Byzantine Pottery from Sagalassos

I was pretty surprise to see an article entitled “Middle-Late Byzantine Pottery from Sagalassos: Typo-Chronology and Sociocultural Interpretation” in the very recent Hesperia (A.K. Vionis, J. Poblome, B. De Cupere, M. Waelkens, Hesperia 79 (2010), 423-464).  It’s not so much that the subject matter is late, but that the site of Sagalassos is a Belgian project in Turkey rather than an American project in Greece.  As some of my more observant friends pointed out, Hesperia has published the results of project from Albania, so maybe this should not have caught be so off guard.  But it did and it indicates to me that Hesperia is continuing to expand its purview to include the wider world of Mediterranean archaeology. Hooray!

The article on the Middle-Late Byzantine material from Sagalassos is pretty cool as well.  The main focus of the article is on a series of 12th-13th century layers from the Alexander Hill at the site of Sagalassos.  Over three seasons of excavation, the excavators uncovered the remains of a “heavily burned” destruction layer containing the remains of a short-lived occupation containing a significant and robust quantity of 12th-13th century Byzantine pottery.  This layer appears to represent the final phase of activity on this dramatic hill overlooking the ancient site of Sagalassos.  Early occupation on the hill included a 5th-6th century basilica that was almost completely removed and a later “refuge” of some description with a fortification wall and a substantial cistern.  Apparently the church was almost completely dismantled for the construction of the later refuge. The final destruction layer, which seems to represent the final layer of occupation, may represent an effort to dismantle the refuge to prevent it from being used again.

While the site history of the Alexander Hill is pretty interesting (particularly the dismantling of the church), the real meat of the article is in the analysis of the ceramic assemblage from the final layer.  While I would like to have understood the sampling method the produced the assemblage, the authors nevertheless conduct a rigorous and thorough examination of the material and take into account both “common ware” (which we would call medium coarse, coarse, and kitchen/cooking ware in chronotype terminology) and glazed table wares (fine and and semi-fine wares in our terminology).  Some of the glazed wares were repaired indicating that the objects had significant value to their owners.  The presence of repaired pots in an assemblage associated with the destruction of the site, however, suggests (to me at least) that these vessels were either discarded by the last occupants of the refuge or brought to the site by work crews commissioned to destroy or salvage the remains of the site. I wish the article had made considered more thoroughly the formation processes at play in the creation of the assemblage from the burned layer including the possible nature of activities at the final occupation phase of the site.  If these materials were left by work crews (like the material associated with the final phase of activity at Kourion), then the diet, ceramics used, and social standing of the individuals could suggest a different assemblage from that left behind by a family.

Despite the origin of the pottery in a layer associated with the site’s destruction and short term occupation, they regard the material as sufficient diverse to qualify as a use assemblage and, therefore, suitable for making larger arguments for the nature of Byzantine cooking practices, diet, and the circulation of Byzantine glazed pottery and utility ware forms.  This was all supported by residue analysis of individual vessels and the quantitative analysis of the entire assemblage.  Apparently the individuals at Sagalassos ate more beef and game than their Late Roman predecessors (who preferred lamb and goat).  Pretty neat stuff.

The article places the material from the assemblage at Sagalassos in the context of the Byzantine Eastern Mediterranean and it will be really useful as we look to document a site with a similar history at Polis in Cyprus.  The material present at Sagalassos has comparanda both on Cyprus and, unsurprisingly, at Corinth in Greece where the study of Byzantine pottery has long held pride of place.  The careful publication of an assemblage from a site like Sagalassos expands the base of evidence for the further study of Byzantine pottery.  The appearance of an article like this in Hesperia should show scholars that there are high-quality journals prepared and willing to publish similar papers.

P.S. Lest you think that I’m just a blogger, you’ll notice that David Pettegrew, Sarah James, and I also have an article in this volume: “Towers and Fortifications at Vayia in the Southeast Corinthia,” Hesperia 79 (2010), 385-415.

Artifact Level Analysis and Places of History

Over the last few weeks, I’ve been working with David Pettegrew to finish writing the analysis of the survey data from the Pyla-Koutsopetria Archaeological Project (PKAP).  Followers of this blog know that this work is a long a term project and involves challenges both on the level of analysis but also organization and description.  In other words, we’ve been working to figure out both how to interpret our survey results, but also how do we organize and describe this data in way that is useful to scholars who are likely to ask different questions from the one’s that our survey set out to consider.

The biggest challenge is moving from the highly granular, artifact level analysis of individual groups of pot sherds to the level of historical time and space.  After all, very few important things happened in the space of a pot sherd or in a time framed absolutely by the life-span or production cycle of an individual vessel.  It is essential to aggregate sherds, space, and time in order to produce historical arguments.  The chronological ranges for artifacts through time depend, in particular, on our understanding of ceramic typologies based on the fabric, shape, and in some cases decoration.  These the chronology assigned to these various typologies are not necessarily meaningful in a historical sense and can be quite individualize to particular objects.

In other words, artifact level analysis is separate from the process of interpreting artifacts across the survey area as chronologically and historically meaningful groups.  Part of the interpretive process involves grouping artifacts together into more or less contemporary groups of object.  This process involves judgement on our part and cannot be applied in the same way across the entire assemblage.

