dendrochronological dating of historic log cabins on...

11
ORIGINAL ARTICLE Dendrochronological dating of historic log cabins on the Belle Meade Plantation, Davidson County, Tennessee, USA Lauren A. Stachowiak 1 Elizabeth A. Schneider 2 Maegen L. Rochner 1 Savannah A. Collins 1 Henri D. Grissino-Mayer 1 Received: 14 November 2014 / Accepted: 1 May 2015 / Published online: 10 May 2015 Ó Springer-Verlag Berlin Heidelberg 2015 Abstract The Harding Cabin is a one and one half story, double pen log cabin located on the Belle Meade Plantation in Davidson County, Tennessee, USA. Historical accounts attribute the eastern cabin to Daniel Dunham who originally settled in the area ca. 1780 and the western cabin to John Harding who purchased the land from the Dunham family in February 1807. We used dendrochronological dating techniques to determine the years of harvest for trees cut and used to build both cabins. We found that both cabins were predominantly made from two genera not commonly used to build log cabins in the southeastern U.S., Fraxinus (ash) and Ulmus (elm). We obtained ring width measurements from 30 cores from all walls on both cabins and used graphical and statistical techniques to crossdate the tree-ring series and create floating chronolo- gies for both species. We crossdated these with reference chronologies from the ITRDB. The Fraxinus chronology was dated from 1657 to 1805 while the Ulmus chronology was dated from 1583 to 1805. Cutting dates indicated that most if not all trees used to build both cabins were cut in late winter/early spring of 1807 before the trees had broken dormancy. We propose that both cabins were built the same year but by different families. We propose that the Harding family built the eastern cabin while Harding’s slaves built the western cabin, originally at a different lo- cation on the property. Once the mansion was completed by 1820, the slave cabin was relocated beside the Harding cabin. Keywords Dendroarchaeology Á Tree rings Á Construction history Á Log cabin Á Belle Meade Plantation Á Southeastern U.S. Introduction Agencies that manage state and national historic sites recognize the importance of the accuracy provided by dendrochronological dating and have increased the demand for tree-ring dating of historic structures. These agencies are also recognizing that written records and oral histories may have enhanced the historical significance of a structure and may have added one to two generations to the age of a structure (Grissino-Mayer 2009). Any wooden structure can be analysed using tree-ring dating techniques provided certain criteria are met and these criteria are carefully evaluated prior to sampling for tree-ring analysis. For ex- ample, the species of tree(s), number of rings, accessibility, and preservation quality dictate whether tree-ring dating will be successful. Furthermore, such empirical analyses of tree-ring dates for historic structures provide valuable in- formation for those wishing to achieve listings in the Na- tional Register of Historic Places managed by the National Park Service (Grissino-Mayer 2009). The use of tree rings in archaeological research in the southeastern U.S. has developed into a reliable and replicable science over the past 35 years. Stahle (1979) spearheaded the use of Communicated by L. A. Newsom. Electronic supplementary material The online version of this article (doi:10.1007/s00334-015-0535-x) contains supplementary material, which is available to authorized users. & Henri D. Grissino-Mayer [email protected] 1 Laboratory of Tree-Ring Science, Department of Geography, The University of Tennessee, Knoxville, TN 37996, USA 2 Department of Geography, Environment and Society, The University of Minnesota, Minneapolis, MN 55455, USA 123 Veget Hist Archaeobot (2016) 25:105–115 DOI 10.1007/s00334-015-0535-x

Upload: hoangkiet

Post on 08-May-2018

214 views

Category:

Documents


1 download

TRANSCRIPT

ORIGINAL ARTICLE

Dendrochronological dating of historic log cabins on the BelleMeade Plantation, Davidson County, Tennessee, USA

Lauren A. Stachowiak1 • Elizabeth A. Schneider2 • Maegen L. Rochner1 •

Savannah A. Collins1 • Henri D. Grissino-Mayer1

Received: 14 November 2014 / Accepted: 1 May 2015 / Published online: 10 May 2015

