Markus Hiekkanen & Tapio Seger BEYOND POST -HOLES: AN ...

Fennoscandia archaeologica V (1988)Markus Hiekkanen & Tapio SegerBEYOND POST -HOLES: AN INVESTIGATION OF PRE-ROMAN HOUSE REMAINS ATMICKELS IN ESPOO, S. FINLANDAbstractThe starting-point of this article is the problem of defining the floor-plan of a prehistoricdwelling investigated in trial excavations at the site of Mickels in the city of Espoo,S.Finland. Reconstructions of plan on the basis of post-holes are not always convincingand with the lack of them are altogether impossible. At the Mickels site the distributionof daub from the walls of the building was used in connection with methods of recordingand documentation that were more detailed and precise than normally. Statistical andcartographic methods were used to outline the form of the walls with the result thatthere was originally a four-sided construction (or constructions) at the site. Ceramicfinds date Mickels to the Pre-Roman Iron Age and the site was also a locus of contactfor ceramic traditions of the coastal regions and the inland.Markus Hiekkanen, National Board of Antiquities, Department of Monuments and Sites,Box 187, SF-OOl71 Helsinki.Tapio Seger, National Board of Antiquities, Section for Prehistory, Box 913, SF-00101Helsinki"Snart sagt i alia tider tycks det ha fun nits baderunda, fyrkantiga och rektanguliira hus, och dethar verkligen inte sa stor betydelse vilketderasom iir fallet.,,1 - Mats P. Maimer (1976 43).INTRODUCfIONExtensive trial excavations were conducted in1978 at the prehistoric site of Mickels (Fi.Mikkelii) in Espoo due to planned housing construction.The site is approximately one kilometrefrom the centre of Espoo and covers ca.700 x 400 metres in area (Fig. 1). The excavationwas conducted partly by earth-movingmachinery and partly by hand. In addition tosmaller areas of finds it was possible to locate aconcentration of daub and sherds that was densewith limited boundaries. The daub and the1 "Briefly put, there appear to have been both round,square and rectangular houses in all periods and it isreally of not much importance which of these thecase in question is."special features of the ceramics suggested adwelling site of the beginning of the Iron Age,although the elevation of the area relates to LateNeolithic shorelines (ct. Meinander 195467-68). On the basis of these results it wasnecessary to conduct a detailed excavation of thesite the following season.The trial excavation had already raised thequestion of identifying the shape of the buildingby archaeological methods. At the time archaeologicalreconstructions of the floors and plans ofprehistoric dwellings were mainly based on socalledpost-holes and these results were extensivelycriticized (Meinander 1976; ct. Matiskainen& lussila 1984). The trial excavations atMickels did not reveal any features that indicatedpost-holes. In planning the 1979 excavationsit was necessary to employ a differentapproach. The basic question in this respect was,what happened to clay lining and daub when thedwelling or building was burnt; in archaeologicalterms only ceramified daub remains in additionto soot and charcoal where these are preserved.This led to the basically simple assumption that21

within 2 x 2 metre excavation squares. Thismethod is naturally a painstaking one and thereis not much reason for its use unless there aredefinite aims and goals set for the excavation.THE MICKELS SITE, EXCAVATIONRESULTS AND FINDSLocation of the Mickels sitein a fire most of the daub would fall by the walls,although definite sources of error must be takeninto account, such as the falling of the walls inwardsor outwards, the construction of a newbuilding on the ruins, recent disturbances etc.Accordingly it could be possible, at least in principle,to outline the shape of the walls and thusthe whole dwelling through statistical and cartographicmethods.This in turn led to the question of the fieldmethods to be applied. The chief aim was to recordthe finds with such precision and detail thatstatistical noise could be minimized. The systemof recording and measuring had to be flexibleand of a routine nature, as the schedule for thesalvage excavation was limited. Of the standardtechniques, the conventional 2 x 2 metre squareand even the alternate 1 x 1 metre square weretoo large. On the other hand, the measuring ofcoordinates for each find of daub would havebeen too laborious and entailed the possibilityfor error. A solution was reached whereby all ofthe finds at Mickels were recorded and documentedin 10 x 10 centimetre units (or cells)The site of Mickels is on a southern inclinegently sloping towards the Espoonjoki River.According to a map from 1698 by the surveyorSamuel Broterus (Archives of the NationalBoard of Survey, Helsinki) the site was undercultivation already at that time. Precise levellingcarried out in 1978 (Hiekkanen 1978; 1979)showed the dwelling site to be on a small hillockrising slightly from the surrounding terrain (Fig.1). This is a lenticular feature of fine yellow sandmeasuring ca. 30 x 40 metres and surrounded byclayey soil. The sand changes into clayey soil atan elevation of 18.1 metres above sea level onthe side of the river. The dwelling site extendedto only the south-east part of the sandy featureas shown by the distribution of finds and culturallayer.The cultural layer was patchy and relativelythin (15-20 cm). Along with clay daub twokinds of features were discovered - hearths andrefuse pits. There were two or three hearths dependingon whether one can interpret two adjacentand partly concentric concentrations ofstones as one or two hearths. One of the hearthswas approximately 8 metres to the south of thedwelling in the clayey zone. It was of round regularform and approximately one metre in diameter.The stones were fire-cracked with soot inbetween them.A total of six features interpreted as refusepits were found. Of these three were in the areaanalyzed in closer detail and marked in the mapin Fig. 9. The refuse pits were in different partsof the site, but most of them appear to havebeen outside the dwelling. The pits measured1-2 metres in diameter and 0.3-0.5 metres indepth (measured from the boundary of topsoiland cultural layer). Containing a large amountof soot, the pits were clearly distinct from thesurrounding light-coloured and broken culturallayer. The pits also contained more fragments ofceramics and daub than the surrounding areas,which was especially evident in a large refuse pitin the north-west part of the excavated area (Fig.9). The finds from the site include ceramics,daub, stone artefacts, quartz flakes, fragmentsof burnt bone and a flake of flint. Table 1 shows22