As an example, our analysis of material representing activity across our site from the Classical to Hellenistic (BC 475 to BC 100) periods involves artifacts dated to at least 8 different, overlapping chronological ranges: Archaic-Classical, Archaic-Hellenistic, Classical, Classical-Hellenistic, Classical-Roman, Hellenistic, Hellenistic-Early Roman, and Protogeometric-Hellenistic.  In contrast, our analysis of activities on our site from the Roman period involves artifacts dated to three chronological ranges: Roman, Early Roman, and Late Roman.  Our ceramicist established the date ranges for individual artifacts largely based upon dates established through stratigraphic excavation and completely independent from our interpretation of the site as a whole.  It is common for individual classes of artifacts to receive have different chronological ranges. A sherd from a cooking ware pot might represent a vessel-type produced over a 500 year periods (say, any time during the Classical-Hellenistic period), whereas a fragment of fine ware might derive from a vessel produced during a 4 or 5 decade span of time (say, the early 4th century).  Each of the objects receives a different date and chronological range when documented in the survey area. As a very general rule, utility wares tend to be produced over longer spans of time than fine and table wares, but this has no necessary impact on how and when they were used.

The process of interpreting the artifacts documented by our ceramicist involves us aggregating these objects into chronologically, functionally, and spatially meaningful groups.  Past human activities took place in particular spaces and made use of object produced at different times and for different functions. To produce a picture of what happened in the past at our site that has meaning within these human terms, it is necessary to group together material with different date ranges into assemblages that have meaning in human terms.

For example, here are various maps showing some of the periods aggregated to produce our analysis of the Classical to Hellenistic period at our site:

Archaic-Classical Period

Classical Period

Classical-Hellenistic Period

Classical-Roman Period

Hellenistic Period

Hellenistic-Early Roman Period

To understand trends at our site from Classical to Hellenistic period, the data contained in each of these maps must be analyzed together.  Occupants at our site may have used coarse ware datable only to the Classical-Roman period alongside table wares dated more narrowly to the Classical period.  The Classical period table ware may have represented a households investment in public display, the same household may have stored their agricultural wealth in a series of amphoras that have forms and fabrics used for over 500 years.  To establish the potential spatial relationship between these two activities in an archaeological setting, it is necessary to plot artifacts assigned to different chronological ranges across our site in order to produce assemblages that reflect historical activities.

This task is central to the analysis of artifact level survey data and is the key interpretive move in mediating between the results of archaeological work and historical events in the past.  Our goal as we work to prepare this kind of analysis for publication is to keep this interpretative move as transparent as possible.  Transparency, while sometimes tedious for the reader, opens our analysis for critique on both evidentiary and methodological grounds and reinforced the idea that archaeologists produce the landscape that they interpret.

A Comparison Between a Survey Assemblage and an Excavation Assemblage

For the past few months, David Pettegrew and I have been “publishing” the preliminary results of some experimental analysis conducted over the past year at the Pyla-Koutsopetria Archaeological Project on Cyprus.  As part of the experimental component to our project we were interested in documenting the relationship between the surface assemblage and the assemblage of material produced by excxavation

The archaeological fieldwork conducted on the elevated height of Vigla provided us with an opportunity to compare assemblages produced by intensive pedestrian survey and the excavation of a trench in the survey area.

Longstanding critiques of survey have suggested that the relationship between the surface assemblage and the subsurface material is too problematic for survey to be a technique used to produce a comprehensive view of the landscape. While it is true that there are more variables in the formation processes that impact the creation of a surface assemblage, we should be aware of the potential for a false dichotomy. Excavated assemblages are every bit as much a product of formation processes as those on the surface and as a result, we always have to temper our interpretation of past events with the understanding of the archaeological record as the product of a whole range of physical and cultural transformations. The goal of this comparison then is not to test the surface assemblage against the subsurface material, but rather to suggest that their correspondence indicates that the area may have endured similar archaeological processes.

As with all of our experimental units, the comparison is influenced by significant differences in the spatial comparison between the two sample areas. The surface area of our trench EU 8 represented only 6 sq meters; The two survey units 500 and 500.1 combined to covere over 6000 sq m. Any comparison of area, however, is problematic; the trench had volume and the relatively two-dimensional surface of the survey area did not.

  

The excavation unit produced significantly more ceramic material. The excavation unit produced over 4000 artifacts. By comparison, we counted anywhere from 366 to close to 1000 artifacts from our survey of various areas on the top of Vigla (depending on surface conditions and the number of walkers available), and these samples allowed us to estimate an overall artifact density of between 15,000 – 11,000 artifacts per ha. These are astronomical densities by any reckoning.

While we counted every artifact visible in our 20% sample of the surface, we collected artifacts using the chronotype sampling strategy which required us only to collect each unique type of sherd from each swath.  Using this technique in two campaigns of field walking on Vigla, we collected 963 artifacts with a weight of 27.6 kg.  In contrast, we collected and analyzed every artifact from the excavation area and this resulted in over 4000 artifacts, but this assemblage weighed less than 10 kgs more (37.0 kg) than the assemblage collected from survey.

The nature of the chronotype sampling method used in the survey makes it difficult to find a metric to compare the quantity of material collected from the survey against the quantity of material collected from excavated contexts. The key point for evaluating the correspondence between the two assemblages is not necessarily the quantities of material but rather the presence or absence of material indicating particular activity, periods, or material types present in the area.