� Springer-Verlag Berlin Heidelberg 2015

Abstract The Harding Cabin is a one and one half story,

double pen log cabin located on the Belle Meade Plantation

in Davidson County, Tennessee, USA. Historical accounts

attribute the eastern cabin to Daniel Dunham who

originally settled in the area ca. 1780 and the western cabin

to John Harding who purchased the land from the Dunham

family in February 1807. We used dendrochronological

dating techniques to determine the years of harvest for trees

cut and used to build both cabins. We found that both

cabins were predominantly made from two genera not

commonly used to build log cabins in the southeastern

U.S., Fraxinus (ash) and Ulmus (elm). We obtained ring

width measurements from 30 cores from all walls on both

cabins and used graphical and statistical techniques to

crossdate the tree-ring series and create floating chronolo-

gies for both species. We crossdated these with reference

chronologies from the ITRDB. The Fraxinus chronology

was dated from 1657 to 1805 while the Ulmus chronology

was dated from 1583 to 1805. Cutting dates indicated that

most if not all trees used to build both cabins were cut in

late winter/early spring of 1807 before the trees had broken

dormancy. We propose that both cabins were built the

same year but by different families. We propose that the

Harding family built the eastern cabin while Harding’s

slaves built the western cabin, originally at a different lo-

cation on the property. Once the mansion was completed

by 1820, the slave cabin was relocated beside the Harding

cabin.

Keywords Dendroarchaeology � Tree rings �Construction history � Log cabin � Belle Meade Plantation �Southeastern U.S.

Introduction

Agencies that manage state and national historic sites

recognize the importance of the accuracy provided by

dendrochronological dating and have increased the demand

for tree-ring dating of historic structures. These agencies

are also recognizing that written records and oral histories

may have enhanced the historical significance of a structure

and may have added one to two generations to the age of a

structure (Grissino-Mayer 2009). Any wooden structure

can be analysed using tree-ring dating techniques provided

certain criteria are met and these criteria are carefully

evaluated prior to sampling for tree-ring analysis. For ex-

ample, the species of tree(s), number of rings, accessibility,

and preservation quality dictate whether tree-ring dating

will be successful. Furthermore, such empirical analyses of

tree-ring dates for historic structures provide valuable in-

formation for those wishing to achieve listings in the Na-

tional Register of Historic Places managed by the National

Park Service (Grissino-Mayer 2009). The use of tree rings

in archaeological research in the southeastern U.S. has

developed into a reliable and replicable science over the

past 35 years. Stahle (1979) spearheaded the use of

Communicated by L. A. Newsom.

Electronic supplementary material The online version of thisarticle (doi:10.1007/s00334-015-0535-x) contains supplementarymaterial, which is available to authorized users.

& Henri D. Grissino-Mayer

[email protected]

1 Laboratory of Tree-Ring Science, Department of Geography,

The University of Tennessee, Knoxville, TN 37996, USA

2 Department of Geography, Environment and Society, The

University of Minnesota, Minneapolis, MN 55455, USA

123

Veget Hist Archaeobot (2016) 25:105–115

DOI 10.1007/s00334-015-0535-x

dendrochronological methods in the dating of historic

structures throughout much of Arkansas, while Grissino-

Mayer and his students and colleagues have successfully

dated over 50 historic structures in Tennessee, Virginia,

North Carolina, Kentucky, Florida, Georgia and Michigan

(Grissino-Mayer and van de Gevel 2007; Henderson et al.

2009; Mann et al. 2009; Slayton et al. 2009; Garland et al.

2012; DeWeese et al. 2012; Grissino-Mayer et al. 2010;

Harley et al. 2011; Grissino-Mayer et al. 2013; Stachowiak

et al. 2014).

The Belle Meade Plantation, located in Davidson

County in central Tennessee, was added to the U.S. Na-

tional Register of Historic Places in December 1969. Long

before the picturesque plantation structures were built be-

ginning in the early 1800 s, the land was part of a trade

route used by Native American groups, and the forests

were used as primary hunting grounds for wild game. In

1780, Colonel John Donelson led a party of explorers and

settlers by river to central Tennessee and eventually

founded the city of Nashborough (later renamed Nash-

ville), Tennessee. Daniel Dunham (b. 1735, d. 1787) was a

member of this party. In a meander bend of Richland

Creek, south of today’s Nashville, Dunham settled the area

and moved his family to a 640 acre tract of land in 1784

(North Carolina Land Grant #173, 15 January 1784, issued

to Daniel Dunham, Grant Book 66, p. 191). That year, he

built a small fort or station (i.e., a post in a remote location)

at a main crossing to accommodate the high traffic of

travellers through the area at that time. The fort came to be

known as Dunham’s Station. While detailed written

records are scarce for the Dunham family and the region in

general during this period, several documents reveal

hardship for the Dunham family throughout their tenure on

the property. During a raid on the property in May 1787 by

Native Americans, Daniel Dunham and two of his children,

Joseph and William, were killed (Kilgore 1981; Wills

1991; Hoobler 2008). The station was later partially de-

stroyed in a fire in 1792 during another raid. Despite these

hardships, the remaining Dunham family members lived at

the station until the property was purchased by the Harding

family (Kilgore 1981; Hoobler 2008).

