eia fowler

Upload: xxx68

Post on 06-Apr-2018

232 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 EIA Fowler

    1/16

    Ceramic discard and the use of space at EarlyIron Age Ndondondwane, South Africa

    Kent D. Fowler

    University of Manitoba, Winnipeg, Canada

    There are few methods of ceramic analysis in accumulations research that can help distinguish discretedepositional areas and evaluate previous interpretations of activity organization at archaeological sites. This

    paper introduces a quantitative method of ceramic analysis that evaluates whether the fragmentation andfunctional variation of ceramic objects distinguish depositional areas resulting from different accumulation

    activities at early farming sites. Applied to the analysis of ceramics from the 9th10th century A.D. site of

    Ndondondwane in South Africa, this study tests previous functional interpretations of the site and providesmore detailed and dynamic ones for the activities contributing to the development of spatially and temporally

    discrete archaeological deposits during the main occupation of the settlement. This approach to the study ofceramic discard at Ndondondwane illustrates both a method as well as important caveats based on siteformation and accumulations research that must be considered in attempts to interpret variation in the spatial

    organization of activities in past farming settlements in tropical Africa.

    Keywords: accumulations research, ceramic analysis, spatial organization, South Africa, Early Iron Age

    IntroductionFormation studies in archaeology are primarily con-

    cerned with two sets of processes: those leading to the

    disposal of artifacts and those that have affected the

    nature and distribution of artifacts after they wereburied. Accumulations research aims to understand the

    first set of processeswhy and how materials accumu-

    late in the archaeological record (Varian and Mills

    1997: 141). Ceramic artifacts have played a central role in

    accumulations studies because they are abundant,

    durable, and found in most deposits at sites occupied

    by pottery-producing cultures. Accumulations studies

    have been instrumental in defining discard and disposal

    practices associated with food producing societies and

    the longer-term occupation of sites (Santley and

    Kneebone 1993). Such studies have been stimulated bynearly four decades of ethnoarchaeological research on

    ceramic use life and discard (Beck 2006, 2009; Beck and

    Hill 2004; David 1972; DeBoer 1974; DeBoer and

    Lathrap 1979; Longacre 1985; Mayor 1994). One

    dominant concern in this line of research is whether

    ceramics are adequately diagnostic of the activities

    that led to their accumulation in archaeological deposits.

    Two of the most influential lessons from ethnoarch-

    aeological research concern how the frequency of

    functional ceramic types at a site may be used to infer

    past activities. First, we cannot assume that the discard

    of different types of vessels in a particular location

    resulted from the activities that occurred at or near it

    (David 1972; David and Henning 1972; Sullivan 1989).

    Thus, neither the presence nor the relative frequency of

    cooking, storage, and transport vessels can be used forsimple reconstruction of past activities. These interpreta-

    tions must be correlated with other evidence for activities

    in a specific location and across a site. A second lesson is

    that the frequency of different functional types of vessels

    is useful for inferring the intensity and duration of

    activities at an archaeological site (DeBoer 1974; DeBoer

    and Lathrap 1979). Consequently, archaeologists have

    used ethnoarchaeological research to understand site-

    wide formation processes and infer occupation spans,

    population sizes, and mobility patterns (Frankel and

    Webb 2001; Gallivan 2002; Hardy-Smith and Edwards2004; Mills 1989; Rosenswig 2009; Shapiro 1984; Varien

    and Mills 1997; Varien and Ortman 2005).

    This study is concerned with the first issue raised by

    ceramic ethnoarchaeological research: that ceramics

    alone are not valuable for reconstructing past activities.

    As Shott (2006: 2) has stated, the archaeological record

    is not a faithful, complete depiction of the cultural past

    but a systematically refracted one. However unfaithful

    the archaeological record may be, all sites and the

    deposits within them are not created equally. Although

    we can identify variation in the nature and distributionof artifacts, it is unclear whether this variation is useful

    for distinguishing different disposal practices and dis-

    turbances that can be used to more accurately interpret

    Correspondence to: Kent D. Fowler, Department of Anthropology, 435Fletcher Argue Building, University of Manitoba, Winnipeg, Manitoba R3T2M5 Canada. Email: [email protected]

    Trustees of Boston University 2011DOI 10.1179/009346911X12991472411286 Journal of Field Archaeology 2011 VOL . 36 NO. 2 15 1

  • 8/3/2019 EIA Fowler

    2/16

    past behavior (Needham and Spence 1997; Wilson

    1994; Wilson et al. 1991). It is interesting, therefore, that

    studies of ceramic accumulations have not been used

    to test hypotheses generated from the analysis and

    interpretation of other artifacts. Any method designedwith this goal in mind must evaluate linkages between

    the archaeological assemblage (ceramics recovered

    archaeologically), the disposal assemblage (ceramics

    deposited in a location), and the behavioral assemblage

    (ceramics used in a particular behavioral setting, such as

    a household) (Pauketat 1989). The analysis of any one

    of these linkages necessarily incorporates other data

    concerning artifact recovery methods, depositional

    contexts, and the behavioral interpretations that definepast activity settings, such as households.

    This study investigates whether patterns of ceramic

    accumulation are adequately diagnostic of different

    activity areas at farming settlements in tropical Africa.

    Figure 1 Site plan of Ndondondwane.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    15 2 Journal of Field Archaeology 2011 VOL . 36 NO. 2

  • 8/3/2019 EIA Fowler

    3/16

    To test the validity of this approach, a method of

    ceramic analysis that uses patterns of ceramic fragmen-

    tation and the distribution of functional types to

    identify variation amongst deposits at a site is

    employed. The method involves a synthesis of analy-

    tical concepts and techniques used in studies of

    archaeological ceramics in Europe and the Americas.Co-varying patterns of fragmentation and the propor-

    tion of functional types are then used to evaluate

    hypotheses about the activities that led to the accumu-

    lation of deposits in different settings at an early

    farming settlement. Data from the site of Ndondond-

    wane in eastern South Africa (FIG. 1) are used to

    illustrate this approach. This particular settlement is

    well-suited for this analysis because it was occupied

    only once between the late 9th and early 10th centuries

    A.D. and patterns of artifact disposal and site formation

    processes have been studied and are well-understood

    (Fowler 2002; Fowler et al. 2000, 2004; Fread 2007;

    Greenfield and van Schalkwyk 2003; Greenfield et al.

    2000; Singer 2008). Most farming settlements in south-

    ern Africa were occupied repeatedly or for long

    durations of time (Huffman 2007; Maggs 1976, 1980).

    Ndondondwane holds great potential to help under-

    stand the relationship between social and spatial

    organization in early southern African farming com-

    munities because it is less disturbed than other sites

    repeatedly occupied by early farmers.

    The main goal of this study is to evaluate whether

    this method of ceramic analysis can identify discrete

    depositional areas and distinguish disposal practices

    at Ndondondwane to test previous functional inter-

    pretations of the site. An additional aim is to consider

    the applicability of the method beyond this particular

    case. Structural remains are poorly preserved at many

    early farming settlements in the tropical regions

    throughout eastern and southern Africa, but cera-

    mics are ubiquitous (Eggert 1993; MacLean 1994;

    Mercader et al. 2003; Reid 1994; Schmidt and Childs

    1985; Van Noten 1979).

    MethodsWhen conducting intrasettlement spatial analyses, the

    main problem lies in identifying recurrent patterns in

    the spatial distribution of archaeological remains at

    different areas of a settlement. From a ceramic

    perspective, these remains include raw materials (clay,

    temper), production facilities (e.g. firing areas), debris,

    containers, other tools (e.g. tuyeres), and sculptures

    used during the occupation of a settlement (Pool 1992).

    The method of ceramic analysis proposed here

    identifies distinct depositional sets of ceramics andassociated cultural remains in spatially discrete areas of

    a settlement (Carr 1984: 114). There are two main stages.

    The first generates quantitative sets of ceramics from

    different excavation areas described in terms of their

    fragmentation and the range of use-categories repre-

    sented. Each of these variables can be presented as

    indices of ceramic fragmentation and functional

    diversity for each depositional area. The second stage

    searches for patterning in the spatial distribution of

    these sets (South 1978a, 1978b). During this stage,

    index data for spatially discrete deposits are comparedto evaluate linkages amongst archaeological, disposal,

    and behavioral assemblages and assess previous

    functional interpretations of space.

    Pattern-recognition procedures help identify a broad

    spectrum of ceramic variability. Using these proce-

    dures aids archaeologists in distinguishing previously

    unknown functional patterns and, thus, encourages an

    understanding of the variability in the organization of

    settlements. Another advantage of this method is that

    a broad range of variability in structural and artifactual

    data is considered before functional types of areas are

    defined. This increases the opportunity to infer acti-

    vities, detect changes in their organization over time,

    and identify those that were either were not considered

    in previous interpretations or do not have any pertinent

    ethnographic parallels. The resulting model of cera-

    mic use and discard, when combined with structural

    remains, provides independent data to aid in the inter-

    pretation of site organization and function where

    structural remains are less well preserved (cf. Gallivan

    2002). In this study, several techniques were used in the

    pattern-recognition procedures, each aimed to distin-

    guish spatially discrete depositional areas in settlements

    based on unique types, quantities, and distributions of

    ceramics. A five-step method for deriving data sets and

    identifying and comparing ceramic fragmentation and

    functional patterns at different areas in a settlement is

    outlined below.