Fig. 1. Areas excavated at the Mickels site in Espoo in 1978 and 1979. The map shows trial trenchesopened in 1978, the area excavated in 1979 and the sand/clayey soil boundary. Contours at 20cm intervals. Drawn by P. Honkanen, T. Seger & M. Hiekkanen 1978-1979. City of EspoolNational Board of Antiquities.23

Table 1. Frequencies and weights of the various find categories of the investigated area.DAUBCERAMICSN Grams N GramsTopsoil2268 7690 602 1491Cells 1701 7501 364 1284Total 3696 15191 966 2775QUARTZ FRAGMENTS BURNTOF TOOLS BONEN Grams N Grams N Grams21 262 3 447 3 1,510 350 5 1115 10 731 612 8 1562 13 8.5the distribution of finds between the topsoillayer and the cultural layer. It can be noted thatthe mean weight of the fragments of pottery anddaub from the topsoil layer was considerably lessthan that of corresponding material from the culturallayer. Centuries of field cultivation hadcrushed the sherds in the topsoil and they hadalso been affected by weather. The excavationsalso revealed that despite the working andploughing of the field the topsoil finds displayedlittle lateral translocation. This was demonstratedby the refuse pits; the composition of thetopsoil was clearly anomalous at the locations ofthe latter. Also in the topsoil there were considerablymore potsherds and fragments of daubat these locations.With the exception of ceramics the finds donot display any marked features. Quartz wasfound only in flakes and it was impossible to analyzethe fragments of burnt bone due to theirFig. 2. Pottery of Bronze Age type from Mickels withpit decoration. NM 20602:171. Drawing by R.AholalNational Board of Antiquities.minute size. The stone artefacts are mainly largehand-held grinding and polishing stones of irregularform. A blade fragment of an adze wasalso found which may be classified among thegouges of the so-called Kiukainen type (Meinander1954a 102-103).The potsherds found at the site are all of postneolithicdate and fall into two main groups. Inthe first group the temper consists mainly ofcrushed feldspar along with other lithic materials.The vessels were of even base and in manycases they had a distinct neck part. The rim is ofeven width and is often bent slightly outwards.The surface is smoothed ("scratched") either onthe inside or the outside with what appears tohave been a bunch of straw (cf. Jaanusson 198143): in some of the sherds the outer surface hadweathered away, a feature evident in ceramicsof the same type elsewhere (e.g. Edgren 196922). At least one of the vessels was made on atextile surface as indicated by impressions on abase sherd. There are no textile impressions onany of the wall sherds.The vessels were sparsely decorated and ornamentationwas limited to three parts: below therim , the widest part of the vessel and the uppersurface of the rim. The motifs applied to thesurface consisted of different stamp impressionssuch as short comb and oval impressions andpits. In two of the vessels there was a single rowof pits below the rim (Fig. 2) which can be interpretedas a feature of Bronze Age pottery (Meinander1954b 171-173). In at least two of thevessels the widest part is covered by a banddesign possibly of comb stamps with rows of pitsbelow and above it (Fig. 3). In some cases thedecoration displays a rhythmic application ofmotifs (Figs. 4 & 6).The upper surface of most of the rim sherds isdecorated with irregular impressions that arehard to interpret in any further detail (Figs. 2,4-6). There are also rim sherds with distinct pit24