Comparing the period date between the two collection strategies reveals that the survey collection produced more chronotype period categories (16 compared to 14) and nine of the periods represented in the survey assemblage were also represented in the excavation assemblage. In general, the survey material represented a longer chronological range with material from later periods present on the surface including material from the Late Roman, Medieval-Modern, and Modern periods. The excavated area, in contrast, produced more material from narrower periods and at least one object from a period earlier than those represented in the survey, a sherd potentially dating to the Bronze Age (for more on broad and narrow periods, see here). This artifact appears to pre-date the earliest phases of architecture present in our trenches and may not represent a past activity on the site. In general, the material from both the survey and the excavation overlap, but the excavation material offered slightly more chronological resolution than the material from the survey.

The diversity of chronological periods in the survey material would appear to extend to the chronotypes represented in each unit. The excavation produced 54 chronotypes, while the survey unit produced 57. There are 30 overlapping chronotypes between the two collection methods. While the different sampling techniques make it difficult to compare the assemblages in a meaningful way, the quantity of material from each area nevertheless provides a very basic matrix for comparing the relative quantity of various types of material from each unit. The survey and excavation both produce a significant number of artifacts from the three rather general chronotypes: ‘Coarse ware, ancient historic’, ‘medium coarse ware ancient historic’, ‘kitchen ware ancient historic’.  The excavation also produced a significant proportion of material from two additional chronotype that were poorly represented in the surface assemblage: ‘animal bone’ and ‘fineware, Hellenistic-Roman, Early’ which made up 6.6% and 5.5% of the excavated material respectively, but less than 1% of the material from the survey. The absence of animal bone on the surface of the ground could be an issue with visibility (white and tan bones do not stand out as well against the buff colored soil) and certainly preservation. 

It is notably harder to compare the potential range of activities present in the area. The chronotype method of collection privileges larger, better preserved sherds (walkers will often discard small or poorly preserved sherds if they find larger examples of the same chronotype). It also tends to under represent very common chronotypes in proportion to the total assemblage. In other words, there are fewer examples of chronotypes such as “medium coarse body sherd, ancient historic” in the survey sample in part because field walkers were instructed not collect multiple examples of this very common type of artifact. In the excavation, excavators collected every example of a “medium coarse body sherd ancient-historic” causing sherds of this type to make up a larger proportion of the total assemblage.

This tendency can be seen in the relative size of artifacts collected from the survey and excavation. From the survey, the collected artifacts were much larger and this probably reflects both our field walkers’ tendency to select larger sherds more frequently than smaller sherds for collection and the difficulty seeing the smallest sherds on the ground from a standing posture. These two tendencies combined to produce an average survey artifact weight almost 30 g as compared to the average size of an excavation sherd that was under 9 g.

The fabric groups present show some significant differences in the assemblage that we can largely trace to different sampling strategies. The survey unit preserved more coarse ware
(47%) whereas the majority of material from the excavated unit was medium coarse ware. The weight of the two fabric groups as a percent of the total assemblage sheds more light on the situation. Medium Coarse wares from the excavation represented 53% by volume, but only 22% of the assemblage by weight. In fact, the average weight of a medium coarse ware sherd is less than 4 grams. In other words, many of the medium coarse fragments of pottery from the excavation are quite small, and these sherds are the most likely to be overlooked during survey. Cooking/kitchen ware, coarse ware, and amphora represented the other significant parts of the excavation assemblage. As the chart below indicates the percentage of weight is significantly different from the proportions determined by counts. In weight amphora and coarse wares combine to make up the majority of material.

The material from survey shows different proportion, but these proportions are significantly biased by our sampling technique that suppressed the collection of redundant artifacts.

Coarse ware is the most common fabric group by quantity and makes up the majority of material by weight. Amphora sherds, which tended to be handles or very large body sherds, represent a massive quantity by weight, but significantly lower percentage by quantity. The opposite is true of medium coarse ware and kitchen/cooking ware.

Similar tendencies are visible from rim-base-handle-sherd analysis (for more on R-B-H-S Analysis, see here). 

The results of this comparison suggest that for the height of Vigla the most major differences between the assemblage produced by survey and that produced by excavation are tied to the different sampling strategies used in these different contexts. At the same time, the basic patterns present in the survey assemblage were also present in the assemblage from the excavation.  The presence of material from the Classical and Hellenistic period, the presence of fine ware, kitchen/cooking ware, and utility wares, and the almost complete absence of earlier material allows us to argue that the site was first occupied in the Archaic to Classical period, saw domestic activities, and then was used less intensively in later periods.  This close correlation of survey and excavation assemblages reflects, in part, the stability of the soils on Vigla and the relative lack of erosion, on the one hand, and the lack of intensive activity during later periods, on the other.  In other words, the surface assemblage and excavation assemblage enjoyed similar sets of formation processes which produced similar assemblages. 

David Pettegrew’s Setting Stage for St. Paul’s Corinth Available as Podcast or Streaming Video

For those of you who could not make it to the David Pettegrew’s 2nd Annual Cyprus Research Fund Lecture, fear not!  We have made David’s lecture available as both a downloadable podcast and as a streaming video.

Here’s the podcast.

Here’s the video.

David’s two days on campus were really exciting.  Not only did he speak to over 50 faculty,undergraduates, graduate students, and members of the community on the Thursday afternoon talk, but he also contributed to the history department’s “brown bag” lecture series on Friday.  At his Friday talk, he presented a great primer to intensive survey archaeology and discussed the ideas of “source criticism” as applied to ancient material culture.  Finally, David took a couple of hours and read Latin with some of our graduate students and undergraduates at our weekly “Latin Friday Morning” reading group.