    Step 1: dataThe first step involves screening appropriate data for

    analysis. Construction materials, remodeling, and post-

    depositional processes influence the preservation of the

    built environment. Unlike ceramic studies in regions thathave an excellent preservation, it must be determined

    whether ceramics are adequate indicators of activities

    when the precise size of structures and nature of other

    architectural features are unknown or poorly under-

    stood. This analysis focuses on the variation in the

    distribution of ceramics found within, nearby, and away

    from structural remains. Variation in artifact distribution

    may be a product of temporal differences in the use of

    areas, differential patterns of artifact disposal, and

    differential preservation. Each of the sources must be

    considered before we can conclude that the variation

    identified is a product of how areas were used.

    Step 2: quantificationThe second step involves utilizing appropriate quantifi-

    cation techniques designed to describe the proportion of

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    Journal of Field Archaeology 2011 VOL . 36 NO. 2 15 3

  • 8/3/2019 EIA Fowler

    4/16

    ceramic types within an assemblage. In selecting quantifi-

    cation and analytic techniques, I heeded Schiffers (1987:

    279) call to use simple variables such has total quantity,

    ratios, and frequency distributions to supply insights into

    formation processes. Ceramic data are presented and

    discussed in two ways; the first unit is the number of

    sherds. A sherd is either a single fragment of pottery or agroup of fragments that can be conjoined or reasonably

    demonstrated to come from the same vessel based upon

    a range of shared attributes (comparable decoration,

    fabric, residues, rim diameter, etc.). Sherd counts are

    therefore equivalent to the number of individual speci-

    mens (NISP) commonly used in faunal analyses.

    The second unit is an estimated equivalent of the

    total number of complete ceramic objects represented

    in the assemblage. Two types of equivalents were

    combined into one measure in this analysis. The first

    type is an estimated vessel equivalent (EVE) (Orton

    1985; Tyers and Orton 1991). An EVE assesses the

    part of a vessel form that is a proportion of the

    whole. An element of a vessel that represents a

    portion, such as a rim, base, or handle, has to be

    measured. In the following analysis, I based EVE

    calculations on rim sherds, or rim-EVEs. Because the

    rims of Early Iron Age (EIA) vessels are circular, they

    are easily identifiable in assemblages and appendages

    (e.g. handles) are exceedingly rare. Calculations of

    rim-EVEs are based upon a measure of the propor-

    tion of rims present for different types of vessels.

    Rim-EVEs can be presented as 1.0 (whole) or lessthan 1.0, or as percentages (e.g. 5 EVEs5500% or five

    estimated vessel equivalents).

    A different measure is needed to determine the

    frequency of other kinds of ceramic objects. A

    variation of the minimum number of individuals

    (MNI) concept was developed for this purpose. The

    estimated objects represented (EOR) measure estab-

    lishes frequencies of figurines, sculpture, beads, and

    other objects based on the total number of parts (such

    as heads, arms, torsos, etc.) in the assemblage. The

    sum of each part for an object type is therefore

    equivalent to a MNI for that type (i.e., five left arms of

    figurines equal five figurines). To refer to the frequency

    of all ceramic objects, both EVE and EOR data must

    be combined. When an EVE is rounded up it is

    equivalent to MNI (an EVE for a rim sherd between 1

    Table 1 The proposed activities and ceramic data for all spatially discrete settings during the main occupation ofNdondondwane.

    Excavation

    area

    Activity area/

    features Inferred activities Ns* Ne Nf RI CI

    Dung Area Livestock Zone Livestock enclosure. Animal penning,

    large accumulations of cultural debrisalong stockade walls.

    490 84 6 60 171

    Human Activity Zone Outside upslope entrance to livestock

    enclosure. Iron forging, food preparationand consumption, livestock tethering.

    4612 216 9 90 47

    Mound Area Ovaloid structureand open midden.

    Wild animal processing, ivory bangle manufacture,mens assembly area, site of initiation ceremonies(ritual paraphernalia including polished

    bones and ceramic sculpture).

    6086 183 8 80 30

    Transect 1 Open space.Low-density

    scatter of debris.

    Possible location of communal gatheringsand grain and root crop gardens.

    570 13 5 50 23

    Transect 2 Open space.Low-densityscatter of debris.

    Possible location of communal gatheringsand grain and root crop gardens.

    104 1 1 10 10

    Midden 1 Domestic complex Composite of ceramics within between

    and features of a domestic area

    1066 89 8 80 83

    Midden A1 Open midden 115 28 3 30 243

    Midden A2 Fire pit 104 5 3 30 48Midden A3 Open midden 11 0 0 0 0

    Midden A5 Open midden 48 8 1 10 167Midden A6 Open midden 50 12 3 30 240Pit 1 Grain storage, domestic refuse{ 43 6 4 40 140

    Pit 2 Grain storage, domestic refuse 85 2 2 20 24Pit 3 Grain storage, domestic refuse, infant burial 88 1 1 10 11

    Midden 2 Domestic complex Composite of ceramics within and between

    features of a domestic area

    7 1 7 70 143

    Midden 3 Domestic Complex Composite of ceramics within and between

    features of a domestic area

    4933 450 10 100 91

    Eastern Midden Open midden 1523 116 3 30 76Eastern Midden, Pit 2 Grain storage, domestic refuse 13 1 10 100 53

    Western Midden Open midden 3200 332 2 20 104Western Midden, Pit 1 Disposal of single vessel 14 1 1 10 72House Area House floor 310 4 2 20 13

    * Ns5Number of sherds; Ne5Estimated object equivalents; Nf5Number of functional types; RI5Richness Index score;

    CI5Completeness Index score.{ Domestic refuse typically includes organic debris (e.g., bone), pottery, grinding stones, and ash.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    15 4 Journal of Field Archaeology 2011 VOL . 36 NO. 2

  • 8/3/2019 EIA Fowler

    5/16

    and 2 is rounded to 2). By transforming EVE data in

    this way, both EVE and EOR frequency measures are

    equivalent. This combined frequency measure is

    referred to in the following analyses as estimated

    object equivalents (EOE).

    Step 3: fragmentation and richness indexes

    Both sherd counts and estimated object equivalents

    provide the quantitative data required to generate

    two indices that can be used to compare variation in

    the fragmentation and range of functional types that

    occur in deposits across sites. Generating indices of

    ceramic fragmentation and functional variation is the

    third step in analysis.

    Fragmentation is a proxy measure of preservation

    because fragmentation results from breakage rates in

    the systemic context and post-depositional factors in

    the archaeological context (Shott 1989). Both influ-

    ence the analysts ability to identify morphological

    types and use-wear traces. A study of ceramic

    brokenness provides a way to evaluate how complete

    and identifiable objects are in a deposit (Orton 1985;

    Schiffer 1987: 282).

    Sherds counts divided by vessel-equivalents (bro-

    kenness following Orton 1985) and vessel-equivalents

    divided by vessels represented (completeness follow-

    ing Orton 1985 and Schiffer 1987: 282) are the most

    commonly used ratios of measures to explore site

    formation processes. A variation of the latter techni-

    que was developed by Reid (1973) to evaluate spatial

    patterns in ceramic fragmentation. The Relative

    Room Abandonment Measure compares the number

    of whole pots on room floors with the number of

    sherds in the room fill at pueblo sites in the U.S.

    Southwest. The ratio of complete pots to sherds in a

    deposit has been used as a relative dating technique, to

    discern temporal variability in room abandonment

    (Reid 1973, 1978, 1985; Reid et al. 1975; Reid and

    Shimada 1985), and to distinguish the processes

    related to pueblo abandonment (Montgomery 1993).

    In this analysis, completeness was calculated bycomparing the total number sherds in a deposit (ns) to

    the estimated number of complete objects (EOE,

    labeled as ne in TABLE 1) in the same deposit to

    generate a completeness index (CI). A score is arrived

    at by dividing the estimated number of complete

    objects (ne) by the number of sherds (ns) in a deposit

    and globally transforming all scores into whole num-

    bers by multiplying them using the same magnitude of

    ten. Low scores indicate a greater number of sherds

    present in the deposit in relation to the number

    of objects, or a lower degree of completeness. Forexample, 30 pots in a deposit with 500 sherds would

    give an index score of 6 (30/500 6 100). A lower score,

    or low completeness, indicates that a greater number

    of fragmentary objects occur in the deposit. This

    inverse relationship is why the term completeness

    and not fragmentation is used (contra Orton 1985,

    1993: 176; Schiffer 1987: 282). Higher scores indicate a

    greater number of pots in relation to sherds in the

    deposit. For instance, 50 pots in a fill of 200 sherds

    would give an index score of 25 (50/2006100). A

    higher score, indicating high completeness, suggeststhat many relatively complete objects were deposited.

    These two variablesthe frequency of sherds and

    the number of objects in the depositcan be used to

    generate hypotheses about the relationship between

    sherds and object frequencies in deposits. We expect

    high completeness to reflect less human or post-

    depositional disturbance, such as through sherd reuse

    or scavenging. In contrast, deposits with lower scores

    may have resulted from more objects being broken

    before or during discard, greater reuse of deposited

    material after initial discard, and/or significant post-

    depositional disturbance after abandonment.

    A second measure is required to assess relationships

    between ceramic discard and ceramic use. The richness

    index (RI) measure is based on the concept of diversity

    borrowed from population ecology. Diversity has been

    utilized by archaeologists as a means to link artifact

    variability to the function of sites or areas within sites

    (Hietala 1984). For instance, Kent (1999) distin-

    guished trash from storage areas at Kutse (San) camps

    in Botswana because trash areas tend to have greater

    diversity in artifact types than areas used for storage

    (i.e. where there was more de facto refuse resulting

    from caching or curation). Thus, artifact diversity may

    also be used to investigate connections between dis-

    card practices and the nature of activities that occur in

    different areas of sites.

    In archaeological ceramic studies, the diversity

    concept has been used primarily to examine varia-

    bility in assemblages as a means of inferring past

    production arrangements (Deal 1998; Rice 1981).