Fig. 3. Morby Ware from Mickels. Decoration consists of pits, comb stamp impressions and indistinct impressions.NM 20602:461,850. Drawing by R. AholalNational Board fo Antiquities.impressions (Fig. 2). Decoration of the uppersurface of the rim can also be found in vesselsthat are otherwise undecorated (Fig. 5).The temper, form, surface technique and decorationof most of the pottery of this group permitits classification as Morby Ware, dated to thePre-Roman Iron Age (Meinander 1954b173-177; 196940-47). Some of the sherds canbe defined as being of Bronze Age type on thebasis of their pit decoration (Meinander 1954b171-173).Differing completely from the above are approximatelytwenty extremely thin sherds withsmooth surfaces and finely crushed asbestos temper.Two of these are decorated. One of the decoratedspecimens, a rim sherd (Fig. 7), has adesign below the rim consisting of two diagonalcomb stamp impressions above a horizontal banddrawn with a comb stamp and with a pit impression.The upper edge of the other decoratedsherd (Fig. 8) displays part of the above bandbelow which are four vertical comb stamp impressionsnext to each other. In the lower partof the outermost impressions are two diagonallyincised comb stamps. - This asbestos-temperedpottery belongs to the so-called Luukonsaarigroup of the inland regions of Finland and wasin use from the latter part of the Bronze Age tothe Roman Period of the Iron Age (Meinander1969 59-63). A few sherds of Corded Warewere found in the northernmost trench at an elevationof 25 m. a.s.l.The ceramic material from Mickels indicatedtwo different traditions. The first ceramicFig. 4. Morby Ware from Mickels. Decorated withpits (note grouping) and indistinct impressionson the rim . NM 20602:197. Drawing by R.AholalNational Board of Antiquities.Fig. 5. Ceramics with scratched surface from theMickels site. Decorated with indistinct impressionson the rim . NM 20602:771. Drawingby R . AholalNational Board of Antiquities.25

, : 1Fig. 6. Morby Ware from Mickels. Decorated withrhythmically applied "cat's paw" impressions.NM 20602:1035. Drawing by R. Ahola!National Board of Antiquities.tradition is that of Morby Ware and the closelyrelated West Finnish Bronze Age of the coastalregions and connected with early farming populations(Meinander 1969 48-55). The secondtradition is that of the Luukonsaari group, inturn connected with the hunting-fishing population(Meinander 196955-60).The site is a rarely encountered locus of contactbetween the coastal and inland regions (seee.g. Meinander 1954b). A site of this kind indicatingexogamic practices may also provide suggestionsconcerning the mechanisms by whichslash-and-burn cultivation spread at this time tothe lacustrine regions of Harne in the inland (cf.Donner 1984; Tolonen 1984).As mentioned above, the elevation of the sitecorresponds to late Stone Age shorelines. Thedatable artefacts on the other hand are clearlyFig. 8. Luukonsaari Ware from Mickels. Decoratedwith comb stamps and drawn comb impressions.NM 20602:1275. Drawing by R.AholalNational Board of Antiquities.of the early Iron Age, which must be regardedas the decisive criterion for dating. During itsperiod of use the site of Mickels was not on theshore, but at a distance of several hundred metresfrom Espoonlahti Bay with an interveningarea of clayey soil - an indication in turn ofmeans of livelihood based on cultivation andhusbandry.ANALYSISAs no post-holes were observed in the excavation,the only way of seeking an answer to theproblem of the plan of the dwelling(s) at the sitewas through analysis of the distribution of theFig. 7. Luukonsaari Ware from the Mickels site. Decoration consists of combstamps, pits and drawn comb impressions. Rim undecorated. NM20602:847. Drawing by R. Ahola!National Board of Antiquities.26