It is always gratifying to see how much interest there is in the Ancient Mediterranean at the University of North Dakota.  So, if you enjoyed the lecture with here at UND, thanks for coming out! And if you enjoy the lecture via the streaming video or podcast, thanks for listening!  I also should thank Chad Bushy and Caleb Holthusen from UND’s Center for Instructional and Learning Technologies office for not only preparing the video and podcasts, but trouble shooting during the live webcast.

And, finally, thanks to David Pettegrew for agreeing to spend his fall break with us at the University of North Dakota. For more on his research and the Roman and Late Roman Corinthia, check out his blog Corinthian Matters.

Broad Period Artifacts and Survey Analysis: Quantifying what you don’t know

By far the most vexing issues facing most survey projects is the analysis of artifacts datable only to very broad periods of time (a point I brought up in my blog post from last week).  In the work of the Pyla-Koustopetria Archaeological Project, these artifacts are the equivalent to objects of “unknown date” from other survey projects.  The chronotype identification system required our ceramicist to date each artifact even if these dates are exceedingly broad.  As a result, we have significant quantities of artifacts dated to periods that exceed 1000 years in length. These broad periods tend to represent two types of artifacts. 

1. Artifacts that do not fit into any known typology such as body sherds without particularly characteristic marks, fabrics, or shapes.
2. Artifact types that remained in use for long periods of time.  This is most often the case with various kinds of coarse and medium coarse utilities wares probably produced from local fabrics.

In some cases, the fabrics or shapes can tell use enough to allow us to group the artifact into a relatively well-defined, yet still exceedingly broad, date range.  For example, the most common period for an artifact dated to a broad period is “Ancient Historic”.  This is a date range that extends across the entire period of historical antiquity on the island of Cyprus: 750 BC – AD 750.  Almost all of these sherds (89%) are body sherds. The artifacts datable to this period appear over 77% of the total area of our survey and in 84% of the units where artifacts occur.  Statistically, the distribution of “Ancient Historic” artifacts correlates more closely to the overall artifact densities across the entire study area than any other period, broad or narrow (the correlation is .674).  This is particularly significant because artifact counts and the number of artifacts assigned to a particular period are independent variables: our artifact counts are based on the total number of artifacts visible on the ground according to clicker counts and the number artifacts dated to a particular period is a subset of the number of artifacts sampled from the units.  Finally, the fabric types present in Ancient-Historic period more or less parallel the fabric groups present in narrower, better-known, or at least more clearly defined periods (e.g. Classical or Classical-Hellenistic). For the Koutsopetria plain, for example, “Ancient Historic” material appears as coarse ware, medium coarse ware, and kitchen/cooking ware which finds rough parallels with the groups of material present from other periods and the general functional character of the area (for those of you keeping track at home, we call the Koutsopetria plain Zone 1)

Other broad periods from our site represent small quantities of obscure material that stands outside traditional typologies.  For example, there are only two sherds assigned to two chronotypes dated to the Ceramic Age (a Red Micaceous Pithos and coarse ware). Only four sherds assigned to four chronotypes received the generic Ancient date (Ancient Lekane, fineware, kitchen ware, medium coarse ware). Only one chronotype, amphoras, receive the designation Post-Prehistoric.

In old days of survey, sherds dated to broad periods tended to be neglected either at the analysis phase, or more commonly at the sampling and collection phase.  The vast majority of broad period sherds are body sherds (85%) and most of these would not appear to be diagnostic. As a result many traditional collection strategies that privileged diagnostic sherds (feature sherds with distinct marks, rims, handles, bases) would have overlooked broad period material. More recent work has at least assigned the designation of “unknown date” to these broad period artifacts, but rarely do they appear documented in the survey publication.

This material is difficult to correlate with past human activities. At best, it reinforces the notion that certain types of productive practices may have endured for long periods of time without much in the way of visible changing.  It suggests that certain vessel shapes, fabrics, and pottery categories may have continued to serve basic functions within the community, the household, and the economy for long periods of time as well. In Braudelian terms, the apparently long, slow, and relatively unchanging character of such a large part of our ceramic assemblage represents the slow swells of the sea.  The more closely dated and rapidly changing character of fine wares or even the more diagnostic parts of the vessel , for example, which tend to allow us to produce our narrow period assemblages, show the more fickle and rapidly changing nature of ancient ceramic habits.

The value then of our effort to understand the distribution and character of artifacts datable to broad periods from PKAP is that they give us a real measure of how much we do not know about material from our survey area.  And at the same time, reveal that much of the most basic practices typical of the ancient world likewise continues to elude our grasp.

Stones that Speak and some other data from the Pyla-Koutsopetria Archaeological Project

Over the last few weeks I’ve been running what will hopefully be the final set of unique queries on the data from the Pyla-Koutsopetria Archaeological Project’s survey of the site of Pyla-Koutsopetria and its environs.  These queries are mostly following little hunches or the comments that my co-director David Pettegrew made in the margins.  It is re-assuring in some ways to find that I have not overlooked much (and I hope to circulate a working paper of my distributional analysis by the end of this calendar year), and its always fun to find little patterns.  So here are two small PKAP patterns.

