caryophyllaceae and
Prim-ulaceae, etc.) are often recognizablefrom their state of
immaturity (seefootnote 4 of table l).
Barring these minor complications, itis hopefully clear from
table 1 thata) there are relatively few crop pro-ducts that are
likely to be found onarcheological- sites in charred form,b) the
principal components of eachproduct are sufficiently different(both
in type and in their relativeabundance) for charred renains of
thedifferent products to be readily distin-guished.
3. THE PRODUCTS OF TABLE 1 MOSTCOMMONLY ENCOUNTERED ON
ARCHAEOLOGICALSITES
On most of the small 'primary-producer'(e.9. farmstead) sites in
Britain thathave been examined so far, the productsmost commonly
preserved by charring arethe 'fine sievings' from stages 12 and13.
These may or may not incfude aminor admixture of some coarse
sievingsfrom stage Il. As indicated in fig. 3,these two by-products
are today quiteoften amalgamated in a common 'clean-ings' store,
though the decision toamalgamate depends on the eventual
usesanticipated for eitber by-product; e.g.
if the 'fine cleaningsr with their tailgrains and weed seeds are
likely to beneeded as famine food, then coarser'cleanings' would
not be added, If, onthe other hand, the cleanings are forfeeding
fowl or for burning, the twoby-products will often be
analgamated.In addition to these cleanings, recentexcavations at
small primary-producersites have afso produced occasionalcaches of
prime grain.
However, it is on larger siteswhether 'manorial-' farming sites
or'consumer' sites - that it is moreconunon to find charred remains
of primeproducts in quantity, whether in theform of grain or
spikelets. In manycases, entire grain or spikelet storeshave been
charred in the course ofwholesale destruction of the entiresite.
Examples are too numerous to
In other cases, cbarring of primeproducts (grain or spikelets)
seems tohave occurred as a resul-t of accidentsduring large-scale
parching of spike-Iets prior to pounding or drying ofmalt
(gerninated grain or spikelets)intended for alcoholic fermentation.
Arecent example of malted products camefrom the Roman manorial-
site of Cats-gore in Somerset (England). Here 4 outof 5 large
'drying kilns' producedrenains of spelt which appeared torepresent
deLiberately sprouted (i"e.mal-ted) spikelets that had been
accidenta-Ily overheated in the course of drying(Hiflman, L9B2a).
Samples of charredrenains that may again represent maltedproducts
have also been reported fromthe post-Roman site of Poundbury
inEngland by Monk (f983). Agaj.n, itseems to be the larger sites
with non-domestic modes of production where suchaccidents occurred
most often. Samplessinilar to those from Catsgore andPoundbury have
doubtless been pubLishedfrom a number of sites in continentalEurope
of which I am at present ignor-:ni. .Fhi c nrnar d^6c n^+ n] :cc i
Frr
^rvrqJJ r !fdiscuss nalted products separately frombufk-stored
spikelets and grain, asthe difference lies merely in thegrains
being deliberatefy (and thereforerelatively evenly and extensively)
ger-minated. In any case, criteria fordistinguishing between malt
and othergrain products are discussed in somedetail elsewhere (in
Hillman, 19B2 a).
12
-
3.1 Waterlogged sitesIn stark contrast to everything so
farsuggested in this section, there is oneclass of site on which
models forinferpreting plant remains in terms ofagrarian practice
cannot be simplifiedas described under 2.f above. Theseare
habitation sites such as FeddersenWierde where plant remains are
preser-ved in bulk as a result of large-scalewaterlogging of
habitation deposits.The daunting task that the analysis ofsuch
assemblages represents is bestappreciated from reading the
remarkableaccount of the plant remains from Fed-dersen Wierde
presented by Prof.K6rber-Grohne (1967). On such sites, any andevery
one of the products and by-pro-ducts listed in figs. 2, 3 and 4
cantheoretically be preserved in quantityby waterlogging. On sites
of thistype, therefore, the narrow range ofproducts summarised in
table I isentirely inadequate, and the modelssummarised in figs. 2,
3 and 4 mustrpgrettably be used as they stand.
Waterlogged people
The gut contents of Toflund man andGrauballe man studied by
HeJ-baek (1950'1951, 1958) are a class of non-charredplant remains
worthy of a supplementarynote of their own.
The use of 'fine cleanings' as faminefood was mentioned above
(in sect.3 ).(The 'fine cleaningsr referred to hereare those from
stages l-2 and 13 offigs. 3 and 4). In the author's view,the
compositon of the contents of bothsets of guts (Tollund &
Grauballe)accord closely with the composition of'fine cleanings'
enriched with a Iittleextra prime grain. As a typical formulafor
famine food, such fare would,perhaps, not have been deemed
inapprop-riate for condemrred prisoners
- if that
is what the two men were. Argumentsfor this interpretation are
offered inHiIlman 1981, 156-8.
4. PROBLEMS OF ON-SITE MIXING OF CROPPRODUCTS
The limitations that on-site mixing ofcrop products could impose
on the inter-pretation of plant remains vras firstdiscussed in
detail- by Richard Hubbard(L976a, 1976b). Certainly, mixing ofcrop
remains can be expected to have
occurred when, for example, the charredresidues from various
minor accidentsduring spikelet-parching or grain-roast-ing were
dumped on the same midden asthe ashes from the burned fuel
ofhearths and ovens. It could be argued,therefore, that the onJ-y
unmixed samp-les wil-f come from those 'primary'contexts where the
products were init-ially charred. If this is true, then,all that
can be retrieved is informa-tion on the last event in each
contextprior to its final abandonment.
rn practice, however, it seems thatmixing of products from
different oper-ations was not always so widespread.Indeed, it is
feasible to use samplesrecovered from even tsecondary' con-texts
such as middens so long as thecomposition of the remains
suggestsderivation from a single class of oper-ation. At
third-century AD Wilders-pool, for example, the composition
ofcharred spelt remains from a very largemidden deposit matched
precisely thecomposition characteristic of the hrastefraction from
step 12 of the processingof present-day glume-wheats (see figure3)
together with a few straw nodesrepresenting, perhaps, the waste
fromstep l-I. Furthermore, the cornpositionwas precisely the same
in alf samplestaken from different parts of theextensive deposits
concerned (Hillman,1983 b).
Purity and uniformity of this sortwould not, perhaps, have
surprised usas it did if we had considered exactlywhich crop
products were likely to havebeen regularly exposed to fire in
theday-to-day life of a Romano-Britishfarmstead in a r^ret climate.
Indeedinspection of the ethnographic modelssummarised in figs. 2, 3
and 4 and thecharred products classification in tab-le I suggests
that the only productswhich, on any one day, are 1ikely toleave
charred remains in habitationdeposits are precisely these
sanecleanings from stage 12 (+ 13) and, inmuch smaller quantities,
the cleaningsfrom stage 11.