    Diversity refers to the range of use categories present

    in different assemblages (Rice 1981, 1987: 202203).

    As a measure of variation in a ceramic assemblage, it

    evaluates the degree of homogeneity or heterogeneity

    of different consumer assemblages within sites or

    regions. The variation is described in terms of its

    richness and evenness. The richness of the

    assemblage refers to the number of categories present

    (i.e., the number of taxonomic units, form classes,

    decorative styles, or functional types), while evenness

    is the range of variation within the categories. Only

    the richness of functional types in deposits is relevant

    for this analysis.

    An assemblage RI was developed to evaluate howthe distribution of ceramics reflects disposal behavior

    and the range and interrelatedness of activities under-

    taken in different spatially discrete areas of a site. This

    is accomplished by considering the range of use

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    Journal of Field Archaeology 2011 VOL . 36 NO. 2 15 5

  • 8/3/2019 EIA Fowler

    6/16

    categories present in different deposits at a settlement.

    Operationally, richness is evaluated as a weighted

    measure. It is calculated by dividing the number of use

    categories in each depositional locus by the total

    number of identified categories in the assemblage. This

    measure of intrasettlement variation in ceramics

    evaluates the richness of ceramic use categories foundin different areas in a settlement in relation to the total

    functional repertoire represented in the assemblage.

    Disposal areas with a wide range of use categories are

    considered to have high richness, while those with few

    categories have lower richness. To arrive at a richness

    score, the number of object equivalents (ne) is divided

    by the number of functional types (nf) and the score is

    transformed into whole numbers. High richness

    indicates that many different use categories of

    ceramics were deposited, while low richness scores

    indicate that fewer categories were deposited. It cannotbe assumed there is a one-to-one correlation between

    richness and the range of activities that occurred in an

    area. Rather, the ratio of object frequency to the

    number of use categories provides a basis for

    proposing hypotheses about the relationships amongst

    ceramic use, area function, and disposal practices.

    The number of sherds, completeness scores, and

    richness scores are not good indices of discard be-

    havior when used alone because the number of object

    equivalents may be a function of the number of

    sherds; as the number of sherds increases, so may the

    number of object equivalents. The indices provide a

    form of quantitative ceramic data to conduct intra-

    site comparison of deposits and depositional areas,

    and only when the relative distributions of the index

    scores are compared can they be evaluated against

    other information about the depositional context.

    Step 4 and 5: intrasite comparison

    The last two steps in the procedure involve intrasite

    analysis. Step four entails generating a scatter-plot

    of index values for each deposit at a site in order

    to compare the relationship between fragmentation(completeness) and functional variability (richness).

    The fifth step in this analysis has two parts. The first

    is to determine if spatially and functionally discrete

    areas at a settlement have distinctive patterns of

    ceramic fragmentation and functional variability.

    The aim is to identify whether deposits accumulated

    from different behaviors. For instance, it may be

    asked whether the index values distinguish house

    floors from refuse pits. If so, it can be inferred that

    ceramics accumulated in these locations as a result

    of different discard practices. The second part utilizesdifferential patterns of discard to test previous

    interpretations for the function, organization, and

    abandonment of space proposed using other archae-

    ological evidence.

    NdondondwaneThe site of Ndondondwane provides a case study for

    the application of this method. Ndondondwane is one

    of several small, permanent settlements located by

    rivers and lakes that were established after A.D. 400 by

    early farmers in southeastern Africa (FIG. 1). Three

    research teams have worked at Ndondondwane over

    the past 20 years (Greenfield and Miller 2004;

    Greenfield and van Schalkwyk 2003; Greenfield et al.

    1997, 2000; Loubser 1993; Maggs 1984; van Schalkwyk

    et al. 1997). A program of surface and sub-surface

    reconnaissance and excavations has distinguished

    spatially discrete depositional areas within the single

    occupation settlement. It has also been determined that

    overlapping activities occurred in these areas during the

    occupation of the site (Fowler and Greenfield 2009;

    Fowler 2002; Fowler et al. 2000, 2004; Greenfield and

    Miller 2004; Greenfield and van Schalkwyk 2003).

    Settlement layout and chronologyActivity areas in the settlement are divided into a Central

    Zone surrounded by an arc of peripheral areas (FIG. 1).

    The Central Zone is closest to the rivers edge and is

    composed of three main activity areas arranged about

    40 m apart in a line from north to south: a livestock

    enclosure (Dung Area), a large hut floor (Transect 1),

    and an area (Mound Area) reserved for iron smelting,

    ivory working, and possibly ritual activities, based on the

    presence of zoomorphic and human sculpture. The

    Peripheral Zone forms a rough north-south arc of

    features upslope of the Central Zone. It is comprised of

    three domestic areas designated Midden 1, 2, and 3.

    These domestic areas are separated from the Central

    Zone by a large open space 60100 m across. A fourth

    area used for preparing charcoal and iron ore for

    smelting is located some 60 m from the Mound Area at

    the southernmost end of the Peripheral Zone.

    Extensive auguring and targeted excavation across

    the site revealed a cultural stratigraphy consisting of

    three cultural horizons (Lower, Middle, and Upper)

    between a sterile base and plough zone (Fowler and

    Greenfield 2009; Greenfield et al. 2005; Greenfield

    and van Schalkwyk 2003). There are no dates for

    each of the horizons at Ndondondwane for reasons

    explained by Fowler, Greenfield, and van Schalkwyk

    (2004). Radiocarbon dates from the Lower and

    Upper Horizons are statistically contemporaneous.

    Stratigraphic analysis and ceramic cross-dating con-

    firm the three pan-site horizons. Based upon other

    indicators (e.g., soil and dung accumulations and

    compactions, quantity of remains) it seems likely the

    site was occupied for little more than 60 years.

    The functional interpretations of space at Ndon-dondwane are based on the analysis of architectural,

    faunal, and metallurgical remains. The ceramic data

    from the Middle Horizon at Ndondondwane is used

    here to evaluate interpretations of the sites use and

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    15 6 Journal of Field Archaeology 2011 VOL . 36 NO. 2

  • 8/3/2019 EIA Fowler

    7/16

    organization. During this time, the settlement reached

    its maximum size and there is the greatest variety of

    deposits across the site.

    The Ndondondwane ceramicsNearly 50,000 ceramic sherds were recovered from

    Ndondondwane during five field seasons between 1978

    and 1997, one of the largest and best-preserved EIA

    ceramic assemblages analyzed in southern Africa to

    date. Some 2200 containers were identified from large

    vessel fragments and reconstructed vessels and mor-

    phological analysis of the assemblage defined 12 vessel

    forms that could be placed into unrestricted (where the

    rim is the maximum circumference of the vessel) or

    restricted (where rim diameter smaller than the max-

    imum circumference of the vessel) categories (Fowler

    2002). Four unrestricted vessel forms and eight

    restricted forms were identified. A further 350 pieces

    of sculpture, including anthropomorphic, zoomorphicfigurines, large ceramic heads with zoomorphic

    characteristics figurines (Fowler 2008; Loubser 1993),

    and a miscellany of other objects including a star-

    shaped object, clay coils, and clay disks (vessel sherds

    with smoothed and rounded edges) make up the

    remainder of the assemblage.

    To infer the function of vessels, both the properties

    affecting the performance of containers (e.g. strength,

    hardness, color, texture, size, and shape) and alterations

    made to vessels while in use, such as use-wear patterns

    (abrasions, scraping, and pitting), residues of use(sooting, burning, fabric, and vitrification), visible

    organic residues, and perforations on vessel surfaces

    (for vessel repair) were considered (Skibo 1992). Based

    on the analysis of performance characteristics, use-wear

    and experimental data, the sample was classified into

    ten functional categories (Fowler 2002). Over 35,000

    sherds representing nearly 1900 objects were identified

    as cooking vessels; soaking vessels; parching vessels;

    small and medium-sized serving/eating vessels; medium

    to large sized serving vessels or serving/storage contain-

    ers; very large storage vessels; medium and large sizedstorage and/or transport containers; tools (likely for

    skimming iron slag during smelting); sculpture and

    figurines; and an other category (e.g. disks and coils).

    Seldom do specific ceramic forms serve a single

    function. Many vessels have traces of use resulting

    from more than one activity. The vessels found at

    Ndondondwane were either designed as multifunc-

    tional containers, or were used for multiple purposes

    and this overlap is reflected in the functional categories.

    Sample size

    Temporal differences were controlled by using only theceramics that can be securely assigned to the main

    occupation of the settlement (Fowler and Greenfield

    2009; Greenfield et al. 2005; Greenfield and van

    Schalkwyk 2003). The Middle Horizon sample includes

    ceramics from 16 spatially discrete features and deposits.

    The total sample of sherds and estimated objects

    excludes those with questionable or indeterminate

    provenance, reducing the sample from 2561 to 1047

    objects (TABLE 1).

    Ceramic Discard at NdondondwaneIndex scores for each feature or deposit dating to the

    Middle Horizon at Ndondondwane were compared to

    establish pattern combinations and clustering. Four

    basic pattern combinations were identified: low com-

    pleteness and low richness, high completeness and low

    richness, high completeness and high richness, and low

    completeness and high richness. Each basic pattern

    distinguishes spatially- and temporally-discrete depo-

    sitional areas that correspond to the four main types

    of functional areas identified at the site: a central

    livestock enclosure, domestic complexes, the open

    space between these areas, and a location for specialproduction and ritual activities. The ceramic analysis

    was also able to distinguish smaller sub-sets of

    depositional areas (pits, open middens) within the

    larger functional areas (domestic complexes). I now

    turn to a more detailed discussion of the index data

    from each functional area beginning with the domi-

    nant feature of the site, the livestock enclosure.