For some reason on the edges of comprehension our ceramicist, Scott Moore, documented unworked stones collected in the bags of ceramics collected by our field teams.  Unworked stones collected from the fields are not traditionally regarded as archaeological material (except that their presence in a bag of ceramics has associated them with the archaeological method).  But Scott’s unworked stones do show a pattern. In the last few years, archaeologists have suggested that “background disturbance” or the presence of stones or other materials that look like ceramic objects has a clear correlation with our ability to recover artifacts from the field (the best discussion of this is in Knapp and Given, Sydney Cyprus Survey Project volume).  Presumably our field walker’s tendency to collect stones from the field might reflect a similar pattern. The map below shows the distribution of unworked stones.

And as you can see, a pattern does emerge. Most of our unworked stone comes from units with high or modern background disturbance and this suggests two things.  First, it confirms that the unworked stones are most likely unworked (in some cases Scott documented unworked stone because he was not entirely sure that they were unworked and wanted Nick Kardulias our lithics expert to check them out).  Next, it suggests that background disturbance does influence our field walkers ability to recognize artifacts. It is encouraging to note, albeit in a tentative way, that our field walkers collected objects that they thought might be ceramics and this might give us enough confidence to at least suggest that they did not overlook objects that might be stones.

The second little analysis I ran was on the distribution of faunal remains across the site.  David Reese examined the faunal remains from our excavations in 2008 and 2009 and at the same time looked over a small quantity of faunal remains collected from the survey.  I’ve added to the map the major roads in the area (rather inelegantly displayed unfortunately).  Most of the faunal remains are near the major roads suggesting that at least some of them – particularly the chicken bones – were discarded by passing traffic.  The remains of sheep or goat bones appear cluster in the lowest lying area of the Pyla-Koutsopetria plain.  This area is pretty marshy despite efforts to keep it drained and as a result not generally under cultivation.  This kind of marginal land seems likely to have served as pasture for local flocks.

The final analysis run over the last few days was on some very broad chronological periods into which we grouped material from the survey.  Among the broadest is the “Ancient Historic” period which stretches from around 750 BC to the end of antiquity in AD 749. The transparent dots on the map below show the distribution of artifacts datable only to this long period in the past. Their distribution more or less follows over all artifact densities (with the exception of Kokkinokremos where the ceramicist who read our Iron Age to Bronze Age material used a slightly different designation).  This suggests that artifacts grouped into this broad period are not likely to represent a single class of difficult to identify material, but rather a whole group of artifacts from multiple periods that remains outside of traditional ceramic typologies and chronologies. It is never heartening to see how much material from a survey goes unidentified (or identified in only the broadest possible way), but it is encouraging to see that it does not cluster in suggestive ways.

Some Notes on RBHS Analysis of the Pyla-Koutsopetria Survey Data

No RBHS is not a local high-school to whom I’ve outsourced PKAP data analysis, nor is it a new type of digital hi-def television.  Those letters stand for Rim, base, handle, sherd and represent the basic parts of a ceramic vessel.  Since most of the vessels one finds in survey and even excavation are not whole or are broken and mangled, documenting the rim, base, handle, and sherds from each vessel is an important way to understand how we as archaeologists are able to identify an particular object and assign it to a date, function, and even, sometimes, place of manufacture. It is also helpful in secure, stratigraphic contexts (that not in an unstratified survey context) for identifying the minimum number of possible vessels of a particular type because we know that some kinds of vessels on have, say, one-handle, then a four handles would represent at least four vessels of this type.

David Pettegrew’s research has really set the stage for applying this kind of analysis to the PKAP survey data. He has argued that certainly highly diagnostic artifact types (for example Late Roman 1 amphora handles or Late Roman “combed ware” body sherds) can distort the chronological distribution of material at a site.  Periods characterized by less diagnostic artifact types tend to be less easily associated with a narrow chronology or function and under represented in relation to period defined by more easily identified vessels types.  So isolating the way in which particular periods become visible using our Rim/Base/Handle/Sherd analysis becomes an important to critique our survey data.

Fortunately, the basic system that we use to document our ceramics, the chronotype system, took into account rbhs. The chronotype system required the ceramicist to separate and document as a group, called a batch, according the extant part of each type of vessel present . In other words, we counted in one batch all of the rims from, say, a Roman Amphora and in another batch all the handles from the same kind of amphora. This has allowed us to parse quite finely the character of our assemblages and its relationship to our ability to identify particular types of artifacts based on their individual parts.

So here are some basic observations:

• Of the 19 periods with more than 20 sherds collected using our standard survey procedure, 13 counted the majority of artifacts as body sherds. In other words, for most periods, body sherds represent both the most common and the most chronologically diagnostic type of material.
• Only for the Archaic period were the majority of artifacts identified by one part of a vessel, and these almost all came from one type of vessel, so-called Archaic basket handled storage jars.
• Of the 258 chronotype (that is discrete types of artifacts) that produced extant parts (some chronotypes, like shells or wall plaster fragments, do not produce extant parts that we can easily record), 138 of 55% of these chronotypes were identifiable based on only one extant part. 76% are recognized by only two extant parts and 90% by three. 99% by four extant parts (mainly RBHS).  In other words, most artifacts are only recognizable by one part of the vessel.
• It is interesting to note that the number of chronotypes associated with a particular period has almost no influence on the average number of extant parts by which a vessel is identified. Large number of chronotypes identifiable by a large number of extant parts (4+) come from Roman (40), Late Bronze Age-Hellenistic (18), Ancient Historic (39), Hellenistic-Early Roman (24) vessels. At the same time 4 or more extant parts also appeared for periods with fewer chronotypes, like Classical-Roman (6), Late-Cypriot II-Late Cypriot III (4), and Post-Prehistoric (4).  This means that while the majority of sherds from each period are body sherds, they nevertheless have vessels that are identifiable based on other parts of the artifact.  In other words, our ability to date artifacts to a particular period is independent from the number of vessels with identifiable extent parts. Some periods have three or four chronotypes with lots of identifiable fragments; others have 25 different chronotypes with a mix more and less easily identifiable artifact types.  There does not seem to be a pattern.
• Far more central to the number of parts of the vessel that we can identify is the kind of vessel and their function. Kitchen/Cooking ware produce the most possible extant parts (4+) followed by coarse ware and amphora chronotypes (3.8). Medium coarse ware produced 3.5, while pithos, semi-fine, and fine all produced 2.4 or fewer extant parts per chronotype. This likely has more to do with the shapes of the vessel than the size of the vessel.