In wet clirnates, the dehusking (bypounding) of stored spikelets
of glumewheats and the cleaning of the grain(by smafl-scale
winnowing and sieving)occurs on a day-to-day basis. In
suchclimates, this work is generally doneindoors, and, indoors, the
most obviousplace to sweep the winnowings and dumpthe cleanings is
into the fire burningin the hearth. Here, any light chaffis burned
away; but surviving in the
IJ
-
ashes are two classes of charredremains: firstly, wood charcoal
(ifwood was used as fuel) ; secondly aIIthe denser components of
the cleanings- as summarised in table f (see productno.5).
In stark contrast, accidents duringparching or roasting are
bound to havebeen rare relative to this daily accum-ulation of
hearth ashes laced with thecharred remains of cereal cleaningsfrom
steps II, 12 and 13: primary pro-ducts could scarcely have been
des-troyed through carelessness very often.The risk of different
crop productsbeing mixed in the same midden is fur-tber reduced in
cases where the middencontents were regularly cleared out foruse as
manure. In such cases the accum-ufation of charred remains is
derivedfrom a reduced number of events (i.e.from a shorter period
of hearth use),and the chances of their including anadmixture of
the charred products of arelatively unusual accident during,say,
grain roasting, are corresponding-Iy reduced. This situation seems
toapply to the middens-cum-compost-heapsat the Romano-British site
of CefnGraeanog II (see R,B. White et dl.,forthcoming). It is
perhaps no accid-^-t Lh^-^F^-^ !L-t ; L i ^ ^--^l ^- ^CgrrLt
LllCfCLUIet LIIOL !L r- SOILLPIg> ULrelatively unmixed crop
remains ofprecisely the same composition as pres-ent-day
'cleanings' which are beingrecovered with such consistency
fromsmalf ruraf sites such as Wilderspooland Cefn Graeanog.
On the other hand, we must expect analtogether different range
of charredproducts on larger, more conplex sites,especially those
with rich destructionleve1s in which aff crop products pre-sent in
the settlement at the time ofdestruction could have been exposed
tothe sort of smoul-dering fires typicalof collapsing burnt
buildings. Byexcluding most of the oxygen, suchfires often allow
even light chaff tobe preserved by charring.
In destruction sites such as these,some mixing is inevitable.
Despitethis, where mixing occurs, it is gener-ally restricted to
equivalent productsfrom different crops, e.9. it is notunusual to
find a mixture of barleygrain and Eruner grain
- both fron bulk
srora9e.In summary, then, a) mixing of equi-
valent products from different cropsis not unusual; b) mixtures
of en-tirely different classes of crop product
seems to be relatively rare, except inthe cases of deliberate
amalgamationindicated in figs. 2, 3 and 4.
c) with non-segetalspecies (classes A2 - A6 below), mixingis
quite usual-. Thus mixtures of hazelsbell-s (a food by-product)
with remainsof, sdy, a bedding/thatching speciessucb as heather
have occurred at anurnber of sites.
5. THE SORT OF QUESTIONS ANSWERABLE BYUSE OF PLANT REMAINS
INTERPRETED VIAETHNOGRAPHIC I'IODELS
Before consider ing how a large andcomplex body of data from an
assemblageof nl Anf rema i ns 66n be related toethnographic models,
it is appropriate+^ hrioFlrr nnnciiar l-ho cnri nF drr6c-tions
which can be answered. However,they have been discussed in
detailelsewhere, and littfe more than a bib-J-iography of examples
is offered here.
5.I Was the site a 'primary producer'(i.e. farming) settlement
or strictly atconsumer' settlement?Features of composition which
can beused to distinguish between remainsfrom 'primary producers'
and, sd!r apastoralist'grain-consumer' are discu-ccad in Hi'l'lm:n
1lQQI :nd l-n: Ia
-
numerous. They include the follwing:lluurman (L979') , Colledge
(forthcorning) ,I)enne11, (1972, )-974, I97'l , 1978);Hillman
(1972, 1973, I9B1 pp.I27 & 143-4); Hubbard (1975, I916a, I976b,
1980);Knbrzer (L981); G. Jones (1981, 1982,forthcoming); G. Jones
and Rowley-Conhry(in press). The mechanics of this formof analysis
are also discussed below.
5.3 Did they till their land with ardsor with
moufdboard-ploughs?Discussed in Hillrnan 1981 (145-6) ,1982.
Cfearfy, this question is rele-vant only to sites post-dating
thedevelopment of the moufdboard plough.
5.4 Did they sow their crops in autumnor spring?Discussed by M.
Jones (1981) and ingreater detail in M. Jones (in prepara-tion) ;
also Reynolds (1981 a & b \ i and,HiIIman (1981) .
5.5 Did they irrigate any of theircrops?- Crop types as(1969,
t_972r' . indicators: Helbaek- Weed floras as indicators:
Charles,(forthcoming); HilIman and CoIledge (inprep). + Several
studies of presentdayphytoecology.
Note: These last three questions (5.3,5.4 and 5.5) are addressed
archaeo-botanlcally not by reference to ethno-graphic models but
rather by referenceto modefs for the ecological behaviourof key
weed species (or species assem-blages). They are given mention
inthis discussion of ethnographic modelsonly because anal-yses
which aim toidentify specific aroups of weeds canbe built into the
sort of analyticalsequences outined below, if only as anextension
of the 'D' classificationsystem. Glynis Jones is aLso in theprocess
of devising novel analyticalsystems specifically for extractingfrom
weed remains information on pastedaphic environments (G. Jones,
thisvol. ) .
5.6 Did they rogue (hreed) their crops?See Reynolds 1981, 1982;
Hillman 1981.
5.7 Harvesting methods?Discussed in van Zeist and Bottema(1971
pp. 537-538); Reynolds (1981,I9B2); Kndrzer (1967) and Hillman
(1981pp.148-153)
5.8 Crop processing: threshing, winno-wing, parchingl dehusking,
sieving andhand-cleaning.The effects of each of these operationson
the composition of crop productswere discused by Hillman (198I) and
arediscussed in quantitatively definedterms by Glynis Jones in this
volume.Results of laboratory experiments withsieving hrere
presented by Robin Dennell(I972). Effects of coarse sievingdetected
in charred plant renains fromTeIl Medhur are discussed by
RichardHubbard (forthcoming). A number ofthese operations and their
productshave also been classified in accessible.tabular form by
Kndrzer (1981).
6 APPLYING THE MODELS TO ARCHAEOLO-GICAL REMAINS
we have our site; werve recovered adozen (perhaps hundreds) of
sampLes ofplant remains; sample by sample wervesorted them and
identified them to a'point of diminishing returns' fixed,inturn, by
reference to questions posed,perhaps, at the outset of
excavation.How, now, do we relate the (often)thousands of
identifications and scoresto the models outlined above?
With smal1 assemblages of up to, say,ten samples, any
simifarities in compo-sition between each of the samples anda
particular modern crop product willoften be obvious from simple
inspec-tion. However, in assemblages invol-ving large numbers of
rich samples ofdiverse composition, siniLarities withmodern crop
products are not alwaysobvious from inspection alone, and itis
generally impossible to recognize'by eye' any significant patterns
ofcorrelation between horizontal varia-tion in sample composition
and thevarious site context types. Some defin-able and repeatable
system is clearlyneeded toa) reduce the numbers of variableswithout
losing what could prove to bevital information,b) extract
information on husbandry and
15
-
ItitiM di@ccw/ T- Nno@ctu!. ti@cN ftenl.) o. I. Nho@ctu
t,ilecM oE !- sefta
I.sFlrE or T.aestivennPaclw - - -
rritiM -
lndet-flee-tueEHng spp.rducd - qsela] bder.