, ,, ,, ,, ,, ,, ," 11 , ,,, ,, ,I I I I"' ,, , ,,",,, ",,, , , , ,, ,",, , ",, , ,," , , ,,",", , ,, " ,",--,.....",/' Il_j"'_/ ", ,,,,,,, ,, , , ,,", I,",,, ,",, , ,,,, :... "............. :,:' :1 ~ .. ", .,.\~, ,",'",": ,','" ,, :, ,"" ,, , ,II1 1 I 1,Fig. 9. Unweighted distribution of daub. Cells with daub marked with the figure 1. The map also shows refuse pits(unbroken lines), three stone settings (dashed lines) and two recent disturbances (dotted lines). Of the latterthe one to the west is a contemporary pit for slaughter waste and the one to the east a disturbance apparentlymade by the back wheel of a tractor. Grid coordinates (266/692) marked in the SW corner of themap. Drawn by T. Seger.finds and especially the fragments of daub. Thedetailed analysis was limited to the above-mentionedclearly demarcated area (ca. 100 m 2 ) withits concentration of daub. The frequencies andweights of the find categories in this area aregiven in Table 1.This procedure entails several potential sourcesof error:(1) The investigated area was a field where thetopsoil had been disturbed by ploughing.(2) Especially the largest of the refuse pits indicatedtwo phases of construction. This featurecontained dark discoloured soil, sherds, quartzflakes and a large amount of burnt daub, whichsuggests the possibility that after the first (?)stage of construction had been destroyed by firethe remains were cleared into the refuse pits.(3) It can be assumed that upon falling down, awall of wattle-and-daub does not leave a trace inthe ground that is straight or perfectly circular.The topsoil finds were excluded from theanalysis. On the other hand, the underlying culturallayer appeared to be untouched with theexception of a couple of recent disturbances thatwere small in area and clearly limited (Fig. 9)and analysis concerned this layer.At the outset it can be thought that the analysiscould be based on a simple points distributionfor each of the finds. In the planning of the excavationit was decided however to record all ofthe subtopsoil finds in 10 x 10 cm units or cells.This technique has the distinct advantage of permittingquantification in units by both frequencyand weight which in turn facilitates more variedstatistical approaches.At first the frequency and weight of daub fragmentsof all of the excavation layers of each cellwere summed. This provided quantitativemeasures to be used as variables in analysis. Thequestion that arose in this connection was whe-27

-1+ 1-------........ 000000 00 o 00 o1+ 1----.... 0 0 0 0+----+- ---+- - --+ ----+ - ---+- - --+- - --+- - --+- - - -+- ---+- - --+ - - --+ - - --+----+ G20. 40 . 60. 80 . 1c])0. 120.Fig. ZO. Distributions of daub (upper) and ceramics (lower) in cells presented as box plots. The axis indicatesgrams, the + symbol the median; the l's hinges and the ·'s and O's outliers. Drawn by T. Seger.ther to use counts of frequency or weight as therelevant values.For this reason the correlation between thenumber of daub fragments and their weight wascomputed and it turned out to be significantlynegative (r = - 0.081847847, N = 627). Thisshows that in the case at hand weight is a morereliable criterion of distribution that the numberof fragments and also demonstrates the reasonwhy a points distribution map would not havebeen suitable. The decisive factor is naturally thevolume of daub in each of the cells and not itsdegree of fragmentation (see also Orton 1980156-178).The simplest possible graphic presentation ofdistribution employing computer methods is amap based on a binary code where the zero indicatesempty cells and the figure 1 cells containingdaub (Fig. 9; zeros deleted). The map also showsthe refuse pits in the area concerned, three smallstone settings and the above-mentioned recentdisturbances. The overall impression of the mapis confusing. With the exception of the large refusepit forming a distinct concentration of findsin the north-west part of the area the distributiondoes not permit one to discern any structuralshapes (the straight boundary at the south-eastcorner is due to the fact that here the excavatedarea bordered on a trial trench of the 1978 excavationwhich was not documented or recordedwith the same techniques). The above indicatedthat a mapping of unweighted distribution wasinsufficient for the set aims.A density contour map (e.g. Seger & Nieminen1983) was not suitable in this connection, either.Interpolated curves formed around theunits clearly provide no answer to the questionof the linearity or curvature of the walls of theprehistoric dwelling.The most suitable alternative appeared to be aweighted distribution map which preserves thecells as such; in other words each cell is given a+ =#I ~ I I i rl --'--'IIr------------'L-----------.......,p-1+ 1-------........ 000000 00 o 00 o+ ----+ ----+ ----+----+ ----+----+ ----+ ----+ ----+ ----+ ----+----+ ----+----+ G20. 40 . 60. 80. 10c]). 120.Fig. 11. Classes with code symbols used in the statistical analysis marked above the box plot displaying the distributionof daub. Drawn by T. Seger.28