This kind of analysis may seem tedious and complicated, but it is important to understand how bias in our ability to identify a particular type of artifact can influence the kinds of chronological and functional landscapes that we create from survey data.  In examining our data in this way, we can really see the point of contact between what our ceramicist does in placing artifacts in particular classes and our historical reconstructions of the landscape.  The entire world of Pyla-Koutsopetria is literally born from the data gleaned from individual artifacts.

Plotting Cut Blocks Across Pyla-Koutsopetria

Between 2005 and 2006, the Pyla-Koutsopetria Archaeological Project documented over 500 features from the Koutsopetria plain.  Most of these features were cut blocks of various sizes, material and descriptions as well as a handful of features associated with ancient agricultural installations (bit of an olive press, some andesite mill fragments, et c.).  Over the past couple of days, I finally got to analyzing this data beyond simply observing that we have lots of cut blocks.  The field team in 2005 and 2006 recorded detailed information regarding the location, size, and in many cases generally descriptions of each block and keyed them into a database that we could integrate with our GIS.

Most of the architectural fragments including cut limestone and gypsum blocks, are concentrated in the immediate Koutsopetria plain where farmers have moved them to stone piles on the edges of the fields. Check out our newest additions (partially edited) to our Omeka Collection: Pyla-Koutsopetria from the Air to get an idea of what these stone piles look like.

 

The most common type of cut block is made of local limestone and probably quarried on site. The majority of the blocks fall between 0.3 and 0.7 m in length and 0.3 and 0.5 in width. For the blocks where all three dimensions are visible, their volume falls between 0.06 and 0.03 cubic meters. This produced blocks of between 75 kg and 140 kg which would be relatively easily moved for construction. Some blocks, of course, could be much larger exceeding 1 m in length and weighing close to 500 kg. With blocks of this size, there is almost no doubt that some large scale, monumental architecture once stood in the immediate area.  Here’s a distribution map.  The grey grid in the background is our survey grid and the color of the dots relates to the volume of the stone.

We also documented a significant quantity of cut gypsum block.  Since marble did not naturally occur on the island, Cypriots often used gypsum as a substitute in more elaborate buildings.  These blocks are generally similar in size to the cut limestone blocks with lengths of around a half a meter and widths of 0.3 meters. The average volume of blocks was similar to that of the cut blocks with only a few blocks exceeding 0.1 cubic meters. There were slightly more smaller blocks owing most likely to the more friable character of gypsum. Most blocks fell between 0.01 and 0.06 cubic meters. Gypsum has a lower density than limestone and the blocks had correspondingly lower weight usually between 25 kg and 140 kg. Many, much smaller fragments of gypsum were scattered across the fields and several very large blocks appeared clustered together. Here’s a map:

 

Finally, we also discovered a small quantity of marble from across the site.  Most of these came from the central area of the Koutsopetria plain embedded in rock piles at the edges of cultivated tracks of land. The marble fragments are small < .30 m in maximum length and relatively thin <.04 m suggesting that all but one marble fragment was revetment or floor slabs. The wide distribution of material perhaps indicates that there were several marble clad buildings on the plain of Koutsopetria even though so little marble survives.  Here’s a map:

The next step in analyzing this material is considering its relationship to the re-used blocks found in the excavations at Koutsopetria and the construction techniques used in the fortification wall surrounding Vigla.  It certainly seems possible that the majority of cut stone blocks scattered around the Koutsopetria plain came from the easily quarried fortifications at Vigla and perhaps also the extensive walls surrounding the Bronze Age site of Kokkinokremos.  Gypsum blocks had fairly limited uses architecturally owing to their lack of strength and value as prestige materials.  The gypsum fragments from around the site probably served in specific places in buildings and comparing their sizes to in situ blocks from elsewhere on the island might give us some idea of how they were used.

 

More Experiments in Intensive Pedestrian Survey

More guest-posting brilliance from our esteemed guest blogger, David Pettegrew, the co-director the Pyla-Koutsopetria Archaeological Project and the 2010 Cyprus Research Fund speaker.  Be sure to check out his posts on Tuesday, Wednesday, and Thursday.

Last Thursday, we introduced the survey experiment that PKAP conducted in June 2010 to assess the relationship between the number of artifacts that we see when we walk across a survey unit and the number of artifacts actually on the ground.  In other words, we wanted to assess how effective our survey methods are in actually assessing what was on the ground.  On Thursday, we compared the artifact densities detected by the project’s untrained student fieldwalkers to those counted by trained senior staff members.  Today we will discuss the second phase in our 2010 experiment, an assessment of the total population of all artifacts on the surface of select subunits.  This part of the experiment was designed to give us a total count of all surface artifacts that can be compared with the artifact counts reported in yesterday’s discussion.