9!abs6
gI. /I@a f.ags
a = hulled / " /xeL.rd ss-basdl ot !e1.1- ctdenEe
DoD-basa!.l
" /2 .d/ (bdicate'a + h' wI$ Ecolc" /2 ot s loBl deDBe ealed. /'
" /raeued" / " " Aasalo!re" ,/ hdet. iragsgr&E
- a66Fek1cr
- 6Pet!1c
" - ureier&Ie.n f.aFents
_
Ep. (9!a55 spp. ) gtai.s
(hon-tv1sted, lower
avena fatua- - - - - -lvaa stetjljs (lnc.)- -
C*aoviciaalcf.Avetutatoa/AsterilislyenasF-1ndet.----
? Aveha sp-
(:IMS IMET- orah (@uted as no.ol dryos)(no att@pt ude hele b
couE tall qra
Cereal6 / o$e! 1ar9e qlasses - cub bses
Egre of cetumty of rdentifictaon is indld by psrtron of each
score wi$in its veftrcl 61tm: sores on left =etuin identifr-tlons;
smres on rrght = d$rous rdstlfrctlons; rnteffirate psrtions =
orresgillnqly urefrdiate
hci .lm aos. are rmbores G.lrl]rw 1976; revlsd 1919
Ficfure 5. EiAlipLF oF oNE sEtil oF pnrljARy scolE TAB;ig usED
foR clRxAr tlflltNs i,r,1rr slTE ofCEFN GRAEANOC
16
-
processing methods, andc) test the significance of any appa-rent
relationships bethreen the horizon-tal distribution of site
contexts andthe various components of variation insample
composition,
As our example, we will use the re-sults f-rom a site in North
Wales (Bri-tain). This site is Cefn Graeanog' anative British
farmstead of the LateIron Age and Roman Periodf Located onan
exposed ridge of the Lleyn Peninsulain Gwynedd (NW wales). It
provides auseful exarnple becausea) excavation revealed a wide
range ofclearly defined structural contexts,(see n. White,
forthcorning);b) the excavator, Richard White, reco-vered cbarred
remains of plants fronalrnost every one of these contexts
(280samples from 44 contexts);c) the remains hrere quite rich:
theyincluded over 250 taxa and chaff/grai.nclasses;d) the site is
of a type common inparts of Britain and perhaps Europe aswell.
Fig. 5 is a copy gf one of the cereal-score sheets, and Fig. 6
is a coPY of asmall section of a score sbeet for thel-84 non-cereal
spp. identified from thesite. Each vertical column representsone
sample. Scores on the far left ofany one column represent the
nunber ofitems of certain identity, while scoreson the far right
represent numbers ofdubious identity. Intermediate posi-tions
represent intermediate levels ofcertainty or uncertainty of
identifica-tion. This device obviates the use of'cf.s' and other
formulae for indica-ting various degrees of confidence
inidentifications. (Expl-anation of otherfeatures of the scoring
system willappear in Hil-lman, forthcorning b).
In analysing the data from the CefnGraeanog plant remains, two
strategieswere open to us:a) The first strategy would have
beensimply to feed into principal componentsanalysis (P.C.A.) the
separate scoresof every taxon from each sample, seehow the samples
get grouped, and then- firstly compare the composition ofsamples in
each of these computer-produced groups with the
generalisedcomposition of crop products in ourmodern models to see
if the groups makeany sense in agrarran terms,
- second-
Iy, test the distribution of suchgroups for any significant
correlation
with the distribution ofstructures.
excavated
we abandoned this strategy for thefollowing reasons: (i) we had
too manyvariables for the matrix size of anyP.C.A. progralnme
availabfe to us at therima /ii\ l mrinril-rr nf l-ho l.ava a^r\rrl
.r rrreJv!4LJ v! s'r!chaff components) were present in onlya few of
our samples and' treated asseparate variables, would nornally
beunusable. OnIy by amalgamating scoresas explained below could we
avoid thisIoss of potentialy valuable information.(iii) P.C.A.
systems tend automatical-ly to be biased by (i.e. over-weightthe
significance of) isolated rarities.For example, we found that in
the cour-se of a small-scale 'trial- run' thecomputer had separated
certain samplesinto separate categories of their ownsimply because
they had two seeds ofNardus stricta which was a species notpresent
in other samples. This prob-Iem, too, could be circumvented only
bythe sorts of amalgamations used in thealternative strategy.b) Our
alternative strategy (and the
one finally adopted) was to-
classify each taxon and each classof chaff and grain by direct
referenceto the ethnographic models;- within each sample'
amalgamate the
scores of all items of like class fromany one sample;-
eliminate all classes whichr despitethe amalgamations, are still
represen-ted in very few samples;-
convert class fequencies to thoseratios which, fron the
ethnographicmodels, could be expected to provideanswers to
questions relating to husba-ndry and processing methods.
(Theseratios are thus used to characterisethe key features of
sarnpl-e conposi-tion) .-
using P.C.A.,(i) test for simil-arities in composi-tion between
samples from any onephase and between samples fron anyone class of
context (within phase)e.g. hut floors, hearths, middens;(ii) group
all refated samples on thisbasis;(iii) test for systematic
correlationbetween distribution of sample groupsand distribution of
excavated featuresin any one phase of the site.(iv) Having thus
accounted for (i.e.eliminated) the major components oflateral
(horizontal) variation withineach phase, test for any
systematicchange through time (i.e. variationbetween phases).
1a
-
To
g
s solted bv (1nitl
145 |nbel.146 Euplo14b Polt)q,
148 P. ld149 P. hst50 P. hu151 P- d152 p. hr153 p. nierc
15-S E!4eI1l-1 8i__99
159 Stell1r,0 urrlcn,r E;TTi
1.
.1-l!_t2_lll_JJ!
Fio 6, EXAMPTn oF oNE SHEET oF PRTMAR? scoFE TABLES usED FoR NoN
CEREAL REMATNS A1 l'HE srrtr ofill N ul'q AN, i
Scores = nrrD.:rs ot rtens rdc.Lrtred sLoros on IeIL of c!l"rrs
coiLdlr) identjficdLtorr5; those on llghL = dublous
ldentlflcattons.(AlI foocnoLes aFPea. on separace sheeEj -
1B
-
Each of these analytical steps wiII nowbe discussed in more
detail-
6.I CLASSIFICATION OF REMAINS (AT CEFNGRAEANOG)
STEP l: CLASSIFICATION OF EACH SPECIESBY THE PROBABLE MODE OF
ARRIVAL OF ITSSEEDS ON THE SITE
This classification does not drawdirectly on the ethnograhic
modelsoutlined above and is necessarify high-ly subjective.