_1-2111"2... ~.~., .. .... ., .,. . ... ...,.. ... .. .. - ....... ~: .. . • ....... p. pp- ...••• P P+. . ... .p-• top,~ ., .. ..pI..+ . •.. .p -.+,. -..+ - •.. . .. -. -;.. ...#. .... ..+ ...'mf++ ....... --+---+----;1"Fig. 12. Statistically weighted coded map of the distribution of daub. All cells with daub included. Drawn by T.Seger.certain coded designation in relation to the totalweight of daub found in it. A tabular map withnumeric values would not have been very useful- psychological studies show that only 7 to 9 figurescan be properly grasped at the same time(Miller 1956). For these reasons it was necessaryto combine the unit weights into quantitativeclasses and to mark these with coded symbols.The resulting map attemps to show whether theunits with large amounts of daub form discerniblepatterns.In order to select a suitable order of classes ofcell weights we must consider the nature of thedistributions. The weight distributions of dauband, by nature of comparison, ceramics aregiven in Fig. 10 as box plots (Tukey 1977; McGillet al. 1978; Velleman & Hoaglin 1982; Ryan etal. 1982; Seger 1985). The box plot is one of themost simple and graphic means of representationin EDA (Exploratory Data Analysis). The mainquantities considered are the median of the cellvalues, dividing the values placed in running orderinto two parts of equal number, and the socalledhinges (in practice quartiles; see Tukey1977, in turn dividing the halves around the medianinto similar parts). In practice half of all ofthe cell values remains inside the box (i .e. withinthe hinges). A quarter remains to the left of thelower hinge and the remaining quarter to theright of the upper hinge. The values indicatedwith asterisks and zeros are those differingclearly from the whole (outliers).The distributions of daub and ceramics arevery skewed and resemble each other closelydespite their absolute differences. Criteria ofclassification used e.g. in phosphate surveys suchas the mean and the standard deviation and theirderivatives are bound to normal distributionsand are not suitable in collections of materialthat differ from the Gaussian distribution to sucha degree as the present one (e.g. Seger & Nieminen1983).The methods of EDA suggest suitable criteriafor the construction of classes. A coded table29

.,+~,....+ ..+ p........ P "PP' ... .. ,:~ 'p p ...•p+ ,• +.p +• . •++++ , to + •.+-++ ,++ .,o~I~'~~---+--~--~I,_Fig. 13. Statistically weighted coded map of the distribution of daub. Only cell values higher than the upper hingeincluded. Drawn by T. Seger..pp,.0PI.pPp,OP P.,--'-- 2UIIU,­~~--~---+--~IFig. 14. Statistically weighted coded map of the distribution of daub. Only exceptionally high values (outliers) areincluded. Drawn by T. Seger.30

..p.'. ... 'ppp pp". ".p ,..p"-,/,/

•..'

Matiskainen, H. & T. Jussila, 1984. Naarajarvenkampakeraaminen asumus. SM 1984.McGiU, R., Tukey, J.W. & W.A. Larsen, 1978. Variationsof Box Plots. The American Statistician 32.Meinander, C.F. 1954a. Die Kiukaiskultur. SMYA 53.- 1954b. Die Bronzezeit in Finnland. SMYA 54.- 1969. Dilvits. En essii om fOrromersk jiirnillder.FM1969.- 1976. Kiltabottnar av Madeneva-typ. lskos 1.Miller, G.A., 1956. The Magical Number Seven, Plusor Minus Two: Some Limits on Our Capacity forProcessing Information. Psychological Review 63.Orton, C., 1980. Mathematics in Archaeology. Cambridge.Ryan, T.A.Jr., Joiner, B.L. & B.F. Ryan, 1982. MinitabReference Manual. Boston, Mass.Seger, T., 1985. Exploratory Data Analysis: "Soft"Statistics for Archaeologists. Proceedings of theThird Nordic Conference on the Application ofScientific Methods in Archaeology, Mariehamn,Aland, 8.-11.10.1984. lskos 5.Seger, T. & E .-L. Nieminen 1983. Kaivausdokumentoinnistaja loytojen levinniin graafisistaesityksistii. Karhunhammas 7.Tolonen, K., 1984. Paleoekologin pUheenvuoro. InSuomen viieston esihistorialliset juuret. Bidrag tillkiinnedom om Finlands natur och folk. H. 131.Helsinki.Tukey, J.W., 1977. Exploratory Data Analysis. Reading,Mass.Velleman, P.F. & D .C. Hoaglin, 1981. Applications,Basics and Computing of Exploratory Data Analysis.Boston, Mass.ABBREVIATIONSFM = Finskt MuseumSM = Suomen MuseoSMY A Suomen MuinaismuistoyhdistyksenAikakauskirja333 - FennoscGlldia