We began by selecting four 10 x 10 m subunits based on the densities of the 10 x 10 m artifact densities counted by the experienced senior staff members.  As with past experiments (published in the RDAC 2007), we selected our 4 subunits to represent the range of density variation: the lowest density quartile (G15), highest density quartile (G9), and two middle quartiles (G1 and G6).  Each total subunit was 10 x 10 m, representing 1/16 (6.25%) of the 1,600 sq m survey unit.

To vacuum a high-density unit, you really have to spend a lot of time picking individual artifacts off the ground.  For each of our units, students Andrew, Zane, Valerie, and Luke, and I  walked very slowly in adjacent passes across each selected square gathering together in 1 or 2 corners of the unit all the artifacts present.  An initial pass was never enough for we observed how many artifacts we missed initially.  Usually two additional passes were necessary to vacuum the surface completely, and each pass involved either crawling on hands and knees, or bending so that you had a closer view of the ground.  I have to admit that my back and neck got sore after a while of this.

The results of this “total collection”, shown below, are interesting to compare with the “pedestrian survey counts” discussed yesterday.  You have to keep in mind with the comparison that the pedestrian counts represent a 20% sample of each subunit while the total collection counts represent a 100% sample.  You have to multiply the pedestrian count by a factor of 5 to estimate the “total putative count” (i.e., an estimation of what the total count would be for 100% of the unit) for the pedestrian-walked unit.

The first outlined set of grid units below shows the total counts from each of the total collection units.

The second set of grids compares the total collection counts with the pedestrian survey counts in parentheses (multiplied by 5 to create the 100% putative sample).  

The third shows the factor difference between these two types of counts.  

Here is where it gets even more interesting.  We can estimate that the 940 artifacts experienced fieldwalkers counted through pedestrian survey across the entire unit (i.e., the pedestrian counts from 4 walker swaths) would produce a putative pedestrian survey count (factoring for the 20% sample) of 4,700 artifacts.  In other words, had we walked 100% of the unit, we would have counted about 4,700 artifacts.  Now, if total collection (vacuuming) produces on average 2.96 times the number of artifacts as pedestrian survey, we can estimate that there were 13,212 artifacts actually on the surface of the ground.  To provide some perspective, we collected and brought back to the museum 8,788 total artifacts from the 252 grid squares of Koutsopetria and 19,657 total artifacts from our survey of the entire Pyla-Koutsopetria area.  A single survey unit at Koutsopetria totally collected would produce 1.5 times the number of artifacts sampled from all 252 grid squares at Koutsopetria and .67 of the total artifacts sampled across the entire Pyla area.  If we were to apply the same multipliers to all 252 forty x forty meter grid squares, i.e., the main part of the site of Koutsopetria, the total artifact count of 19,182 would produce a putative total count of 95,910.  Our estimated total population of artifacts (based on the 2.96 factor) is at least 284,894 (and in reality, poor visibility in many units often limited our sample to 50% of the ground).  This is *why* sampling is important!

As for TIME, total collection requires a huge commitment.  Although we (for clarification here, “we” means David – Bill) initially considered surveying all 16 subunits, i.e., an entire 40 x 40 m unit, this proved unrealistic given the time it took for 5 individuals to vacuum a single subunit: 1.5 hours each for G1 and G6, 2 hours for G9, and 1 hour for G15.  Using the total time it took to hoover 25% of the grid square (6 hours) as an index for hoovering this unit, we estimate that 5 individuals could hoover a high-density 40 x 40 m unit in about 24 work hours or well over 100 work hours!  If the typical survey work day is 6 hours long (say, 6AM-noon), it would require 4 full days of a team collecting artifacts from the surface.  Truly this would be an incredibly time intensive task!  By contrast, sampling 20% of the unit through pedestrian survey takes about 20-30 minutes.  In this perspective, total collection requires 72 times more time than pedestrian survey collection!

One final comparative result is interesting to note here.  The “other” category increases dramatically through total collection, including numerous pieces of ancient glass (9), lithic stone artifacts (7), shells (24), slabs (13), gypsum (141), ceramic bricks (2), stone vessel (1), marble revetment (3), and a ceramic tessera or gaming piece.  Although total collection was time intensive, this sort of qualitative information is quite useful in filling out our picture of the overall survey unit and indicates something of the functional variability within each survey unit.

Tomorrow, we will conclude our discussion of experiments with an overview of ceramic fabric categories.  Stay tuned!

 

Experiments in Intensive Survey at Pyla-Koutsopetria

Another guest post from our esteemed guest blogger, David Pettegrew, the co-director the Pyla-Koutsopetria Archaeological Project and the 2010 Cyprus Research Fund speaker.  Be sure to check out his posts on Tuesday and Wednesday.