However, it is anunavoidable first step.
(seetable2-below)Different sites wifl clearly requiredifferent
systems of classification,
depending on which species are represe-nted in the remains. At
Cefn Graeanog,for e*ample, a diverse range of modesof arrival (i.e.
types of usage) had tobe grouped under the single class (A2)simply
because large numbers of Calluna
(heather) seeds, capsules and flowerslvere recovered from the
site, and theycoufd have come from plants which hadbeen used either
as bedding (for humansor animals) and/or as fodder and/or asfuel
and/or as thatcbing. As many aspossibJ-e of these different
productstherefore had to be incl-uded in the oneclass (A2) .
Simil-ar constraints forcedus to amalgamate four disparate
producttypes within another of the classes(A3), and this class
consequentlyincorporates as diverse a mixture asclass A2.
Assj.gning any one species to a singleclass was problematic,
even when theclasses are as broad as some in thisclassification
which was developed forCefn Graeanog. Most difficult of aI1hras
deciding which of the 184 specieswere likely to have been segetals
(i.e.weeds of crops). The nature of ourdilemma is perhaps apparent
from table3 on the next page.
(see table 3 - next page)Tabl-e 3 represents a smafl segnent
of
the fu1l cfassification of all 184
OF ARRIVAL OF SEEDS ONTO SITE
s determined by which species were recov-fn Graeanog). Other
sites wilf require
ng as components of crop products, i.e.ves, separately harvested
ears or strahr.thered as fodder (inc. hay) , bedding anded only by
wood charcoal-s).condiments, medicines or dyes.
ishings' such as rush-matting or reed-
urned as fuel, Such seeds are probably
systematicaly) by other means; e.g. seedse rv,rel-1y-boot ef
fect') , dead ruderals cutas decoration, etc. (Al-though many
spp.pecific categories \rere sought wheneverbability that any given
seed had arrived
l9
-
Table 3. SITTALL SECTION OF ONE OF THE GENERAL CLASSIFICATION
TABLES OFSPECIES IDENTIFIED IN CHARRED REMAINS FROM THE SITE OF
CEFN GRAEANOGNote: The only species which were used in subsequent
analyses were those for whichonly one of the alternative
classifications (within any one of t}te systdns A,B,C orwere
probable. (The less probable alternatives are given in brackets).
on thisbasis, therefore, both Hlzpericun and Lychnis were
furnediately deleted frcrn furtheranalysis.!{here no urrlcracketed
B or C classl' is offered, the plant was probably not a
segetal.Where no unloracketed D classification is offered, the
plant was quite possibly asegretal, so a firm'D'classification
would be nisleading-
D)
computercode for
1*-",
r11rL21I3114
IlsII6r17rr8
LirLO Z
ir6 4-6
r o / -o
qmniaq iAan+ifialin rernains
Hypericr.nn sp. or spp.Lychnis flos-cuculiStellaria
nediaSpergula arvensisIrbntia fontanaScleranthus perennisCheno@irm
albrmC. murale
Salix albaQ:l i v n:nrar
CaIIuna vulgaris/ ^^^l^ 1,. r^ \\ wc@, IVJ./ !D./
Erica tetralix
classif ication systemsR/-
A2 A6AI A2 (A5 A6 )Al (A6 )AIA1 A5 A6A1
A1(A3 46)Al(A3 A6)
(A5 )(As )
(A5 )
(As )
(A5 )
r,DJ DI l/p? )
D)B3rR? l
EJ
bJ
bJ
(c3 )(c3 )
c3 (c4 )(c4 )
c3c3 (c2)
(Dr )(D2 D4)
D4 DID4 DI
AO
AO
N.
(Dr )D4
(DI )(Dr )(Dl )
D2 D3
D2 D3
D2 D3
(1r,c., fls.)169-jI Erica cinerea(seeds, 1vs., fls.)
(The full table of classifi-cations offnrfhcminc hl
all species identified at Cefn Graeanog will___l
non-cereal spp. at Cefn Graeanog usingall four (A to D) of the
classificationsystems discussed in this Paper.
In table 3 it may seem strange thatplants of marsh or damp
meadow such asLychnis flos-cuculi and Rorippa islan-dica, should be
assigned a segetalclassification. This reflects the factthat, in
this system' all species iden-tified in the site remains are
classi-fied according to the way their seedsprobably got onto the
site, and not
according to local habitat's in whichthey nay have been most
prolific. Anexample is needed here. At Cefn Graea-Do9, the land
most 1ikely to have beencultivated ran dorlrn into a mire. As
aresult, the lower ends of the fieldswere probably narshy (as they
are to-day) and heavily invaded by a wetlandlveed flora including
plants such as R.islandica and L. flos-cuculi. Eventhough such
plants would inevitablyhave been more abundant in the adjacentmire
(they still thrive there today)
'
20
-
their charred seeds recovered from thesite were consistently
mixed with cropcleanings and are likely to have ar-rived as crop
contaminants. ft is forthis reason that wetl-and plants of thistype
were classified firstly under AI(and only secondarily under A2, A5
andA6, see table 3). On the other hand,plants such as Caltha
palustris andCarex pauciflora which are typical ofeven wetter
habitats seem less Iikelyto have survived as weeds of wet
arabl-e1
-hi u^r.'^' '^r ; F":ci ^n
hrz r^ro1- l:nrj qnnIqllg. , lrlvsJlvrr u) vrsu, u"u -Yy.has
been observed even in ar:-d areas. Forr---al a
--^i^ /T^ri ^ernc ) nF Al i cm: l:nca-gAarLUlcr JcsuJ \llgr
luaLyJolata were identified in 'fine sievings'frorn the processrng
of crop products frcman- rmirriqated wheat field at Aqvan
(1.T\rrkey). This occurrence sesns explicalrleonl:; on the basis.
of occasional plants rrav-inE invaded wet patches in the field
adja-cent to water -channels. But while manyperer,nials can,
indeed, survive cultivati-
1^^^ -^ |L^ I--r I^ ^^r,, Ii^L+1.,ult JU f ut Lv d5 Lllc ldllq
f5 uttrw r r qrlL r)
tilled with an ard (see Hillnran 1981, 145- 6), the presenCe of
Alisma in thesc-ar6n nrrvir rc1-q mrrcl crrral rr r6nroq6n+ anan
extrerne case, ald Alisma would nevernormally be classed as a
segetal.
Distinguishing between segetals andruderals may appear to be
even moreproblematic. However, precisely thesame principal applies
here as it doesfor the wetland species discussed above.Once again,
charred seeds of typicallyruderal species found consistently
inassociation with crop 'cleanings' arelikely to have arrived on
the site (andgot into fires) primarily as contami-nants of crop
products; they are farless likely to have arrived via one ofthe
rcasual' routes grouped under classA6 (see table 3). In most cases,
there-fore, typically ruderal species areassigned to class A1 as
the mode ofarrival offering the most plausibleexplanation for their
seeds gettingonto the site and into fires. (Oneexception here is
Urtica dioica which Ihave never observed as a segetal, evenunder
the lightest ard cultivation).