John Harding (b. 1777, d. 1865) moved with his family

to the area in 1798 and purchased the property from Daniel

A. Dunham (b. ca. 1784, d. ca. 1840, the son of Daniel

Dunham) on 21 February 1807. Scholars believe that the

original station structure and Dunham home currently oc-

cupies the eastern half of the current two-pen structure

(Kilgore 1981; Wills 1991) (Fig. 1). Henceforth, we call

this structure ‘‘Cabin 1.’’ Sometime later, the family of

John Harding joined a second log cabin (which we call

‘‘Cabin 2’’) on the western side of Cabin 1 (Wills 1991). By

1820, however, John Harding had built a large house on the

former ‘‘Richland Creek Plantation,’’ renaming the

mansion and the property ‘‘Belle Meade Plantation’’

(Fig. 2). He continued to purchase adjoining land such that

the plantation totalled 3,800 acres (1,530 ha) by the mid-

1820 s. Harding had income from the standing of stallions

and boarding of horses, and already owned several horses

he trained and raced. When his son William (b. 1808, d.

1886) took over management of the plantation in 1839, he

continued the horse breeding and racing tradition his father

began, making Belle Meade one of the most renowned

breeders of thoroughbred horses in the U.S., a legacy that

lasted through the rest of the 19th century. After his death

in 1886, national financial crises, a withering economy, and

evangelical reform that disfavoured gambling slowly took

their toll on horse racing and breeding, and final dispersal

sales of the plantation’s horses took place in 1903 and

1904.

Fig. 1 View of the south walls of Cabin 1 (right) and Cabin 2 (left).

The dogtrot or breezeway is seen in the centre. Photo orientation is to

the north (photograph by H. D. Grissino-Mayer)

Fig. 2 The Tennessee Valley region with locations of the Belle

Meade Plantation and the four reference chronologies used in this

project (map created by L. A. Stachowiak)

106 Veget Hist Archaeobot (2016) 25:105–115

123

For over 200 years since the land was purchased by

John Harding, the two conjoined cabins remained on the

plantation, used as housing for slaves and later by freedmen

and other tenants during the 19th and 20th centuries. The

cabins are located approximately 200 metres from what is

now the main plantation house, and 100 metres from the

meander in Richland Creek. The cabins are separated by a

central breezeway (also called a dogtrot) in a style referred

to as a double pen cabin (Rehder 2012). The cabins feature

a second floor, or half story, accessible via steps in the

breezeway between the two cabins (seen in Fig. 1). Each

cabin has a limestone chimney located on either side of the

structure. A porch exists on both the north and south sides

of the cabins, and entrance doors to each pen are currently

located only through the breezeway. However, 19th cen-

tury photographs and evidence of cutting and replacement

of logs in each pen indicate that original doorways once

existed only on the south-facing walls.

Despite claims that Dunham was the builder of Cabin 1,

‘‘the true age of the cabin, alleged to be the original or

described as built on the site of Dunham’s Station, remains

conjecture’’ (Warden 1988). Because doubt exists which of

the two could have been built by Daniel Dunham, doubt

also exists about which cabin was built by John Harding.

The overarching goal of our project was to date precisely

the years of construction for both cabins and thereby help

evaluate the likely builder of each structure.

Materials and methods

Field methods

We extracted 12 mm diameter cores from both log struc-

tures using a specially designed, electrically powered hol-

low drill bit. Cores were specifically located and collected

from logs based on several factors. First, we collected cores

from the basal end of the log to ensure as many rings as

possible were included. Second, we collected cores only

from the underside of logs to ensure the holes left were

inconspicuous and to allow the sawdust to quickly evacuate

the drill bit. Third, we cored only along the rounded edges

of the log (and through bark, when possible) to capture the

outermost rings and thus determine the year(s) the trees

were harvested. The angle of the drill bit was carefully

monitored while coring by a second person who stood at

the log end and ensured alignment of the drill bit tip with

the pith. This allowed us to reach as close to the pith as

possible and therefore obtain the maximum number of

rings possible. Once extracted, the cores were immediately

mounted on core mounts with the cells vertically aligned.