    Central livestock enclosureIn the Central Zone of Ndondondwane, the main

    feature of the Dung Area is a livestock enclosure used

    during the entire occupation of the site (FIG. 2A, FIG. 3).

    A stockade wall divides the area into two distinct areas:

    one used for penning livestock in the west (the

    Livestock Zone) and the other used for human activities

    in the east (the Human Activity Zone).

    The Livestock Zone consists of a byre (animal pen).

    In modern byres, cattle are separately penned from

    sheep and goat. Objects used within and around them

    are often discarded or stored along their edges.

    Archaeologically, we would therefore expect a mate-

    rial signature to reflect the many functional types of

    ceramics used and stored in the immediate area, and

    deposited along the edge of the livestock enclosure.

    During the Middle Horizon loose and compact

    dung layers are surrounded by a large, oval-shaped

    stockade with a smaller, adjacent stockade to the west

    (Fowler et al. 2004; Greenfield and van Schalkwyk

    2003; Loubser 1993). This pattern supports the idea

    that cattle were separated from sheep and goat. Over

    80 ceramic objects found in the Livestock Zone were

    deposited along the edge of the stockade wall during

    the Middle Horizon (TABLE 1, FIG. 3). The high RI

    and CI scores indicate that many functional typesof ceramics were discarded as both primary and

    secondary refuse (FIG. 4). Since no human activities

    are attested in the byre at Ndondondwane, objects

    found within it were likely used during activities in

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    Journal of Field Archaeology 2011 VOL . 36 NO. 2 15 7

  • 8/3/2019 EIA Fowler

    8/16

    the Human Activity Zone and subsequently dis-

    carded against the stockade wall during maintenance

    of the area.

    Human activity is attested in the eastern half of theDung Area, upslope of the Livestock Zone (Fowler

    et al. 2004; Greenfield and van Schalkwyk 2003;

    Greenfield et al. 1997; Loubser 1993). In the Human

    Activity Zone of the Dung Area there is evidence for

    the reworking of iron implements (i.e. charcoal and

    forging slag), the dumping of cultural debris (animal

    bone, pottery, etc.), and repeated roasting of meat over

    a fire-pit in a bowl-like depression, resulting in theaccumulation of much ash, charcoal, and burnt bone

    within the depression (FIGS. 2A, 3). During the main

    occupation, the palisade wall of the livestock enclosure

    shifts slightly to the east (12 m) and the associated

    Figure 2 Simplified plans of the main activity areas at Ndondondwane discussed in the text: A) The central livestock

    enclosure; B) Midden 1; C) Midden 3; D) Mound Area.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    15 8 Journal of Field Archaeology 2011 VOL . 36 NO. 2

  • 8/3/2019 EIA Fowler

    9/16

    human activity area shrinks in size and has fewer

    definite associated activity areas. Iron forging takes

    place, but to a much smaller extent than in the

    previous horizon.

    Low CI and high RI scores indicate a high breakage

    pattern and a broad range of functional types during

    the Middle Horizon (TABLE 1, FIG. 4). During this time,

    fewer activities occurred in the Human Activity Zone

    relative to the Lower Horizon, and the space was used

    less intensively (Greenfield and van Schalkwyk 2003).

    Several lines of evidence suggest this area is where

    feasts, celebrations, ceremonies, and assemblies took

    place in the settlement. Two ceramic figurines were

    deposited in this area, and regional ethnography has

    suggested that figurines served some instructional

    purpose in initiation ceremonies (Loubser 1993;

    Whitelaw 1994b). The figurines deposited in central

    middens at the sites of Ndondondwane and

    Kwagandaganda (Whitelaw 1994a) conform to this

    pattern. There are also very large quantities of food

    debris and waste in the Human Activity Zone,

    particularly remains of domesticated animals whose

    bones were discarded in the western part of the zone

    near the palisade wall (Fread 2007). Lastly, there is an

    unusually high quantity of food preparation and

    serving vessels in the area, and more than half of the

    finely decorated ceramics at the site are deposited in the

    area (Fowler 2002). Nearly two-thirds of food pre-

    paration vessels, three-quarters of the serving vessels,

    and one-third of the storage and storage/transport

    containers recovered from Ndondondwane come from

    the Human Activity Zone and the edge of the

    Livestock Zone. Most of the ceramics used forpreparing, eating, and transporting food are deposited

    in the central, communal use area of the settlement and

    not in domestic contexts, which we may have expected.

    The presence of finely decorated ceramics in public

    settings is not surprising, as ethnographic data from the

    region suggest that highly decorated containers often

    serve as display items in public settings (Fowler 2006).Based on these criteria, the Human Activity Zone

    in the Dung Area is a central area with a special

    public character serving as the symbolic center of

    social, political, and economic life in the community.

    The adjacent byre supports the idea of an area in

    which residents communally penned livestock and

    stored or discarded ceramics that were used for

    consumption and performances that had generated

    large amounts of debris. These activities occurred in

    the adjacent Human Activity Zone of the enclosure

    area. Once these activities had ceased, ceramics wereno longer stored or discarded along the stockade

    walls. When viewed alongside other evidence for

    activities in the Dung Area, the ceramic data support

    the interpretation of this as a communal area, defined

    by a central, open space that was accessible to at least

    all adult members of the resident households.

    The central position of this area and the high

    volume of food-related debris also indicate its social

    and economic importance. The best explanation for

    the discard pattern during the main occupation of the

    settlement is that ceramics accumulated in the Human

    Activity Zone from repeated use that occurred both

    within the Human Activity Zone and in the open space

    between the Dung Area and the peripheral domestic

    complexes. Its central location and the symbolic

    significance of livestock (in particular cattle) to south-

    ern African farming peoples (Huffman 1982, 1986a,

    1986b, 2001), suggest that the space around the byre

    would have been of considerable importance and

    prestige to residents. As such, people strove to improve

    the appearance of this area and maintained it by

    discarding trash in the Livestock Zone.

    Domestic complexes

    Turning to the Peripheral Zone at Ndondondwane,

    each domestic complex is associated with a circular

    structure with an interior fireplace. Grain storage

    Figure 3 The Dung Area at Ndondondwane facing west

    toward the river shwing several major features of the Middle

    Horizon: A) Near complete ceramic vessel in the middle of a

    proposed livestock tethering area; B) Row of postholes for

    the palisade of the livestock enclosure; and C) Large

    depression used for cooking.

    Figure 4 Ceramic index values for the central livestock

    enclosure area.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    Journal of Field Archaeology 2011 VOL . 36 NO. 2 15 9

  • 8/3/2019 EIA Fowler

    10/16

    pits, open middens, hearths, and ash dumps are

    located downslope and/or to the sides of most house

    floors (Greenfield and van Schalkwyk 2003). These

    features and their contents indicate that many typical

    household-related tasks such as food storage and

    preparation and tool production occurred within and

    outside domestic structures,

    House complexes tend to exhibit low CI and high RI

    comparable to the Human Activity Zone (TABLE 1,

    FIG. 5). We would expect high richness scores to

    complement the other evidence for food preparation

    and consumption in these areas, suggesting that these

    were deposited as primary refuse from each house.

    However, the low completeness scores indicate that

    ceramic vessels were also deposited as secondary refuse

    in complexes. Area maintenance, recycling to make

    clay disks, and random displacement such as tram-

    pling would have contributed to the formation of

    refuse areas while the settlement was occupied.

    Our previous studies have shown that depositional

    and post-depositional activities moved sherds from

    their original place of discard (Fowler et al. 2004).

    Plowing and erosion also contributed to lateral sherd

    displacement in most excavation areas, primarily

    affecting the top 10 cm of deposits. Field observa-

    tions and previous analyses indicate that Midden 1and 2 were more affected by plowing and erosion

    than Midden 3 (Fowler et al. 2004; Greenfield and

    van Schalkwyk 2003). The index data show a weak

    linear relationship between the CI and RI for domes-

    tic complexes (r250.464). This relationship does not

    support the conclusion that Midden 2 is the least well

    preserved and Midden 3 is the best preserved.

    The low correlation between the index data and

    the preservation of domestic complexes can be linked

    to the different recovery techniques used during their

    excavation. The best example is Midden 2, initiallyexcavated during the 19821983 season using 262 m

    units and both arbitrary levels and natural strati-

    graphic divisions (Fowler et al. 2004: 446448;

    Loubser 1993). Subsequent excavations during the

    1995 season in this area used a variety of different

    trench sizes and excavation techniques. A major

    problem is that nearly all non-diagnostic sherds

    (i.e. undecorated body sherds) from Midden 2 were

    discarded by a field assistant prior to the Canadian

    teams arrival (Haskell Greenfield, personal commu-

    nication 2009). The selective retention of rim sherdsduring excavation has artificially increased the CI

    score for Midden 2. Consequently, these data are not

    comparable to that recovered from other excavation

    areas and are not included in the analysis.

    In contrast, excavations in Middens 1 and 3 were more

    systematic (using either 161 m or 262 m units, which

    were often sectioned into halves), followed natural

    stratigraphic divisions, and all ceramic objects were

    retained (Fowler et al. 2004: 446). The combination of

    systemic excavation and recovery techniques in Middens

    1 and 3 allows their data to be compared.

    The ceramic index data are indicative of both the

    nature of recovery techniques used during different

    seasons of excavation and the differential impact of post-

    depositional processes across the site. Importantly, des-

    pite differences in preservation and recovery techniques,

    the ceramic index data clearly distinguish the domestic

    complexes from other activity areas at the site. The index

    data further indicate that the scores for domestic

    complexes are different from their component deposi-

    tional areas. The data from refuse middens and pits best

    demonstrate how ceramic index data may distinguish

    discrete areas of discard within a domestic complex.