A1I probable segetals (i.e. all spe-cies with a relatively
unequivocal Al-classification) can no$, be furtherclassified under
the 'B' system. Ofthe examples given in table 3, there-fore, the
species that could be carriedforward into the 'Br
classificationwere - Stell-aria media, Spergula arven-sis,
Scleranthus perennis, Chenopodiunalbum and C.mura1e.
STEP 2:SYSTEM B: SUB-CLASSIFICATION OF CROPSAND SEGETAL WEEDS
(FROM CLASS IA.IIABOVE) BY THE TYPE OF CROP PRODUCT ORBY-PRODUCT IN
WHICH EACH ITEM IS NORMAL-LY FOI,JND.
This step effects a sub-division ofthose species which were
classifiedunder class AI in the preceding step.Here, then, each
segetal weed seed andeach cereal component identified in tbesite
remains is now assigned to a par-ticular crop product. These
assign-ments are made strictly by reference tothe ethnographic
models summarisedabove, and the overall objective is todiscover
which processing stages arerepresented on the site. Of the
weedspecies, the only ones used are thosewith a relatively
unequivocal Al clas-sification in the preceding step (step1). And
of the crop products, onlythose listed in tabl-e I are
consideredhere. This'B system' classification issurunarised in
tabl-e 4.
(See table 4 -
next page)a) classification of crop renains underthe B
system:The major components characterisingeach of the crop products
likely to bepreserved in charred forrn on arcbaelog-ical sites were
outlined in table 1.The basis for assigning any one
cereal-component such as glume bases to aparticular product was the
relativeabundance of this component in theseven different products
sumrnarized inthis sane table I. On this basis,however, the only
conponents whichcould be assigned a fairly unequivocal'Br
classification (and therefore beused in subsequent analyses) were
thosewhich were conspicuously abundant inonly one class and
rel-atively rare inall others. Any component that wasfairly common
in two or more differentproducts therefore had to be deleted,As an
example of the mechanics of thissystem, tabfe 5 shohrs the 'Br
classes(i.e. crop product classes) to whichjust 15 of the cereal
components wereassigned. (The full range of 78 typesof cereal
component found at CefnGeaeanog are listed in the pr imaryscore
sheet reproduced as fig.5, above)
21
(See table 5 - next page)
-
i -oT--eFo-FlI rrroNAL Iri
lllote: For the purpose of this tab1e, only glume-wheat products
are considered andlof these, only those that are like1y survive in
archaeological sites (as listedlin table 1) .t81: winnowingiB2z
'cavings'iwithout'chob'
waste (products of steps 4, 5, & 10 in fig,3).waste from
coarse sieving from steps 6a and lI in fig.3 (with orfrom step 13a
in fig.4).
83: cleanings from fine sieving (from step 12 in fig.3, and step
I3b in fig.4).84: semi-clean prime grain etc. (from accidentatly
burned grain stores in dryareas).85: cleanings from hand-sorting
(from step 14 in fig.4).86: pure prime grain (probably mainly from
step 24 in f.rg.4).Two additional products occur in charred form on
archaeological sites,products I and 3 in table I. However, they are
omitted here as theiris effectively a combination of t$ro or rnore
of the products listed.(In the text, the first of these is coded as
BI).
5. EXAI4PLE OF MECHANICS OFPART OF A TABLE OF CERNAL REMAINS
OF CROP PRODUCT WITH WHICH IS USUALLY ASSOC]IA'I'EIJ IN
PRES!;N'1'-DAY i
STEP 2 OFFROM CEFNEACH ITEM
PRODUCTS
't'= tail grains. rp' - prime grains. These r^rere scored
separately (see fig.5,above) as were the 'unreferable' grains of
intermediate size which cannot bessigned to specific crop products
and which do not appear here.
class of productscomputer with which norm-
code -ally associated
rcoccum or T. spe rachis fragments
-PEr Ld.
.spelta or aestivo-compactum
. aestivo-compactum.
riticum sp, (indet.naked sp.)
spikelet forksglume basesgrainsrrah i c fr:nmanfcspikelet
forksglume basesglume f rags. (non-basal)9ralnsrachis
nodes,/forksgra lnsrachis nodes9ra rnsrachis nodesgrains
83 (B5) iB1 (83)p:85; t:83(85) i83 (85) ip:85; t:83 (85) i
B9
1011T21314l5l6I71819202I
83 (Bs)83 (85 )83 (85 )p:85; t:B.3 (B5)83 (B5)
83 (85)I
82 (Br.83)p:85; t:83 (85)82 (B1.83)p:85; t:B3 (B5)
22
-
lail grain vs. prirne grain::::ring two products (nos.f and 3
in::rle I) which are rarely found in:.:rred form at primary
producer sites,::nains of grains are characteristic of:;c products:
firstly, cleanings from--:re-sieving (product 83) whicb contain::i1
grain; secondly, product 85 which::nsists of semi-cl-ean prine
grain with= minority component of tail grain.- cr'i nn,r'i chi na
Iri l nr=i n f rnm nrimogqrr lrsrrr !!vrr'::ain can therefore
contribute to the-:entification of product type. How-:'.'er,
distinguishing tail and prime;:ain requires knowledge of mesh
sizes-i the fine sieves used, because it is::ring'fine-sieving
(stages l-2 and l3)--:atmuch of the tail grain is unavoidab--,
eliminated from the prine grain along;-th small lreed seeds and
glume-bases.
Clues to the nesh diameters of the::eves used can, in fact, be
extracted::orn measurements of the maximum diame--5rc /hr6AdFhc\ n
fha ar:inc nr6e6hi-
-n uncontaminated samples of 'fine:-eanings'. The grain in fine
clean-'-.c i c ownl rrci rrol.' ts-i I ^--i h ^f-.,JJ L^eruJrvrr),
Lqtf Yrqfrrt
-^ !L^,^ i^ ^^,,-., LL^r l-v^^-JU!5Ct d5 LllElg I5 rrU Wdy LridL
ro!YYt::ime grains could have passed through--re fine sieve. If,
therefore, the::i.ginal frequency distribution of max--i.un grain
diameters $ras orinally oflaussian form, then thorough sievingi;ith
meshes designed to allow the elim-:nation of most of the smaller
weed:eeds should theoretically result in--re loss of rnost of the
tail grain as;;e11. The theoretical effect on a:ingle batch of
grain from bulk storage;ould then be as follows:
bil grain
a single batch of grain is thereforefollows:Bt
nunbersof
qraus
I-t
rn examinins .;J;."' l'Jurn",however, it is usual to measure
egualnumbers of grains from each of thedifferent samples,
regardless of whet-her they were recovered from niddens
orgranaries. Thus, vrhen equal numbers ofgrarns are measured from
sanples ofcleanings on the one hand and samplesof prime grain on
the other, then thefrequency distributions inevitably ap-pear to
exhibit a rather differentrelationship to each other than
thatillustrated above:
tn
gar'i tro s4le l:LrLl gfhrD fofrrngFlt of !trne cleaflrysc
t
"",*."" "*.*., "*,J nwqs drmters of qrarnsA,f
I
'::"i-..-.1
twilm direbs of graiN
However, sieving is rarely that thor-rugh, and not afl- those
items which:ouId theoretical-Iy pass through thesieve are, in
practice, eliminated, Incther words a variable proportion of:he
tail grains remain with the prime;rain. Armore realistic
representationof the effects of sieving on
the:requency-distributions of grain sizes
The approxirnate diameter of the sievemesh used can then be
esti.mated asindicated. (ClearIy, sieve mesh sizescannot be deduced
from any part of theprime grain curve) .