When I announced my plans to conduct a survey experiment where we would “vacuum” an entire 40 x 40 m unit, Dimitri and Bill both laughed and told me that I had to try it simply for its absurdity.  The 40 x 40 m survey unit was our standard size for the 252 units that we laid out across the Koutsopetria plain .  As far as survey units go, 40 x 40 m (or 1,600 square meters) is a relatively small unit compared to that typically employed by those who conduct distributional survey.  At the same time, when on the group, 40 meters is still vast when compared to the dimension of most lived space.  After all, a 40 x 40 meter unit is over 130 square feet on a side and over 17,000 square feet which makes a single survey unit much larger than even the most over-sized suburban McMansions.  The reason that my suggestion was humorous, however, had to do with the method I proposed for collecting artifacts.  In our typical pedestrian survey, we only looked at 20% of the surface of the unit (for a more reasonable and suburban 3,400 square feet) and only collected each unique artifact from what we saw on the surface.  My proposal was more extreme: get down on our hands and knees and completely “vaccum” (or “hoover”) all the artifacts from 100% of the unit to produce an exhaustive (and exhausting!) total collection rather than a quick 20% sample.

Why?  I had the suspicion that the amount of artifacts we see when we walk across the unit is but a fraction of the total number of artifacts actually on the ground.  The suspicion was based on experiments conducted in 2004 & 2006 where we ‘vacuumed’ artifacts from a 5% sample of our 40 x 40 m units, producing on average artifact counts that were 4 times greater than that produced through our 20% sample using pedestrian survey.  We also proved through these experiments that the substantially larger number of artifacts did not really contribute much new chronological or functional information that warranted the additional investments of time and energy.  We published a report on those experiments in an article by the authors in the Report of the Department of Antiquities, Cyprus 2007.  However, we were aware of the substantial fluctuations of artifacts within 40 x 40 m units and the risk of a 5% sample (80 sq m) being unrepresentative of the unit as a whole (1600 sq m).  The point of our 2010 experiments, then, was to test the results with a much more robust sample.  While I initially wanted to vacuum 100% of the unit, time constraints prohibited me to vacuuming 25% of the unit.  Even still, 25% of the unit is 5 times greater than what we sampled in 2004 and 2006.

Due to the limited time for fieldwork this season (and time constraints were one of the reasons that we sampled the units to begin with!), we could only resurvey a single unit placed in the highest-density area immediately northeast of the excavated apse of the early Christian basilica.  We picked this unit to overlap with our very first Discovery Unit, a grid square of 40 x 40 m surveyed in 2004 northeast of the enclosed excavated part of the site of Koutsopetria.  We divided the 40 x 40 unit into sixteen 10 x 10 m subunits, each representing 6.25% of the overall unit area (1,600 sq m).  The grid squares have been given the prefix of G followed by a number between 1-16, as the following plan shows.

In our interest in comparing artifact counts noted during pedestrian survey—where a surveyor walks across the unit examining a 2 m wide swath and counting all pottery, tile, lithics, and other artifact types—with the total population of artifacts actually on the surface, we implemented two stages to the experiments.  The first stage (pedestrian survey) we will report on today.

We began by having four fieldwalkers walk across the unit, recording all artifacts visible in their swath, giving a 20% sample of every 10 m of space across a 40 m transect.  We collected ‘sub-tract’ artifact counts every 10 meters to produce density figures for each of the subunits (G1-G16) and assess the fluctuating density of pottery, tile, and lithic artifacts within a survey unit.

We collected the data for pedestrian survey three times.  The results of these three separate pedestrian survey exercises are shown in the four figures below.  The numbers represent artifact counts of each type (pottery, tile, other, and total), and the gray shaded columns with orange numbers represent the total artifact count for the swath per fieldwalker.  

The first time (see figure 1.1 below) a group of untrained students walked the units—Andrew, Luke, Valerie, and Zane—who who had only seen artifacts at the museum and not in their “natural” (or better, archaeological) contexts.



A steady light rain the following day provided the chance for these same students to rewalk the unit a second time (see figure 1.2 below) with artifacts slightly more visible as a result of the washing of the dust.

Finally, a group of experienced fieldwalkers—David Pettegrew (DKP), Dimitri Nakassis (DN), and Bill Caraher (WRC) —walked the unit and counted artifacts (see figure 1.3).

Hence, the variables in these three episodes of pedestrian survey were experience, and, to a lesser extent, the amount of dust and dirt obscuring the surface of the pottery.  Otherwise, between episodes environmental factors were constant, as were methodological factors and figure 1.4 shows the average of all the counts produced.

We walked these units on June 9 and 10 and each took between 15 minutes and half an hour.

Comparing simply the total artifact counts (the bottom right grid within each of the outlined figures), it is interesting to note that the rain appears not to have made a difference overall in density counts between units [1.1] and [1.2].  Although one student count went up significantly after the rain (LHM: 118  243), and another student count was slightly greater (AMH: 200  241), VAW’s total counts were essentially unchanged (335 to 334), while ZRB’s total counts actually declined (238).

As far as the other variable (experience) goes, there were some significant disparities between experienced walkers and inexperienced walkers as evident in counts for particular grid squares (compare G1 for [1.1] and [1.3]).  Otherwise, the overall artifact counts were comparable for the units: the lowest-density and highest-density subunits occurred between all three walking episodes.  If we look at total artifact counts for each unit as a whole, students counted 942 artifacts in [1.1] and 1056 artifacts in [1.2] while experienced walkers counted 940 artifacts in [1.3].  That is remarkably close!  

We noticed one major difference, however, in the “other” category, which includes all artifacts besides pottery and tile: marble revetment, gypsum, shell, ancient glass, and ground stone agricultural implements.  The experienced field walkers noted 2-4 times the number of other artifacts in [1.3] than inexperienced fieldwalkers in [1.1] and [1.2].  An experienced walker counted 4 lithic artifacts (chipped stone & ground stone) in G3 and G7 that an inexperienced walker missed.