All cores were then labelled according to cabin, compass

direction of wall, the log number, and core letter (if

needed), e.g., C1N005A = Cabin 1, north wall, log 5 (se-

quential from the bottom log), core ‘‘A.’’ Lastly, locations

from where all cores on all logs were extracted were

sketched for permanent archiving.

Sanding and measuring

Each core was sanded with a tabletop sander using progres-

sively finer sandpaper beginning with ANSI 80-grit

(177–210 lm) and endingwithANSI 400-grit (20.6–23.6 lm)

(Orvis and Grissino-Mayer 2002) to polish the wood to a fine

finish. Once sanded, tree rings on all cores were marked using

standard decadal dot notation (Stokes and Smiley 1996; Speer

2010) and then measured to 0.001 mm accuracy using a Vel-

mex measuring system coupled with MeasureJ2X software.

Measuring began with the innermost complete ring assigned to

relative year ‘‘1.’’ By convention, the outermost ring is not

measured because it may not be a complete ring if the tree was

harvested during the middle of the growing season.

Internal crossmatching

We used COFECHA to perform segmented time-series

correlation analysis to suggest a possible temporal placement

for all series in the data set (Holmes 1983; Grissino-Mayer

2001). We evaluated 40-year ring segments lagged by

10 years to provide more in-depth diagnostics that helped

ensure the proper temporal alignment had been achieved.

The program statistically compared each individual series

with a master chronology created from all other cores and

calculated the inter-series correlation coefficient, a statistical

metric of how well the ring-width patterns from the core

being tested matched the chronology created from all other

cores. A minimum value of 0.40 (usually p\ 0.001) was

desirable for results to be statistically convincing although

individual correlation values were often much higher (e.g.,

r = 0.55 with p\ 0.0001). A core was considered cross-

matchedwith all other cores when its coefficient was C 0.40

and COFECHA suggested a temporal adjustment that was

identical for all or most of the segments tested (e.g., ‘‘?41’’).

Suggested placements made by COFECHA were confirmed

visually by comparing line graphs of each series being tested

against line graphs of other series. The result of the internal

crossdating exercise was a set of tree-ring measurement

series that were properly aligned in time relative to each

other, but not absolutely dated.

External crossdating

Absolute (external) crossdating is possible because the

overarching regional climate signal imparts a common pat-

tern of wide and narrow rings in tree growth. Our reference

datasets consisted of three Quercus alba L. (white oak)

Veget Hist Archaeobot (2016) 25:105–115 107

123

chronologies and one Quercus stellataWangenh. (post oak)

chronology downloaded from the International Tree-Ring

Data Bank (ITRDB) (Grissino-Mayer and Fritts 1997), all

located in the surrounding region (Fig. 2). The reference

chronologies were created by Edward R. Cook of the La-

mont-Doherty Earth Observatory at Columbia University in

New York (KY004), Daniel N. Duvick (formerly) of Oak

Ridge National Laboratory in Tennessee (TN005 and

TN008), and Michael C. Stambaugh and Richard P. Guyette

of the University of Missouri (KY005):

1. Mammoth Cave Recollect, KY004, 37.1215�N,86.0504�W, 1649–1985, Q. alba;

2. Piney Creek Pocket Wilderness, TN005, 35.6484�N,84.8584�W 1651–1982, Q. alba;

3. Norris Dam State Park, TN008, 36.1734�N,84.1003�W, 1633–1980, Q. alba;

4. Land between the Lakes, KY005, 36.7473�N,88.0664�W, 1689–2005, Q. stellata.

The reference chronologies were selected because of

their close geographic proximity to the Belle Meade

Plantation, and because they exhibited the appropriate

depth back in time to ensure a successful temporal overlap

with the pattern of ring widths from the undated cores.

These four chronologies were combined to create a single

regional reference chronology using the computer program

ARSTAN (Cook 1985).