At Cefn Graeanog, it proved possibleto estimate nesh diameter as
indicatedabove fron one + pure sample of charredremains of 'fine
sievings'. This esti-mate was then used as the basis foridentifying
alf other grain (from thesame phase of occupation) as either'taif'
or 'prime'. Only those grainswell above or below the estimated
meshdiameters were, in fact, referred toeither class; the rest were
left as'unreferable' (see example of scoresheet, fig.5) . It must
be stressed,however, that on many sites, slightcontanination of
rcleanings' is not un-comnon: there are sometimes small ad-
passed tJEough sreve.c
'E* ^F rfino
prre graf'rehined1n steve
cl(aings'
.*"a *"n -*: f
23
-
mixtures from other products, and theseadmixtures commonly
include a littleprime grain. These prime grains inevi-f :hl v
nrndrrna arran orcAf cr orrcrl anbetween the two frequency
distributionsin diagram 'C' above.
(b) .B. SYSTEM CLASSIFICATION OF WEEDSEED REI{AINS:Fron our
studies of present-day cropproducts, it was clear that the
princi-pal factors determining what seeds werepresent in any one
crop product were(i) the ratio of their surface area toweight
(i.e.their winnowability), (ii)seed size (i.e.sievability), (iii)
seed'headedness'. Each of these categoriesand their use in charred
remains willnow be considered in turn:(i) WinnowabifityFor any
given hrind strength, theprobability that a seed can be winnowedout
of the prime products seems todepend primarily on the ratio-of
itssurface area to its weight (run2.g-I ).Tbis ratio tends to
increase with dec-reasing size with the resul-t that win-nowing
tends to eliminate very snallseeds such as those fron
Campanulaspecies. The presence of h'ings on theseeds afso increases
surface area, ofcoursei winnowing consequently elimin-ates the
winged seeds of pJ-ants such asLinaria vulgaris and Rhinanthus
seroti-nus as well- as the winged fruits ofArtedia squamata,
Aellenia autraniand several species of Scabiosa. AtCefn Graeanog,
therefore, any cbarredremains of winged or very snall seedsand
fruits of segetal species brerecfassified as winnowing waste.
(Suchseeds were, in fact, very rare)./ii\ ciorrrl^rilifrrSeed size
is important in that it det-ermines not only 'winnowability' atthe
bottorn of the size range, but al-solciorrrhilifrrl in l-hr
-,. -..e upper srze ran9es.This effect is reflected in the
clear-cut correlation between tire major pro-ducts and the sizes of
tbe seeds con-tained in them, as indicated in tablel. On the basis
of this correlation,seeds and fruits can be grouped i.ntofour
sievability/winnowability classes,each of which is characteristic
of asingle class of crop product. (These 4classes are outl-ined in
table 6).Clearly, therefore, the identificationof charred remains
of these particufar
products is very straightforward soIong as they are not mixed:
it is mere-Iy a matter of observing the size ofthe seeds and noting
whether or notthey show any signs of having once hadwings. (Charred
wings commonly breakoff). (See table 6)(iii) rHeadedness'.Many of
our most common weeds producetheir seeds in capsules or
capitulae(heads) r .9. Papaver spp., Gypsophilapilosa, Vaccaria
pyrarnidata, Cephalariaspp., Circiun spp., Anthemis spp., etc.Many
of the seeds in these capsu]es orcapitulae are refeased only in
thecourse of processing. this is particu-'l:r'l rr f ha c:co i F l-
ho arncrr'l ocimmature at the time of harvesting, andin such cases
the refeased seeds oftenshow clear signs of immaturity, evenwhen
they are charred. Indeed , Lf,after winnowing, the threshed
spikeletsare cleaned thoroughly with a medium-coarse sieve (stage
6b in fig.3), thenalmost all the free seeds found inensuing crop
products and by-productsnecessarily derive from capitulae
orcapsules of about the same size asspikelets. A large proportion
of themalso show signs of immaturity. Thepoint at which these seeds
are liber-ated from immature capsules is stage 9(fig.3) when the
parched spikelets arepounded in order to release the grains.The
effect of this liberation of seedson the composition of ensuing
productsis indicated diagrammatically in tableI (see small crosses
in the last fourcolumns). Most of the liberated seedsare eventually
separated from the primegrain in stages 12 and 13 as usual.Light
fragments of capsule wall are
el-iminated by the second winnowingstage 10
An equivalent refease can occur withseeds in the 'winnowable'
category incases where they have been retained inimmature heads or
capsules (e.9. inimmature Scabiosa heads or Campanulacapsules) .
These seeds will again bereleased in stage 9 (poundj-ng of
spike-lets). But in this case, they areseparated from the grain
along with the'light chaff' during the 2nd winnowing(stage l0)- as
indicated in fig.3. (Seeal-so footnote 3 of this same
figure).Theoretically, the retention of winnow-able seeds in heads
suggests (as ex-plained by Glynis Jones in this volume)that the two
parameters
-
rr,rinnowabif i-ty' and 'headedness' should not be used
)A
-
-ME'D-SEEDS :::ASSIFIED BY B CLASSES AND THE TYPES OF CROP
PRODUCT TN WHICH THEY ARE NORMALLr3UND
c)
Unwinged seeds/fruits whose narrowest Qs aresignificantLy >Q
ot largest prime grain (butnot much ) spikelet width).x e.9.
Tordylium,Aristolochia, Gundel-ia.Unwinged seeds/fruits whose max.
ls are withinrange as widths of prime grain. eg.
Agrostemma,Cephalaria syriaca, infected grains of
Loliumtemulentum.
Unwinged seeds/fruits whose max.ps are signif-icantly 0.5
mm.e.g. Vaccaria pyramidata, Sinapis arvensis,Gypsophila pilosa,
Polygonum aviculare.
i). Winged seeds (a11 sizes) e all seeds
-
independently. Inpractice, however.winnowing waste is very rare
in sitedcnosi fs (hence i ts exclusion fromtable I). Even if light
chaff rich intwinnowabler irunature seeds were recov-ered, there
would be no problem indistinguishing it either from the coar-se
winnowing hraste of the first winnow-ing (stage 5) or from any
other cropproduct.
As for the pattern of occurrence ofthe intact 'weed heads'
themselves inthe major crop products, our studies ofpresent-day
products suggests that thisis a function of sieve mesh-diameters.As
mesh-diameters, in turn, are careful-ly fixed by the sieve makers
to matchgrain and spikelet widths, it is reason-able to take a
rshort cutr and classifythe weed heads by their size relativeto the
breadths of spikelets and grainspresent in contemporaneous
deposits.In this way, the heads are thus automa-tically classified
according to thecrop products in which they are likelyto be found
following sieving. Thisclassification is outlined in table 7.