    Middens

    Midden deposits are a primary component of

    domestic complexes. However, they are variable and

    we can distinguish between those associated with

    structural remains, such as pits and fire-pits, and

    those that have no associated features. The latter are

    referred to here as open middens. The index data

    clearly distinguish the different nature of these

    deposits found in Middens 1 and 3.

    There are five midden deposits in Midden 1 (FIG. 2B).Two are open middens (A1, A5) and three are otherwise

    associated with pits (A3, A6) or a fire-pit (A2) (FIG. 6).

    The midden beside Pit 3 (A3) had no ceramic data useful

    for this analysis so it was not included. All of the middens

    except the one with the fire-pit have high CI and low RI

    scores (FIG. 7). This suggests that ceramic vessels used for

    cooking, serving, and storage/transport in the nearby

    household were discarded as primary refuse. The fire-pit,

    however, has very low CI and higher RI scores. A small

    number of sherds from five fragmented vessels occur in

    and around the fire-pit. Some of these sherds wereblackened but the only vitrified sherd was located 2 m

    away in A6. This suggests that people did not dump

    broken pots in fire pits; rather, they regularly placed

    them in discrete middens located around it. Breakage

    Figure 5 Ceramic index values for the domestic complexes.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    16 0 Journal of Field Archaeology 2011 VOL . 36 NO. 2

  • 8/3/2019 EIA Fowler

    11/16

    during cooking and foot traffic likely contributed more

    significantly to sherd displacement in fire pits.

    The midden deposits in Midden 3 contrast those inMidden 1. Midden 3 has two midden deposits, each

    with one pit. The East Midden is located nearest the

    house floor and the West Midden is situated some

    20 m away from the house floor (FIG. 2C). Neither

    midden had ash deposited in them and they both

    have similar RI scores to open midden deposits

    in Midden 1, but much lower CI scores (FIG. 7). The

    East Midden and the West Midden are dissimilar; the

    West Midden has a wider range of ceramic types, but

    ceramics in both middens are highly fragmented. This

    pattern of high functional diversity and high frag-mentation is more typical of deposits with secondary

    refuse (Kent 1999). This implies that ceramics used

    in activities in and around the nearby household

    and from other parts of the settlement were likely

    deposited in the middens.

    The index data from Middens 1 and 3 clearly have a

    different distribution than those from the Livestock

    Enclosure or the cumulative signature of domestic

    complexes. Richness scores never reach 40, but there is

    a wide range of completeness index scores. Both sets of

    scores potentially capture the different discard beha-

    viors between domestic complexes. The differences

    between these middens highlight the variation that can

    exist within and between domestic complexes.

    Pits

    Pits have a different distribution of index scores than

    those that characterize the Livestock Enclosure, domes-

    tic complexes, or open middens (TABLE 1, FIG. 8). RI

    scores never reach 60 and CI scores never exceed 80.

    However, the pits in Midden 1 and Midden 3 cluster

    into two groups: those that have very low RI and CI

    scores, and those that have comparably higher scores.These clusters straddle the distribution of index scores

    for the open middens in each domestic area.

    Pits with low index scores include Pit 2 and 3 in

    Midden 1 and Pit 1 in the West Midden of Midden 3.

    These pits are more similar to the ceramic debris

    remaining on the abandoned house floor in Midden 3.

    In Pit 2 (Midden 1), nearly all the grinding stonesrecovered from the site were collected and dumped in

    this pit prior to the abandonment of the site

    (Greenfield and van Schalkwyk 2003). Very low RI

    and CI scores also characterize the burial in Pit 3 (in

    the open midden A3) located 20 m west of the house

    floor in Midden 1 (FIGS. 2B, 8). In this instance, the pit

    was filled rapidly with cultural debris (animal bone,

    pottery sherds, and organic debris). An infant (likely a

    neonate) was interred in a large cooking vessel inverted

    at the top of the pit, which was then subsequently

    capped by several large stones (Greenfield et al. 1997).

    The ceramic analysis also distinguished the shallow pit

    in the West Midden of Midden 3 (FIG. 8). This pit does

    not cut through the midden debris, so it was excavated

    prior to the accumulation of the midden. It has a

    higher CI score than other pits at the site because

    fragments from one vessel were deposited with no

    other artifacts or organic materials.

    In contrast, Pit 1 in Midden 1 and Pit 2 in Midden

    3 have higher index scores (FIG. 8). Each pit was

    originally dung-lined and used for storage. Only later

    were they used for the discard of refuse. In Midden 1,

    Pit 1 has higher CI and RI scores than Pit 2 or Pit 3.

    It is more similar to Pit 2 in the Eastern Midden of

    Midden 3. Pit 2 and Pit 3 have about twice as many

    sherds as Pit 1, so the fragmentation ratio in this case

    is not a function of sherd frequency. A comparable

    pattern of fragmentation is found in Midden 3

    between Pit 1 and Pit 2.

    Relative to the middens in which they occur, the

    index scores distinguish pits used for different purposes

    (TABLE 1, FIG. 8). The low index scores characterize pits

    used for the dumping of grinding stones, an infant

    burial in Midden 1, and a single pot burial in Midden 3.These pits should be considered special instances of

    discard. The scores for pits in Midden 1 (Pit 1) and the

    East Midden of Midden 3 (Pit 2) indicate another

    pattern of discard. The few, well-preserved functional

    Figure 6 Open midden A2 in Midden 1 from the west

    showing a substantial accumulation of ash.

    Figure 7 Ceramic index values for middens in the three

    domestic complexes.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    Journal of Field Archaeology 2011 VOL . 36 NO. 2 16 1

  • 8/3/2019 EIA Fowler

    12/16

    types in these pits may distinguish the typical discard of

    ceramics from households that accumulated during the

    maintenance of the area. In sum, each cluster of index

    scores for pit deposits at Ndondondwane differentiated

    dung-lined pits that were initially used for grain storage,

    pits used for burials, and pits used only for the discard

    of specific cultural debris.

    Variation in domestic complexes

    The ceramic index data suggest variation in discard

    behaviors within and between domestic complexes.

    Within complexes, the pit data suggest that index scores

    distinguish regular maintenance and discard episodes

    from special instances of discard. Regular discard

    patterns are characterized by high richness and com-

    pleteness scores, while burials, the discard of single pots,

    and pits rapidly filled with cultural debris during

    the abandonment of a household have much lower

    scores. Additionally, regular discard in open middens is

    characterized by a distribution of index scores that lay

    between those that distinguish pit deposits (FIG. 8).

    Between complexes, clear differences were identified

    in the analysis of the open middens in Midden 1 and

    Midden 3. The variation in ceramic index scores mirror

    the differences in the features and artifact types found in

    each domestic area. For instance, forging slag, large ash

    deposits, and abundant daga fragments (a clay-dung

    mixture used to waterproof granaries and other

    structures) are not found in Midden 3. The absence of

    forging debris, the large open middens (containing a

    ceramic figurine) situated down slope of the well-

    preserved house floor, and the high volume of ceramics

    and decorated vessels set this complex apart from

    others. The different types, quantities, and distribution

    of artifacts between Midden 1 and Midden 3 highlight

    the variation that can occur in such areas at Early Iron

    Age settlements. The ceramic index data are reliableindicators of such variation and help define domestic

    activity areas with greater precision.

    In sum, the distribution of ceramics and other

    remains (metal-working debris, faunal remains, daga,

    etc.) from domestic areas indicate that spaces to the

    front and sides of houses were multi-functional. A

    child was interred in a pit in the front of one house,

    characterized by a rapid accumulation of highly

    fragmented ceramic debris in the pit. Other spaces

    were used to entertain guests, prepare food, in addition

    to the production, maintenance, and recycling of ironobjects. The refuse generated during these activities

    was discarded in regularly maintained, open middens,

    located 515 m away from dwellings during the

    occupation of the site. Prior to the abandonment of

    the site, this pattern changed and refuse was discarded

    in dung-lined storage pits and, in at least one instance,

    a midden some distance from a dwelling.

    Special purpose areasThree excavation areas at Ndondondwane, referred

    to as special purpose areas, have yielded evidence for

    activities unrelated to livestock keeping, communalpublic events, and everyday domestic life. The open

    space between the Central and Peripheral Zones and

    the Mound Area were utilized for the entire occupa-

    tion of the site. The Charcoal Preparation Area

    accumulated from activities that occurred exclusively

    during the Upper Horizon. Only the open space and

    Mound Area deposits dating to the Middle Horizon

    are considered in the following analysis.

    Open space

    Similar to other Early Iron Age sites in the region such

    as Kwagandaganda (Whitelaw 1994a, Huffman 1993,

    2001), there is a large open space at Ndondondwane

    that spans about 60 m from the Dung Area to Midden

    1 and 150 m from Midden 2 to Midden 3 (FIG. 1).

    Above, it was posited that a portion of the open space

    was used as extension of the Human Activity Zone.

    We have also recently argued that this space could

    have contained household or communal gardens

    (Greenfield et al. 2005).

    The ceramic analysis detected several depositional

    areas in this open space. Deposits to the southwest of

    the livestock enclosure, in Transects 1 and 2, have low

    CI and RI scores (TABLE 1, FIG. 9), typical of deposits

    associated with abandoned house floors and pits

    (FIGS. 7, 8). The few vessels deposited in these areas

    are mainly for cooking and serving. These deposits can

    best interpreted as a sheet midden accumulated

    through occasional or casual use of the area.

    Historically, boys take meals in these areas of home-

    steads while tending livestock (Fowler 2002: 6667,

    34041; Greenfield et al. 2005). Alternatively, Transect

    1 may have served as a peripheral toft, or casual

    discard area (Hayden and Cannon 1983) during themain occupation of the settlement, accounting for the

    higher RI score in Transect 1. While it is uncertain how

    these deposits accumulated, the absence of structural

    remains and their ceramic signatures set them apart

    Figure 8 Ceramic index values for pits and an abandoned

    house floor in domestic complexes relative to open middens

    (shaded area; see FIG. 7).