We used COFECHA to statistically crossdate the undated

series by entering the reference Quercus chronology as the

dated dataset and the file containing the temporally adjusted

measurement series as the undated dataset. We again tested

40-year segments in the undated series lagged by 10 years to

maximize the number of segments tested and to develop

useful diagnostics for evaluating problematic segments or

cores. Crossdating was achieved when COFECHA suggested

a common temporal adjustment for all or most tested seg-

ments (e.g., ‘‘?1582’’) and these segments also displayed

statistically significant (usually p\0.001) correlations

against the reference chronology. Once crossdated, we used

program EDRM (Edit Ring Measurement) (Holmes 1992) to

manually adjust the rings to their exact calendar years and

created a chronology for each cabin using program ARSTAN.

We graphically verified the crossdating by overlaying the two

dated cabin chronologies along with the reference chronol-

ogy. Crossdating had to be convincing both graphically and

statistically (Grissino-Mayer 2001).

Determining cutting dates

Cutting dates for logs from a cabin indicate the years in

which the trees were harvested and are determined by

careful examination of the terminal rings on the cores un-

der high magnification. Only a few logs will have bark

attached to the outermost ring, making the year of the last

formed tree ring easily known. Most logs, however, do not

have intact bark and others may have the outermost rings

lost to decay or trimming during construction. Therefore, a

system has been established to determine the cutting dates

of each log and the likely year(s) of construction for the

cabin (Bannister et al. 1966; Nash 1999; Grissino-Mayer

et al. 2009).

B: Bark was present and attached to the last formed ring,

indicating the outermost ring is intact (a firm cutting date);

r: The outermost ring is intact around a smooth curved

surface of the log (considered a cutting date);

v: The date is within a few years of the cutting date,

based on the presence of sapwood (a near cutting date);

vv: A cutting date is not possible because we could not

determine how far we were from the outermost ring ever

formed on this tree (a non-cutting date);

??: A ring count was necessary on the outermost rings

because these were located in a detached sapwood portion of

the core, or the rings couldnot be crossdatedpast a certainpoint.

Once cutting dates for each sample were determined, we

noted any clustering of dates around certain calendar years

that could indicate the likely year(s) of tree harvesting for

cabin construction.

Results

Species used

Once we were able to view the sanded cores under the

microscope, we learned that the cabins were constructed

from multiple tree species (Table 1). Logs from Cabin 1

were cut from one of several Fraxinus (ash) species and

from one of several Ulmus (elm) species that grow in the

area (Fig. 3). Cabin 2 was made from four different tree

genera: Fraxinus, Ulmus, Quercus (oak), and Carya

(hickory) (Table 1). Each cabin was analysed by grouping

series from each genus together and testing the crossdating

of the floating series from the individual genus internally

against each other and externally against the reference

chronologies. Only Fraxinus and Ulmus were analysed

because we did not have enough cores that had enough

rings to test the Quercus and Carya series. We also com-

pared the final Fraxinus and Ulmus chronologies together

to evaluate crossdating between two hardwood tree genera,

and to test the crossdating of series between the two cabins.

Internal crossdating for Cabin 1

We were able to internally crossdate the tree rings from

seven cores representing all seven logs cut from Fraxinus

trees (ESMTable 1). The average interseries correlationwas

108 Veget Hist Archaeobot (2016) 25:105–115

123

0.53, which would be considered high for southeastern tree

species, while the average mean sensitivity was 0.23, a value

typical for trees with a moderate climate signal. COFECHA

flagged 15 of the 59 40-year segments tested as being

problematic (25 %), which is considerably higher than the

cut-off of 10 % we desire when crossdating tree rings in the

southeastern U.S. Of these 15 problem segments, eight oc-

curred on one core, C1N006A, which was the longest of the

seven Fraxinus cores in Cabin 1 (149 years) and also the

most challenging to crossdate. These eight segments had

correlations that were nonetheless all positive, indicating

some inherent climate signal, andwere bracketed by younger

and older segments that were not flagged (ESM Table 1).

Furthermore, this series, despite the problem segments, was

crossdated as having the same outermost ring (ring 149) as

five of the other six series, further confirming this series was

crossdated correctly.