(See table 7 - previous page)It should be noted that the
classifi-
cation of weed heads in table 7 doesnot make use of their
absolute size.Instead, it merely uses their sizerelative to the
width of prime spike-lets. Such a classification is there-fore
easily applied to weed heads incharred remains: it merely requires
thewidth of the weed heads to be comparedwith the width of any
spikelets (orwell-preserved spikeletforks) recov-ered from
contemporaneous deposits. (Amore exact approach to quantificationof
'headedness' is presented by cfynisJones in the following
paper).
At the site of Cefn Graeanog, thisrBr classification system was
appliedto every cereal fragment, segetal rreedseed and segetal weed
head from everysample of charred remains recoveredfrom the site.
Anal-ysis of cl-ass to-tals (as outfined below) reveal-ed manyof
the samples to be dominated by re-mains of specific crop products
oftypes still to be found in the presentday. The sort of notation
used in therB' classifications which were appliedto these remains
was illustrated intable 3, above.
STEP 3.
SYSTEM C: FURTHER CLASSIFICATION OFSEGETAL WEEDS (FROM CLASS A1,
ABOVE) BYTHEIR GROWTH HABIT AND HEIGHT WHENGROWING IN CROP
STANDSThis step represents a further sub-classification of the
segetal weedsalready assigned to class AI. Ourobjective, here, is
to extract informa-tion on the harvesting methods appJ-iedto tbe
crops represented in fhe charredremains fron the site.
(See table 8 - facing page).while twining habit is a
relatively
absolute criterion, classificati.on ofweeds by the height at
which they formfruits can be rather arbitrary. First-Ly, the height
of any one weed speciesvaries dramatically in response tofactors
such as soil-water availabilj-ty:nd Aanci t-rr nf |-hn
-rlh^ T!vgrrJr L-I v! rrrc ur9y sLollu. I Lfurthermore seems
unlikely that thesefactors consistently affect the heightof the
host crop by precisely the saneamount. SecondIy, v,reed heights
areexpressed relative to an 'average' cropheight (see note to table
B), and evenunder a standard set of conditions, itis possible to
observe enormous differ-ences in the average heights both with-in
one population as well- as betweendifferent varieties of any one
cropspecies. For example, some of theshortest present-day Turkish
Emmersbarely exceed 60 cm., while the taflestexceed 150 cm. It is
impossible to becertain, therefore, whether or notreaping high on
the straw of an ancientcrop would have included heads of,sayf
Agrosteruna.
Because of these uncertaintj-es (espe-cially in assigning
certain species toC2 or C3), the 'C system' classifica-tion is, I
feel, to be regarded as nornore than an optional (and
sometimesdubious) supplement to the A and Bsystems described above.
Certainly,identification of harvesting by uproot-ing, at least, is
perhaps better attem-nl- cd hrr rrqo 6f l-ha nroconca aF aoro:
ltsvvgtslvlgrrglv!nrr'r m hrcac f coa f ig.5 bel_ow) . (Forexamples
of rC systemr classifications,see table 3).
Identification of reaping heightsfrom charred remains is aLso
discussedby van Zeist (1968), Glynis Jones(L9791 , Hiflman (1973
and 1981), andReynolds (I98I).
26
-
8. SYSTEM C: CLASSIFICATION OF SEGETAL WEEDS (FROM CLASS A1) By
THErRFORM AND HEIGHT WHEN GROWING IN CROP STANDS.
l;3Le: weed heights are expressed relative to supposed beight of
the host crop,=: it is this relationship which determines which
weeds get harvested. The exam-.-: used here is an'averager crop of
Spelt wheat. 'Weed height' is taken as the--lrest point at which
the plant forms fruits.
twining weeds These are automaticafly harvested when crop is
uproot-ed or when reaped either l-ow or at mediun height
, on the strawi e.g. Polygonum convolvulus.free-standing
weeds
- 3/4 height of crop or taller: These are harvestedby medium and
low reaping; also by reaping highon straw when this is done
carelessly; e.g.Agrostemma githago, Gypsophila pil-osa, etc.
free-standing weeds -
I/4 to 3/4 height of crop: These are harvested(together with C2
hreeds) when the crop is reapedfairly Iolv on straw; e.g. Anthemis
cotula,Bupleurum rotundifolium, Papaver dubium, etc.
free-standing weeds - I VA height of crop: These are harvested
only by
very low reaping or by uprooting when this is perform-without
tborough 'root beatingr; eg. Aphanes microcar-pa, Polygonum
arenastrum (prostrate forms), Galium ar-ticulatum, Aristol-ochia
clematitis.
-_:.SSIFICATION OF NON-SEGETAL SPECIES:.e. CLASSES A2 TO A6) BY
THE HABITAT
:l; h]{ICH THEY PROBABLY GREW..:--I species other than crops
and
--.^.:ir probable weeds were next classi-:-:d according to the
sort of habitat:::n which they were likely to have-=:ived in the
catchment of this parti-:--ar site (Cefn Graeanog). OnIy
those=;:cies assigned to classes A2 to A6.::e usd in this step of
the analysis;-.:. we used only those species which-::e unlikely to
have been able to:::vive as weeds of crops (see=-=:t.6.1, step 1,
above) . Such plants:::cably arrived on Lhe site direct:::n
non-arabl-e habitats in the area=:j may therefore offer clues to
the--.'-:es of vegetationaf resources avaif-=--e in the area of the
site durinq its-.-,ul/oLrurr.
:ne objective of this step in the::.alysis is therefore to allow
amalgam-:::on of records of plants of like-.:oitat with a vievr to
extracting infor--:tron on past patterns of exploita-:-cn of the
pfant resources provided by=:ch habitat type.
By reference to present-day vegeta-tion in the area of Cefn
Graeanog today,it seemed that the assumed non-segetalspecies
present in charred remains fromthe site could be crudely divided
intofour groups as follows:Df:
- weeds of waste land (ruderal-s);
n?. -
n.cl- rrr6 ^r
harf l.r cnani ac.vrvv 4vv,D3:
- plants of cleared woodland orwoodland fringes and glades;
o
D4: -
marsh and bog species.However, none of these classes are
mutually exclusive, and it was eventua-IIy decided that the
particular speciesrepresented in the Graeanog remainsoffered no
possibility of distinguish-ing betsreen habitats D2 and D3,
forexampl-e. This dilemma is perhaps appa-rent from the few
examples of 'Dr clas-sification offered in table 3 (above) ,and in
table 1I (below) in which class-es D2 and D3 have been
amalgamated.