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    16 2 Journal of Field Archaeology 2011 VOL . 36 NO. 2

  • 8/3/2019 EIA Fowler

    13/16

    from the specific and redundant sets of activities that

    occur in adjacent areas.

    The Mound Area

    To the south of the open space, the Mound Area was

    a main area of activity in the centre of the site

    (FIG. 2D). After the house in the area was demolished

    in the Lower Horizon, a fence or the walls of an

    ovaloid structure were constructed (Loubser 1993:

    117), marking the main occupation (Middle Horizon)

    of the site in this area. The activities that took place

    in and around this structure include manufacturing

    ivory bangles and butchering wild animals and

    processing their hides. Several human figures and

    the fragments of at least four ceramic sculptures with

    zoomorphic features were also found in this area.

    Most of this debris accumulated in what has been

    termed the Grey Midden (Maggs 1984).

    Ceramics from this horizon have lower CI and higher

    RI scores than the Human Activity Zone (TABLE 1,

    FIG. 9). These scores suggest the functional range of

    ceramics in this area is more limited than in the

    contemporary Human Activity Zone of the Dung Area,

    and that many ceramics may have been discarded as

    secondary refuse. However, several lines of ethno-

    graphic evidence suggest that the ceramics could have

    accumulated from primary activities that occurred in

    this area.

    First, the sculptures recovered from the fill of Grey

    Midden led Loubser (1993) to suggest this is where

    initiation rituals conducted at the site. Throughout

    the Bantu-speaking world in southern Africa, certain

    parts of initiation rituals occur in private, usually in

    buildings where public access was restricted; Loubser

    interpreted the structure in the Mound Area as a

    location for initiation ceremonies.

    Second, such places in settlements are also typicallyassociated with male-related activities, such as ivory-

    and metal-working (Huffman 1986b). Large accumu-

    lations of ash are related to the court or mens

    assembly area, where men gather to discuss political

    matters (Kuper 1980, 1982). The size of middens is

    directly related to the political importance of the

    homestead head because men with higher political

    positions in the regional system hold larger assemblies

    (Huffman 1986b: 316). Further, Loubser (1985: 85)

    has described how ash may play a symbolic role in

    political debate by acting as a cooling agent in hotsituations. Water and ash are two of a series of

    cooling agents linked to healing and fertility that are

    opposed to hot, dangerous, sterilizing forces in south-

    ern Bantu speaking societies (Kuper 1982: 1824).

    These psychological boundaries symbolize proscrip-

    tions for behavior that are linked to the movement of

    men and women in settlements and the nature of the

    activities they undertake in different settings.

    There is evidence for assembly-area middens with

    large ash deposits situated away from livestock byres

    at Iron Age settlements (Denbow 1982, 1983, 1984;

    Eloff and Meyer 1981; Huffman 1986b, 2007;

    Loubser 1985). Based on its position in the settle-

    ment, the structure in the Mound Area was linked to

    political activity at Ndondondwane. The presence of

    only one vessel for serving or drinking during this

    stage of use implies that its contents were consumed

    but not likely prepared here, and the abundant eating

    and serving vessels discarded along the wall of the

    structure may have been used for meals during

    meetings or by initiates during their seclusion.

    Ethnographic and archaeological evidence support

    the argument that the Mound Area at Ndondondwane

    is linked to political activity. However, it may not be the

    only location used for such activities. The location of

    mens assembly areas varies in settlements throughout

    the Bantu-speaking world of southern Africa, and its

    precise location is flexible as long as it occurs in the

    central area of settlements, which is associated with

    male practices linked conceptually to cattle ownership

    (Kuper 1980, 1982). Assembly areas occur both in byres

    and in the open space of settlements. Weather and

    seasonality can influence the location, as meetings can

    be held in shaded areas when warm and hot and in the

    open during the cooler, dryer winter months. Seasonal

    variation and the preference of homestead heads to use

    a particular location in a settlement for meetings are

    factors that would influence the kinds, quantity, and

    distribution of vessels used during gatherings. At

    Ndondondwane, the cooking and serving vessels in

    Transect 1 and 2 are comparable in terms of fra-

    gmentation, but differ only in the range of types

    represented. This pattern may well reflect the intensity

    of repeated use of these locations, and not the kinds of

    activities that occurred at them.

    ConclusionsThe primary aim of this analysis was to evaluate

    whether the fragmentation and functional variation of

    Figure 9 Ceramic index values for special purpose areas

    and open space.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    Journal of Field Archaeology 2011 VOL . 36 NO. 2 16 3

  • 8/3/2019 EIA Fowler

    14/16

    ceramic objects at Ndondondwane could be used to

    distinguish discrete depositional areas in support of

    other evidence for different activities that occurred

    within them. The results support previous conclusions

    concerning the use of space at Ndondondwane. These

    new data refine some earlier conclusions, allowing more

    detailed and dynamic interpretations of the possibleactivities contributing to the formation of 16 spatially-

    and temporally-discrete archaeological deposits.

    The second aim was to evaluate the methods

    applicability for use elsewhere. It was determined that

    the discard of ceramics at Ndondondwane could be used

    to distinguish five main functional areas: domestic com-

    plexes, livestock enclosures, burials, open space, and

    special purpose areas, five areas that characterize farm-

    ing settlements dating to the past two millennia in

    southern Africa (Huffman 2007), perhaps as main

    components of farming settlements throughout tropical

    Africa. Open middens, refuse pits, and house floors are

    distinguishable from the cumulative fragmentation

    signature and total functional inventory represented at

    entire sites. Only the sheet trash from open space is

    difficult to discriminate from the ceramic debris in pits

    and house floors, but spatially-isolated deposits with

    sheet trash are discernable by the absence of structural

    remains. Consequently, the results provide hypotheses

    about the relationships between the kinds, quantity, and

    distribution of ceramics and the activities that generated

    the layout of early farming settlements in southern

    Africa.

    One hypothesis resulting from this analysis is that

    different functional areas will be distinguishable by

    discrete ceramic index scores. In areas without asso-

    ciated features (middens, house floors, and pits), it is

    difficult to identify with any certainty the activities that

    generated certain deposits. However, the patterns allow

    the analyst to entertain a range of possible behaviors,

    and the ceramic index data act as a relative measure of

    the accumulation behaviors at different activity areas at

    Ndondondwane. Comparable analysis of ceramic data

    from other sites will generate indices with different

    structures. With appropriate chronological control and

    ceramic data, the method could be used to identify the

    same kinds of functional areas at different stages of a

    sites development or they could be used to monitor

    changes within activity areas over time.

    This study supports the caution provided by eth-

    noarchaeological research concerning the relationship

    between the place where ceramics are discarded and the

    place where they are used. It was clear that at Ndon-

    dondwane ceramics do not necessarily indicate the

    kinds, range, or intensity of activities at or near the spotthey were discarded. Most domestic middens demon-

    strate reasonable linkages between what Pauketat

    (1989) referred to as the archaeological assemblage

    (ceramics recovered archaeologically), the disposal

    assemblage (ceramics deposited in a location), and the

    behavioral assemblage (ceramics used in a particular

    behavioral setting, such as a household). At Ndon-

    dondwane, the Dung Area provides a counter example.

    The range and intensity of activities declined during the

    Middle Horizon in the Dung Area (Greenfield and van

    Schalkwyk 2003), yet this is when the greatest quantityand functional range of ceramics were discarded. In this

    specific context, ceramics discarded near the livestock

    enclosure were used during activities in the open space

    of the settlement. It is evident that the same category of

    feature (e.g., pits, middens) did not accumulate in the

    same ways, as shown by the variation amongst the

    middens in the domestic complexes at Ndondondwane.

    The different discard patterns observed for open mid-

    dens suggest that different practices produced them. In

    other words, there is an unreasonable correspondence

    between the archaeological, disposal, and behavioral

    assemblages. After considering a range of potential

    disturbance processes, different abandonment discard

    practices can explain why the East Midden in Midden 3

    was unlike other middens in Midden 1.

    This study also identifies issues concerning artifact

    recovery at Ndondondwane. In particular, Midden 2

    was not useful for this analysis because of how data

    were collected during excavation. Clearly, recovery

    and retention methods used by archaeologists can

    have an impact on the potential for the analysis of

    ancient behavior. Large-scale recovery techniques

    may provide data that makes answering questions

    about social behavior and processes at smaller scales

    difficult. Likewise, the selective retention of material

    in the field hinders attempts at comparative analyses,

    either within or between sites.

    This paper demonstrates a method for using ceramic

    accumulations to evaluate the effects of recovery

    methods, the nature of depositional contexts at archae-

    ological sites, and the behavioral interpretations that

    define past activity settings. Accumulations studies do

    not involve an exercise of filtering out natural

    processes to better understand cultural factors influen-cing site formation, but rather target the interplay

    between natural and cultural processes that lead to

    formation of archaeological deposits.

    Both the method and the testable hypotheses

    regarding ceramic accumulations are applicable to

    other farming settlements both within and outside the

    region given appropriately quantified ceramic data.

    The new method may provide a tool for under-

    standing the organization of sites occupied by

    agriculturalists who relied on perishable building

    materials. Utilizing complementary analytical appro-aches provides deeper insight into the dynamic

    processes contributing to the formation and changes

    in farming settlements with widely dispersed deposits

    and few structural remains.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    16 4 Journal of Field Archaeology 2011 VOL . 36 NO. 2

  • 8/3/2019 EIA Fowler

    15/16

    AcknowledgmentsI extend my appreciation to the staff of the Amafa

    aKwaZulu-Natali (Heritage KwaZulu-Natal, South

    Africa) in Ulundi and Pietermaritzberg, the Depart-

    ment of History and Archaeology at the Natal Mu-

    seum, and eThembeni (Pietermaritzberg), each of who

    graciously offered the use of their facilities during thefield and laboratory analysis of the data for this paper.

