Warren Co. Tennessee Example: Group Dynamics, Social Networks and Cultural Capital
(c) Douglas R. White 1998
site under construction: see also new abstracts


Correlations: Warren County, Tennessee 1850 Census Database Project
The graphic below represents our first collaboration with a dedicated U.S. genealogist, contacted through the email exchange below the graphic. The graph shows 181 marriages from Ann Turner's database that relinked different circles of families in Warren Co., Tennessee (map page by Fred Smoot). There are 495 "relinked" families in toto, or about 1000 individuals of the more than 4000 that Ann had identified as linked into one "tree" -- which she defines as people connected by ties of blood or marriage. That's quite a high level of endogamy in this community.
The graph is three dimensional with the size of the nodes showing the third dimension (the larger circles are more ancestral, the smaller ones in lower generations). The families are scaled by the proximities of their linkages. What emerges is a two large clusters of families, one larger than the other: possibly this corresponds to the county seat. There is another large cluster thinly relinked to it (maybe the second largest town?) and then one or more smaller clusters. It would of course be very interesting to "correlate" town of residence from the census with the observed clusters from the proximity scaling, but that starts to be a bit of work.
Examining Ann Turner's database, we find plenty of data for further analysis:
3,677 occupations listed
--772 land values listed
7,433 dates (mostly birth)
--262 death dates
plus locality data as explained in Ann's email below (otherwise only 21 specific instances are given for McMinnville in the GED file, plus 2 for Oakman and 1 for Rock Island). This promises to be a very interesting and profitable collaboration, not only in its first stages. Thanks for contacting us, Ann! Warren County, Tennessee Project of Ann Turner

Date: Fri, 25 Dec 1998 09:34:34 -0800
From: Ann Turner
Reply-To: APTurner@aol.com
To: Douglas R White PhD
Subject: Linkages Projects dataset

Hello, Dr. White --

I came across your web site during some semi-random browsing. I wonder if you might be interested in a database I am developing, based on the 1850 census for Warren County, TN. All the families were entered into a genealogy program (PAF) as family groups whenever it seemed "plausible." I then make corrections and link more and more families together as I correlate the census records with land, marriage, probate, and other types of records. The county had about 8000 people, and now (with a few additions of records) over 4000 of these have been linked into one "tree" -- which I define as people connected by ties of blood or marriage. I have a web site at
http://members.aol.com/apturner/wctnhome.htm which explains the project in more detail. If you are interested in the dataset, you can download a copy of the GEDCOM file at that site.

I would be very interested to hear your analysis if this type of data is of any use to you.

Ann Turner

Date: Fri, 25 Dec 1998 18:25:41 EST
From: APTurner@aol.com
To: drwhite@orion.oac.uci.edu
Subject: Re: Linkages Projects dataset

In response to Doug White's message of December 25th 1998:
> I was interested to see using my methods
> whether in addition to a giant tree of connected families we
> would network of endogamous families. The enclosed graphic is
> the result: it shows 181 marriages that relinked different
> circles of families, for 495 "relinked" families in toto.

What an interesting analysis! I knew there were lots of intermarriages between the same groups of families, but the graphic representation really shows it off.


> the proximities of their linkages. What emerges is a two
> large clusters of families, one larger than the other: possibly
> this corresponds to the county seat. There is another large
> cluster thinly relinked to it (maybe the second largest town?)
> and then one or more smaller clusters. It would of course be
> very interesting to "correlate" town of residence from the
> census with the observed clusters from the proximity scaling, but
> that starts to be a bit of work.

The county is quite rural -- it has one "large" town in the center, McMinnville, and a half dozen very small towns. (I've only been there once, so I'm not speaking from intimate knowledge). The records in the database are in census enumeration order, which usually corresponds to geographic proximity. If the order of records was not preserved when you imported the GEDCOM file into your program, the REFN field can be alpha sorted -- it is the household number and individual number within the household, e.g. 0697-04. Would this give you enough information to correlate the location with your clusters? The county was divided into 16 districts, but the district numbers have changed over the years and no one is quite sure what they were in 1850. It wouldn't surprise me if McMinnville was done first -- it is also the largest district. I'll include a list of REFN for each district, since there might be breaks in geographic proximity at those points.


> and then one or more smaller clusters. It would of course be
> very interesting to "correlated" town of residence from the
> census with the observed clusters from the proximity scaling, but
> that starts to be a bit of work.