It will be apparent that this step(4) makes no use of
ethnographic nodels.Step 4 is included in this paper mere-Iy
because its omission wou.Id have lefta gap in the logical sequence
of analy-^l ^
27
-
The following operations (5.2 Lo 6;4\were next applied to each
sample sePa-rately:
6.2 AMAIGAMATION OF SCORES OF THOSECHARRED REMAINS ASSIGNED TO
THE SAMELLAbS
At the site of Cefn Graeanog, the char-red remains were
classified as outlinedin the preceding section (6.1). Withineach
sample of charred remains, the'scores' (numbers of items) of
allthose taxa assigned to any one classunder the A, B, C or D
classificationsystems $rere next amalgamated to give'class totals'
as indicated in tables9, l-0 and 1l (below) .
Reasons for amalgamating scores:At Cefn Graeanog, most taxa
(consideredindividually) \.,ere present in so fewsanpl-es that, in
isolation, there wasIittle sense in using their pattern
ofoccurrence to indicate differences inhuman manipulations of the
wild ordomestic plant products concerned.This problem of 'patchy'
results is notunusual- on small, farmstead sites, andat Cefn
Graeanog it was overcome onlyby amalgamating (within any one
sample)the scores of those taxa or chaff clas-ses which, on the
basis of present-dayparallels, coul-d be assumed to haveshared a
common relationship with anygiven hurnan activity. In other
words,in each sample, we amalgamated thescores of all those taxa
assigned toany one of the A, B, C and D cl-assesoutlined above.
These amalgamations had two principaleffects: a) Any one class
(within whichscores were amalgamated) was represen-ted in many more
samples than rdere anyof its constituent taxa. b) The classtotals
for any one sample were, ofcourse, much larger than the scores
forindividual taxa. As a resul-t, differ-ences between
archaeological samples inrespect of cfass totals were far
more1ikely to reflect genuine differencesin human activities. In
interpretingthe composition of samples of charredremains in terms
of human activities,we therefore used class totals and notthe
potentially misleading scores ofindividual taxa and chaff
types.(Note: the scores of in
-
-a-ble 9. CEREAL TOTALS AT CEFN GRAEANOG: AI'IAIGAI4ATION BY THE
CTASS OF3OP PRODUCI IN WHICH EACH COMPONEI.M IS }iORMAI,LY FOUI\D
IN TTIE PRESE}TI-DAY:
..\IVPLE OF IYPICA], TUIALS SIT,ET FOR A SI}GI,E
SAI\.{PI,Elkrte: The arnalgarnations in this table are repeatd for
each salrple seperately.
The only cereal crrqDneJrts included in these srsnnations are
tltose r,vhich are;-:rerally abr:ndant in onlv one of tie major
crop products ccnfitcnly preserved by::.arring. Class 84 is
crn-itted here as it is equiv-alent to a ccrnbination of:.asses E}5
and 86.
class of product in which each conponent is cqmpnestlo save
spa.ce, each cdnponent is listed by only its ccrqruter code as
given in fig.s)
B1-----@ing mste
_=-Jaw mste' )pres-nt
'cavirgs' frm@se saevugnG.
-stB3
clmhgs frmfiae siwing
rcs.
.qtB5
claings frmtEnd sorting st
B6Iea prire gr.3adv for mkY
no.pres-slt
-:eys
l423
Irt134lr;tfqI t"-arl4el4r143
71
ng
Hg
>o(
!^,trts { 18[20sals '15irtlet.
gn
n
!*lats
30
(44*tt"r=tt:
ggg
g
l.lone foudcefn Gramog
30
144*""r1X:
E9rgggs{HgH
x
ggg
:: -ss one)o( )o(
.:-j.ng the 'Br sub-totals, and these,-:- turn by adding 'C'
class sub-totals.:.-.: 'C' class totals were, however,:::ained by
adding the sub-totals from::-umns of similar 'C' cl-ass from each-:
tbe rBr clusters; i.e. we added the:jo-totals from each of the C2
columns;-:
-
TABLE 10- SCORE TOTALS OF WILD AND SEGETAL SPECIES FROM CEFN
GRAEANOG: MLGAIIIATIONS BY HUMAN ACTIVITY ASSOCIATIONS (i.e. BY
TTBA, B AND C CINIIICATION OF Tm 3) : LAYOUT oF SCORE TOTALS SHEET
FOR A SINGLE SAJ'IPLE
As before, the only species Eed hse to @ntrj-bute to cl6s tobls
Ee tho* for wtrich onlv one of t'tE alt@tive classifitions Es
probable within eachof tie systs of classificlion (i-e. witldn ey
one of tie @lrJlffi A, B ild c in table 3).witlin my one sdrq)le,
ja producirg grard tobls for aci of the 'B' clcss tie rele\mt 'B'
Ebls frm t}le -Is in tible 9 de gosally added to t}|e@rcpording
tobls in this table. CI6s 'AI*' dpries species of prcbable segetal
sbtE but of metuin clGsifi@tion in ttle 'B' and 'C' systs.ltE A14
total. is added Co tlE 'A' clds qand total as ijrlited. 1\l 9@ sFe,
htjn l]ares have ben abbleviated-
CLASS A1 Al* A3 A5 A6B1 B3
c2 cl n( c2 no
*
*
hth4is@tuIa
NeE fatElcpts.67-70) g
g
*
(n
f!o
l m+lE@tmples re@vn Cefn G!a@
rl
fedrcg)
EagHgg
H
BB
::d" lo!l@rs I n- sh@t Lipsl nki@ bua-l-f tms ls-s!rct tips I
IEac cinq-seeos I s-flms lakie
-
--e 11. 'D' CLASS TOTALS OF SPECIES OF UNCERTAIN MODE OF ARRIVAL
ONTO.: SITE (iC. UNCERTATN 'A' CLASS): AMALGATITTON BY HABITAT (D)
CLASSES
---- --Lr- --^1----!^^ the scores of smcie^ - ^r*.r1^- L^Lr+-*
rfefefenCe WhiCh afe_.-J Lrurs 4rdrYdL6LCs
-- -F_--_5 U! slL[Idt ll(ruI@L I- -iikelrz f. hA\/F Fecn
qcocf:lq (urcdq nf crnnq] :nd- or which no other Lme+rj-vocal 'A'
system classification was possible..,--.: The habitat (D) classes
used here are outlined in 6.1, st4. On accor:nt of the'.,--:osition
of plant ccnnnLrnities in the area of the site today, jt seeried
poinl---r'i to assign species to classes D2 or D3 separately. These
two classes were
-, -^c^-^ ^-^1 ^^-^+^r
-^ - h-Li !-+ ^r 1--^- i n urhi nh o:nh cmai oc ..,-- r ; L^r,.
+^ h1r'^ h-.'- rldulLdL uldbbeb
---.. -F,-,.-.j WaS llKely t'O nave Deen gfo\^/Ing'- fho
rriciilifrr nf fL^ -rt^ ^. ^-.^ ^--^^nm drrrina itc nriain:l
ncnrml-innlLJ v' rrc srLe ur LEr rr uf ocaruj uu frrg f Lr urrjlrar
uLLutEurvrr
IDl I nos.;c
^f ..*^ 1.-^ l---^-rlvl,-crrL
D2&D3pasture & wdlarld spp
nos. D4ma-rsh & bog spp.
nos.
-rtica dioica1: l l n+r ni arr
)oo(>o{>o