    Financial support from the University of Alberta, the

    Social Science and Humanities Research Council of

    Canada (Grant Nos. 752-99-1163 and 756-2002-0381)

    is gratefully acknowledged. Nicholas David, Haskel

    Greenfield, Augustin Holl, Tom Huffman, David

    Lubell, E. Ann McDougall, Tim Maggs, Gavin

    Whitelaw, Andrzej Weber, Pam Willoughby, and the

    reviewers have all contributed to the development of the

    analysis and thoughts expressed in this paper and their

    contributions are greatly appreciated. I wish to thank

    the excavators (Haskel Greenfield and Len van Schalk-

    wyk) for permission to use data from the Ndon-

    dondwane excavations in preparing this manuscript.

    Figures 3 and 6 are courtesy of Haskel Greenfield.

    Kent Fowler (Ph.D. 2002, University of Alberta) is

    currently Assistant Professor in the Department of

    Anthropology at the University of Manitoba. His

    research interests are in complex societies in Europe

    and Africa, which he has explored through the study of

    Neolithic mortuary practices and zooarchaeology in

    Europe, and Iron Age settlement organization andceramic technology in South Africa. His current

    research focuses on the ethnoarchaeology of ceramic

    production and regional identity in South Africa.

    ReferencesBeck, M. E. 2006. Midden Ceramic Assemblage Formation: A

    Case Study from Kalinga, Philippines, American Antiquity 71:2751.

    Beck, M. E. 2009. Counting Pots in Kalinga, Philippines: Short-and Long-Term Change in Household Assemblages,Ethnoarchaeology 1: 79106.

    Beck, M. E., and M. E. Hill, Jr. 2004. Rubbish, Relatives, and

    Residence: The Family Use of Middens, Journal ofArchaeological Method and Theory 11: 297333.

    Carr, C. 1984. The Nature of Organization of IntrasiteArchaeological Records and Spatial Analytic Approaches toTheir Investigation, Advances in Archaeological Method andTheory 7: 103222.

    David, N. 1972. On the Life Span of Pottery Type Frequencies, andArchaeological Inference, American Antiquity 37: 141142.

    David, N., and H. Henning. 1972. The Ethnography of Pottery: AFulani Case Study Seen in Archaeological Perspective.Cambridge: Addison-Wesley.

    Deal, M. 1998. Pottery Ethnoarchaeology in the Central MayaHighlands. Salt Lake City: University of Utah Press.

    DeBoer, W. R. 1974. Ceramic Longevity and ArchaeologicalInterpretation: An Example from the Upper Ucayali, Peru,American Antiquity 39: 335344.

    DeBoer, W. R., and D. W. Lathrap. 1979. The Making andBreaking of Shipibo-Conibo Ceramics, in C. Kramer, ed.,Ethnoarchaeology: Implications of Ethnography forArchaeology. New York: Columbia University Press, 102138.

    Denbow, J. 1982. The Toutswe Tradition: A Study in Socio-Economic Change, in R. R. Hitchcock and M. R. Smith, eds.,Settlement in Botswana. Johannesburg: Heinneman, 7386.

    Denbow, J. 1983. Iron Age Economics: Herding, Wealth and PoliticsAlong the Fringes of the Kalahari During the Early Iron Age,unpublished Ph.D. dissertation, Indiana University, Bloomington.

    Denbow, J. R. 1984. Prehistoric Herders and Foragers of theKalahari: The Evidence for 1500 Years of Interaction, inC. Schrire, ed., Past and Present in Hunter-Gatherer Studies.New York: Academic Press, 175193.

    Eggert, M. K. H. 1993. Vergangenheit in der Gegenwart?Uberlegungen zum interpretatorischen Potential der Ethnoar-chaologie, Ethnographische-Archaologische Zeitschrift 34: 14450.

    Eloff, J. F., and A. Meyer. 1981. The Greefswald Sites, in E. A.Voight, ed., Guide to Archaeological Sites in the Northern andEastern Transvaal. Pretoria: Transvaal Museum, 722.

    Fowler, K. D. 2002. Early Iron Age Community Organization inSouthern Africa: Social and Symbolic Dimensions of CeramicProduction, Use and Discard at Ndondondwane. Ph.D. disserta-tion, University of Alberta, Edmonton. Theses Canada,AMICUS No. 29895381.

    Fowler, K. D. 2006. Classification and Collapse: TheEthnohistory of Zulu Ceramic Use, South Africa, SouthernAfrican Humanities 18: 93117.

    Fowler, K. D. 2008. Social Memory and the Antiquity ofSnake and Crocodile Symbolism in Southern Africa, inS. Badenhorst, P. Mitchell, and J. Driver, eds., Animals andPeople: Archaeozoological Papers in Honour of Ina Plug. BARInternational Series 1949. Oxford: Archaeopress, 169185.

    Fowler, K. D., and H. J. Greenfield. 2009. Unravelling SettlementHistory at Ndondondwane, South Africa: A MicrochronologicalAnalysis, Southern African Humanities 21: 345393.

    Fowler, K. D., H. J. Greenfield, and L. O. van Schalkwyk. 2000.The Identification and Signficance of Ceramic Ecofacts fromNdondondwane, South Africa, Southern African FieldArchaeology 9: 3242.

    Fowler, K. D., H. J. Greenfield, and L. O. van Schalkwyk. 2004.The Effects of Burrowing Activity on Archaeological Sites:Ndondondwane, South Africa, Geoarchaeology 19: 441470.

    Frankel, D., and J. M. Webb. 2001. Population, Households, andCeramic Consumption in a Prehistoric Cypriot Village,Journal of Field Archaeology 28: 115129.

    Fread, E. 2007. Taphonomy and Subsistence Strategies atNdondondwane: A Zooarchaeological Perspective on anEarly Iron Age Village in the Thukela Valley, Republic of

    South Africa, MA thesis, University of Manitoba, Winnipeg.Theses Canada, AMICUS No. 34305851.

    Gallivan, M. D. 2002. Measuring Sedentariness and SettlementPopulation: Accumulations Research in the Middle AtlanticRegion, American Antiquity 67: 535557.

    Greenfield, H. J., K. D. Fowler, and L. O. van Schalkwyk. 2005.Where Are the Gardens? Early Iron Age Horticulture in theThukela River Basin of South Africa, World Archaeology 37:307328.

    Greenfield, H. J., and D. Miller. 2004. Spatial Patterning of EarlyIron Age Metal Production at Ndondondwane, South Africa:A Question of Cultural Continuity between the Early and LateIron Ages, Journal of Archaeological Science 31: 15111532.

    Greenfield, H. J., and L. O. van Schalkwyk, 2003. Intra-Settlement Social and Economic Organization of Early IronAge Farming Communities in Southern Africa: A View from

    Ndondondwane, Azania 38: 123137.Greenfield, H. J., L. O. van Schalkwyk, and T. L. Jongsma. 1997.

    Ndondondwane: Preliminary Report on the 1995 Survey andExcavations, Nyame Akuma 47: 4252.

    Greenfield, H. J., L. O. van Schalkwyk, and T. L. Jongsma. 2000.Surface and Subsurface Reconnaissance at Ndondondwane:Preliminary Results of the 19951997 Field Season, SouthernAfrican Field Archaeology 9: 516.

    Hardy-Smith, T., and P. C. Edwards. 2004. The Garbage Crisis inPrehistory: Artefact Discard Patterns at the Early NatufianSite of Wadi Hammeh 27 and the Origins of Household RefuseDisposal Strategies, Journal of Anthropological Archaeology23: 253289.

    Hayden, B., and A. Cannon. 1983. Where the Garbage Goes:Refuse Disposal in the Maya Highlands, Journal ofAnthropological Archaeology 2: 117163.

    Hietala, H. 1984. Intrasite Spatial Analysis in Archaeology.Cambridge: Cambridge University Press.

    Huffman, T. N. 1982. Archaeology and the Ethnohistory of theAfrican Iron Age, Annual Review of Anthropology 11: 133150.

    Huffman, T. N. 1986a. Archaeological Evidence andConventional Explanations of Southern Bantu SettlementPatterns, Africa 56: 280298.

    Fowler Ceramic discard and the use of space, Early Iron Age, South Africa

    Journal of Field Archaeology 2011 VOL . 36 NO. 2 16 5

    http://www.ingentaconnect.com/content/external-references?article=0278-4165()2L.117[aid=9156255]http://www.ingentaconnect.com/content/external-references?article=0278-4165()2L.117[aid=9156255]http://www.ingentaconnect.com/content/external-references?article=0278-4165()2L.117[aid=9156255]http://www.ingentaconnect.com/content/external-references?article=0278-4165()2L.117[aid=9156255]http://www.ingentaconnect.com/content/external-references?article=0278-4165()2L.117[aid=9156255]http://www.ingentaconnect.com/content/external-references?article=0278-4165()2L.117[aid=9156255]http://www.ingentaconnect.com/content/external-references?article=0278-4165()2L.117[aid=9156255]
  • 8/3/2019 EIA Fowler

    16/16

    Huffman, T. N. 1986b. Iron Age Settlement Patterns and theOrigins of Class Distinction in Southern Africa, Advances inWorld Archaeology 5: 291338.

    Huffman, T. N. 1993. Broederstroom and the Central CattlePattern, South African Journal of Science 89: 220226.