The county is quite rural -- it has one "large" town in the center, McMinnville, and a half dozen very small towns. (I've only been there once, so I'm not speaking from intimate knowledge). The records in the database are in census enumeration order, which usually corresponds to geographic proximity. If the order of records was not preserved when you imported the GEDCOM file into your program, the REFN field can be alpha sorted -- it is the household number and individual number within the household, e.g. 0697-04. Would this give you enough information to correlate the location with your clusters? The county was divided into 16 districts, but the district numbers have changed over the years and no one is quite sure what they were in 1850. It wouldn't surprise me if McMinnville was done first -- it is also the largest district. I'll include a list of REFN for each district, since there might be breaks in geographic proximity at those points. 


REFN      District   NUMBER OF ENTRIES

0001-0177     9         1203       363  Navy Blue     30



0178-0236    15          530       209  Dark Green    36

0237-0273     2          124        37  Light Green    8

0274-0368     3          519       187  Red           41

0369-0389     1          125        42  Yellow        10

0390-0528     4          887       363  Blue          83 (County Seat??)

0529-0613     5          490       163  Pink          35



0614-0656     7          277        97  Orange        20



0657-0690     6          287        68  White          4



0691-0787    14          567       224  Brown         29

0788-0890    11          576       293  Olive         43

0891-0946    16          404       127  Light Blue    26



0947-1009    12          358        78  Grey           2



1010-1123    13          677       107  Black         18



1123-1276    10          887       206  Drab Green    19



1277-1385     8         1717       555  Brown        117 (seem to be dispersed around the County Seat)

Here is a bicomponent of the couples coded by districts that shows a principal basis of the clustering in the graph to be location This GED file that follows is file made by Ann Turner (January 1999) with District codes (from Ruby's work) inserted before the names. Now, here are all 3191 families in a spring embedding, colored by their districts.
Date: Tue, 09 Feb 1999 23:22:32 -0800
From: Fred Smoot [dogtrot@well.com]
To: TNWARREN-L@rootsweb.com
Subject: Re: Populated places

Ann et al,

Only these Post Offices were active in Warren County in the year 1850.

Caney Fork
Clearmont
McMinnville
Rock Island
Rocky River
Rough and Ready
Shippingport (private)
Trousdale

There had been others before 1850 that had closed and many were created after 1850.

Even if a place had a Post Office, it did not mean it was located in a town. One building at crossroads could have a 3rd class Post Office.

The 1850 folks in Warren County were mostly rural. Farms that is. Often they lived along the waterways. The first land claims in a county would be along rivers with its good rich bottom land.

My cousin Lib tells me that she remembers going from Shellsford to Crisp by wagon. If the river was up, they had to go much further west to a bridge, but is the river was down, they would cross a ford just west of Crisp. (Shields Ford)

So think of clusters of farms grouped along a river, with a rural church being the anchor. Warren County did not jump into the 20th Century, it sort of eased its way in just a little later then we would think.

Fred


> [from Doug White to Ann Turner, cont.] If you would like, I could
> put a link to your page from one of my database pages, or even a
> graphic such as this if you like showing some of the structure
> and the great interest which your project might hold for others.

I would be delighted if you would put the graphic on your site, and I could then link to it from my site.

Let me know if there's anything further I can do. I'll be out of town next week, so I might not get back to right away, but this is fascinating stuff! I'm glad to see another use for this data -- I've thought all along that it might have some interest outside the genealogical community.

Ann

Date: Fri, 25 Dec 1998 23:13:23 EST
From: APTurner@aol.com
To: drwhite@orion.oac.uci.edu
Subject: Re: Linkages Projects dataset

Just another stray thought about the Warren County census database. There's also some socioeconomic data in the form of land value and occupation. These are stored in the notes section of the GEDCOM file, but there is also a flat comma-delimited file with separate fields in WCTN50DB.ZIP at
http://members.aol.com/apturner/wctnhome.htm. The REFN field (household #) is the same in both the GEDCOM and the flat file. Maybe one of your students would be interested in analyzing that for a class project.

Ann

UCI Student Intern: Ruby Salmo rsalmo@uci.edu The central blue node in this graphic shrinks the multiconnected component of relinked families and shows the rest of the genealogical tree in 3D generations.



Potential collaborative work on local populations

advances home page

cultural kinetics home page

linkages home page