Antonia, you ask if you are being too optimistic. My answer is "yes."
What Alisdair describes in his response is, I would wager, a universal
reaction to attempts at quantifying--in a rational, repeatable,
comparable, "scientific" way--not only 19th-c ceramics, but nearly
anything archaeological. And besides these methodological problems,
before such quantifications can tell us what we usually want them to
tell us, we have a whole boatload of assumptions about standardization
in data-recovery methods, standardization in site sample reliability
and validity, standardization in classification (e.g. "white granite
and ironstone" etc.). And, on top of all of these, we have to have a
deep faith that there was some "standardization" (or at least
normalization) of refuse disposal behavior and post-depositional
factors, etc.
Don't get me wrong: I love numbers. I can talk half intelligently
about the inferences to be drawn from the distributions of residuals
about a multiple regression, or about the various benefits and
problems with different factor-rotation schemes or clustering
algorithms. I think that's neat stuff, and I always have. I could
translate this letter into FORTRAN, or SAS or SPSS! But I still do not
believe that numbers are necessary to archaeology. In fact, because I
have very good quantitative training, I generally don't believe most
numberical treatments are *appropriate* for archaeology.
Don't get me wrong. I still have minimum vessel counts done for all my
projects, but I try my best to convince those doing the work that it
is an art, not a science; that consistency within the project is more
useful than standardization between projects.
For several years many archaeologists--the Urban archaeology forum of
the SHA comes especially to mind--have bemoaned the lack of
standardization and comparability. It is truly (IMHO) a pipe-dream.
Antonia, I am not suggesting you give up the idea of data-basing your
ceramics. Far from it. Again, I have used not only mainframe data
bases like SPSS, but I've even written my oww using BASIC or VISICALC
since the day I got my first Apple II computer. Today we use FoxPro
and we are dying to figure out how to afford ReDiscovery. I just don't
expect our data to be comparable to yours--at least not in a direct
and unambiguous way. I don't even expect my own data from 15 or 20
years ago to be comparable to what we are doing today. Everybody
thinks there's a pot of gold at the end of that rainbow, but I rather
think the appropriate metaphor has to do with a benighted and naked
emperor.
Dan Mouer
http://www.freedomnet.com/~dmouer/homepage.htm
>
> On Thu, 5 Jun 1997 11:28:40 SAST-2 Antonia Malan wrote:
>
> > (1) In 1994 there was desultry correspondence about standardising 19th
> > century ceramic analysis. Did anything come of this? Did anyone
> > crack the secret? We have sorted out the Dutch 18th century, and the
> > British period up to 1840 or even 1860; now we are onto the later
> > 19th and early 20th century. (Neither POTS nor Miller's Indices are
> > directly relevant to the Cape, so we had to set up our own system.) We beli
ev
> e
> > it is indeed necessary to deal with
> > the later period and its problems. We are tackling MNVs, basic description
an
> d
> > identification, a simple database (based on MSAccess). Are we too
> > optimistic? Any ideas or support welcome.
> > Dr Antonia Malan
>
> Standardising minimum vessel counts present a considerable problem,
> particularly for the period 1780-1820 when creamware and pearlware
> co-exist, but also whenever there are going to be significant amounts
> of decorated and undecorated wares in an assemblage (as will often
> happen in the mid to late 19th century - well, at least it does outside of
> South Africa ;-)
>
> The problem is as follows: Assume that you count the vessels based on rim
> sherd occurrence and unique body sherds. Usually, you will significantly
> underestimate the number of undecorated wares (whether creamware or
> ironstone/white granite) as decorated unique body sherds are far easier to
> identify. Let's assume that you therefore cut out the unique body sherds and
> base your count on rim sherds only. You will _still_ usually underestimate
> the number of undecorated vessels as it is far easier to identify a vessel fro
m
> a
> decorated rim (whether painted, shelledge, transfer-printed, etc.) than from a
> collection of undecorated rims. This discrepancy will be particularly pronoun
ce
> d
> on sites with highly fragmented assemblages, such as ploughed sites.
>
> For example, at the Slave Quarter site at Thomas Jefferson's Poplar
> Forest (a ploughed, exceptionally fragmented assemblage), the 1636 pearlware
> sherds were represented by an MVC of 53 vessels, or 1 vessel for every
> 31 sherds (I've rounded decimals). The 2262 creamware sherds were
> represented by 33 vessels, or 1 vessel to every 68 sherds. The dark green
> bottle glass (arguably even harder to distinguish than undecorated creamware)
> came to 29 vessels for 2650 sherds, or a rather worrying 1 vessel for every
> 91 sherds (Brooks 1996:39). This isn't a matter of people being careless
> when counting, this is a very real problem that has very real consequences
> for comparing data and for analytical techniques that rely on the MVC
> (mean ceramic dates and Miller indices spring to mind).
>
> The basic flaw would appear to be the implicit assumption that a
> minimum vessel count is inherently coherent across an assemblage. It isn't.
> "Each MVC for part of an assemblage is an MVC for that part alone rather being
> a directly comparable absolute. In other words, the pearlware MVC (or perhaps
> more accurately, the decorated vessel count) does indeed estimate the minimum
> number of pearlware (or decorated) vessels, just as the creamware (or, if you
> will, the undecorated) MVC estimates the minimum number of creamware
> (or undecorated) vessels. The key words here are "estimates" and minimum".
> Archaeologists have a tendency to treat MVC's as an absolute quantification of
t
> otals
> rather than estimates of the minimum, and they are almost always the latter"
> (Brooks 1996:40).
>
> Like Dr. Malan, I'm open to suggestions. Count systems based on
> assemblage weight or vessel completeness percentage would not appear to
> be an option. For the former, you usually need to be able to separate all of
th
> e sherds
> by form, and have an analogous complete vessel to weigh. The latter is highly
> subjective in the absence of analagous vessels for comparison, although experi
en
> ce
> will help in that regard. Both of these might be feasible when you have an
> assemblage with fairly large fragments, but they'll border on the nightmarish
> when done with a highly fragmented assemblage. Perhaps all data are more
> subjective than many of us would like to admit...
>
> As far as basic description and identification are concerned, if
> we can agree on whether to refer to "ironstone" or "white granite", we
> might be on to something ;-) Someone else can tackle that one, though.
>
>
> Alasdair M. Brooks
> University of York
>
>
> Source cited:
>
> Brooks, A. 1996 "Analysis of Ceramics and Glass from the Quarter Site"
> Mss. on file, Thomas Jefferson's Poplar Forest (Forest, Virginia)
>
>
>
|