    Huffman, T. N. 2001. The Central Cattle Pattern and Interpretingthe Past, Southern African Humanities Journal 13: 1935.

    Huffman, T. N. 2007. A Handbook to the Iron Age: TheArchaeology of Pre-Colonial Farming Societies in SouthernAfrica. Pietermaritzburg: University of KwaZulu-Natal Press.

    Kent, S. 1999. The Archaeological Visibility of Storage:Delineating Storage from Trash Areas, American Antiquity64: 7994.

    Kuper, A. 1980. Symbolic Dimensions of the Southern BantuHomestead, Africa 50: 823.

    Kuper,A. 1982. Wives for Cattle. London: Routledge and Kegan Paul.Longacre, W. A. 1985. Pottery Use-Life among the Kalinga,

    Northern Luzon, the Phillippines, in B. A. Nelson, ed.,Decoding Prehistoric Ceramics. Carbondale: Southern IllinoisUniversity Press, 95111.

    Loubser, J. H. N. 1985. Buffelshoek: An EthnoarchaeologicalConsideration of a Late Iron Age Settlement in the SouthernTransvaal, South African Archaeological Bulletin 40: 8187.

    Loubser, J. H. N. 1993. Ndondondwane: The Significance of Featuresand Finds from a Ninth-Century Site on the Lower Thukela River,Natal, Natal Musuem Journal of Humanities 5: 109151.

    MacLean, M. R. 1994. Late Stone Age and Early Iron AgeSettlement in the Interlacustrine Region: A District CaseStudy, Azania 2930: 296302.

    Maggs, T. 1976. Iron Age Communities of the Southern Highveld.Pietermaritzburg: Natal Museum.

    Maggs, T. 1980. The Iron Age Sequence South of the Vaal andPongola Rivers: Some Historical Implications, Journal ofAfrican History 21: 145.

    Maggs, T. 1984. Ndondondwane: A Preliminary Report on anEarly Iron Age Site on the Lower Tugela River, Annals of theNatal Museum 24: 7193.

    Mayor, A. 1994. Durees De Vie Des Ceramiques Africaines:Facteurs Responsables Et Implicationsarcheologiques, inTerre Cuite Et Societe: La Ceramique, Document Technique,Economique, Culturel. Xive Rencontres InternationalesDarcheologie Et Dhistoire Dantibes. Juan-les-Pins: Editions

    APDCA, 179198.Mercader, J., R. Mart, I. J. Gonzalez, A. Sanchez, and P. Garca.

    2003. Archaeological Site Formation in Rain Forests: InsightsFrom the Ituri Rock Shelters, Congo, Journal ofArchaeological Science 30(1): 4565.

    Mills, B. J. 1989. Ceramics and Settlement in the Cedar MesaArea, Southeastern Utah: A Methodological Approach,unpublished Ph.D. dissertation, University of New Mexico,Albaquerque.

    Montgomery, B. K. 1993. Ceramic Analysis as a Tool forDiscovering Processes of Pueblo Abandonment, in C. M.Cameron and S. A. Tomka, eds., Abandonment of Settlementsand Regions: Ethnoarchaeological and ArchaeologicalApproaches. Cambridge: Cambridge University Press, 157164.

    Needham, S., and T. Spence. 1997. Refuse and the Formation ofMiddens, Antiquity 71: 7790.

    Orton, C. 1985. Two Useful Parameters for Pottery Research, in E.Webb, ed., Computer Applications in Archaeology 1985. London:University of London Institute of Archaeology, 114120.

    Orton, C. 1993. How Many Pots Make Five? An HistoricalReview of Pottery Quantification, Archaeometry 35: 169184.

    Pauketat, T. R. 1989. Monitoring Mississippian HomesteadOccupation Span and Economy Using Ceramic Refuse,American Antiquity 54: 288310.

    Pool, C. A. 1992. Integrating Ceramic Production andDistribution, in G. J. Bey and C. A. Pool, eds., CeramicProduction and Distribution: An Integrated Approach. Boulder,San Francisco, and Oxford: Westview Press, 275313.

    Reid, A. 1994. Early Settlement and Social Organization in theInterlacustrine Region, Azania 2930: 303313.

    Reid, J. J. 1973. Growth and Response to Stress at GrasshopperPueblo, Arizona, unpublished Ph.D. dissertation, University

    of Arizona, Tucson.Reid, J. J. 1978. Response to Stress at Grasshopper Pueblo,Arizona, in P. F. Grebinger, ed., Discovering Past Behaviour:Experiments in the Archaeology of the American Southeast.London: Gordon and Breach, 195213.

    Reid, J. J. 1985. Formation Processes for the Practical Prehistorian:An Example from theSoutheast, in S. R. Dickens, Jr.and H. T.

    Ward, eds., Structure and Process in Southeastern Archaeology.London: Gordon and Breach, 1133.

    Reid, J. J., M. B. Schiffer, and J. M. Neff. 1975. ArchaeologicalConsiderations of Intrasite Sampling, in J. W. Mueller, ed.,Sampling in Archaeology. Tuscon: University of Arizona Press,209224.

    Reid, J. J., and I. Shimada. 1985. Pueblo Growth at Grasshopper:Methods and Models, in W. A. Longacre, S. J. Holbrook, andM. W. Graves, eds., Multidisciplinary Research at Grasshopper

    Pueblo. Tuscon: University of Arizona Press, 1218.Rice, P. M. 1981. Evolution of Specialized Pottery Production: ATrial Model, Current Anthropology 22: 219240.

    Rice, P. M. 1987. Pottery Analysis: A Sourcebook. Chicago andLondon: University of Chicago Press.

    Rosenswig, R. M. 2009. Early Mesoamerican Garbage: Ceramic andDaub Discard Patterns from Cuauhtemoc, Soconusco, Mexico,Journal of Archaeological Method and Theory 16: 132.

    Santley, R. S., and R. R. Kneebone. 1993. Craft Specialization,Refuse Disposal, and the Creation of Spatial ArchaeologicalRecords in Prehispanic Mesoamerica, in R. S. Santley and K.G. Hirth, eds., Prehispanic Domestic Units in WesternMesoamerica: Studies of Household, Compound andResidence. Boca Raton, FL: CRC Press, 3763.

    Schiffer, M. B. 1987. Formation Processes in the ArchaeologicalRecord. Albuquerque: University of New Mexico Press.

    Schmidt, P., and S. T. Childs. 1985. Innovation and Industry duringthe Early Iron Age in East Africa: the KM2 and KM3 Sites ofNorthwest Tanzania, African Archaeological Review 3: 5394.

    Shapiro, G. 1984. Ceramic Vessels, Site Permanence and Group Size:A Mississippian Example, American Antiquity 49: 696712.

    Shott, M. 2006. Formation Theorys Past and Future: Introductionto the Volume, in M. Shott, ed., Formation Theory inArchaeology: Readings from American Antiquity and LatinAmerican Antiquity. Washington, D.C.: The SAA Press, 116.

    Shott, M. J. 1989. On Tool-Class Use Lives and the Formation ofArchaeological Assemblages, American Antiquity 54: 930.

    Singer, M. 2008. Intra-settlement Community Organization atNdondondwane: An Early Iron Age Settlement in the ThukelaValley, South Africa. MA thesis, University of Manitoba,Winnipeg. Theses Canada, AMICUS No. 37384159.

    Skibo, J. M. 1992. Pottery Function: A Use-Alteration Perspective.

    New York: Plenum.South, S. 1978a. Pattern Recognition in Historical Archaeology,American Antiquity 43: 223230.

    South, S. 1978b. Research Strategies for Archaeological PatternRecognition on Historic Sites, World Archaeology 10: 3650.

    Sullivan, A. P., III. 1989. The Technology of Ceramic Reuse:Formation Processes and Archaeological Evidence, WorldArchaeology 21: 101114.

    Tyers, P., and C. Orton. 1991. Statistical Analysis of CeramicAssemblages, in K. Lockyear and S. P. Q. Rahtz, eds.,Computer Applications and Quantitative Methods inArchaeology 1990. BAR International Series 565. Oxford:B.A.R., 117120.

    Van Noten, F. 1979. The Early Iron Age of the InterlacustrineRegion: The Diffusion of Iron Technology, Azania 14: 6180.

    van Schalkwyk, L. O., H. J. Greenfield, and T. L. Jongsma. 1997.

    The Early Iron Age Site of Ndondondwane, Kwazulu-Natal,South Africa: Preliminary Report on the 1995 Excavations,Southern African Field Archaeology 6: 6177.

    Varien, M. D., and B. J. Mills. 1997. Accumulations Research:Problems and Prospects for Estimating Site Occupation Span,Journal of Archaeological Method and Theory 4: 141191.

    Varien, M. D., and S. G. Ortman. 2005. Accumulations Researchin the Southwest United States: Middle-Range Theory for Big-Picture Problems, World Archaeology 37: 132155.

    Whitelaw, G. D. 1994a. Kwagandaganda: Settlement Patterns inthe Natal Early Iron Age, Natal Musuem Journal ofHumanities 6: 164.

    Whitelaw, G. D. 1994b. Towards an Early Iron Age Worldview:Some Ideas from Kwazulu-Natal, Azania 2930: 3750.

    Wilson, D. C. 1994. Identification and Assessment of SecondaryRefuse Aggregates, Journal of Archaeological Method and

    Theory 1: 4168.Wilson, D. C., W. L. Rathje, and W. W. Hughes. 1991.Household Discards and Modern Refuse: A Principle ofHousehold Resource Use and Waste, in E. Staski and L. D.Sutro, eds., The Ethnoarchaeology of Refuse Disposal.Anthropological Research Papers. Tempe: Arizona StateUniversity, 4151.

    Fowler