The Ulmus cores, overall, contained tree rings that were

problematic to crossdate. Of the 11 logs in Cabin 1 that we

cored that were cut from Ulmus trees, we could crossdate

the tree-ring series from seven logs (ESM Table 2). We

found that Ulmus forms well-defined rings only when

young or middle-aged (Fig. 3a, d), but when some indi-

viduals reach maturity, the ring boundaries are not well-

defined, blending in with the distinct ‘‘ulmiform’’ bands in

the latewood to the point they cannot be distinguished. For

the seven that could be crossdated, Ulmus displayed a very

high average interseries correlation of 0.60 and an equally

impressive average mean sensitivity of 0.34 (ESM

Table 2). Equally important were the overall lengths of

individual series which averaged 145 years, with the

longest/oldest (C1W005A) being 210 years. Three series

had outermost rings with the same relative ring number

(210), suggesting these logs were cut in the same year; two

other logs had outermost ring numbers of 205. A final

notable feature of the crossdated Ulmus cores was the very

few (four) of the 84 40-year segments (5 %) flagged by

COFECHA. In general, when Ulmus can be crossdated, the

tree-ring series are exceptional.

Internal crossdating for Cabin 2

We were able to crossdate the tree rings from eight cores

representing all seven logs that were cut from Fraxinus

trees used to construct Cabin 2. The average interseries

correlation was 0.50 and the average mean sensitivity was

0.25 (ESM Table 3), both values comparable to Fraxinus

in Cabin 1. Of 69 40-year segments tested by COFECHA,

only four were flagged as being problematic (6 %). The

outermost ring number (102) was similar for six of the

seven logs, while the remaining log had a relative ring

number of 100. This finding indicates that these Fraxinus

trees were cut during the same year.

As with Ulmus in Cabin 1, we encountered considerable

difficulty in crossdating the Ulmus tree-ring series for

Cabin 2. We were only able to partially crossdate the tree

rings from eight cores representing five logs (ESM

Table 4). The tree-ring series could be crossdated only

during the earliest to middle sections on each core, such

that up to half the tree rings on any one core could not be

crossdated with confidence. We crossdated the Ulmus tree

rings as far as possible from innermost to outermost ring,

and then counted the rings past that point. At that point, the

thin ring boundaries blended into the ‘‘ulmiform’’ bands in

the latewood and could no longer be distinguished. We

report statistics only for the crossdated portions of the

Fig. 3 a Core taken from an Ulmus log (top) and a Fraxinus log

(bottom). b ‘‘Ulmiform’’ bands in the latewood that were diagnostic

of Ulmus cores. c Attached bark on a Fraxinus core. d Pith of a

Fraxinus core (photographs by L. A. Stachowiak)

Table 1 Tree genera used to build both cabins

Veget Hist Archaeobot (2016) 25:105–115 109

123

cores. The average interseries correlation was 0.48 and the

average mean sensitivity was 0.31, both values somewhat

lower than the values reported for Ulmus from Cabin 1.

Nine segments of the 69 40-year segments (13 %) that

could be crossdated were flagged by COFECHA as being

problematic but all segments were carefully re-inspected

and found to be correctly dated. The crossdated portions of

the cores resulted in a 169 year chronology, but ring counts

brought these cores out enough to create a 222-year long

dataset. Several cores had outermost rings that were within

a few years (e.g., ring numbers 218, 219, 220 and 222)

indicating that the ring counts for Ulmus could provide

some supporting information on possible cutting dates for

Cabin 2.

External (absolute) crossdating

We tested the correlations between the Fraxinus series

from both cabins and found excellent correspondence,

suggesting we could combine the two datasets into one

larger dataset. We created a floating chronology from this

dataset that spanned 149 years and tested this against the

composite reference chronology created from the four re-

gional Quercus chronologies using COFECHA. We found

the absolute crossdating when COFECHA suggested a

‘‘?1656’’ dating adjustment for all 12 40-year segments

tested (lagged 10 years) of the 149-year long dataset with

an average correlation of 0.43. We then added 1656 to all

rings in the Fraxinus measurement series and created an-

other chronology which we again tested against the master

reference chronology. We found a statistically significant

correlation (r = 0.39, n = 149, t = 5.14, p\ 0.0001) that

suggested that the Fraxinus series were now absolutely

dated, spanning from 1657 to 1805 (Fig. 4). The last ring

present on most cores was therefore the year 1806 which

could have been a partial or complete ring.

We tested the correlations between theUlmus series from

both cabins and again found excellent temporal correspon-

dence, suggesting we could combine the two datasets into

one larger dataset. We created a floating chronology from

this dataset that spanned 223 years and tested this against the

composite reference chronology created from the four re-

gional Quercus chronologies. COFECHA found a strong

dating adjustment of ‘‘?1582’’ for 12 of the 19 40-year

segments tested (lagged 10 years) in the 223-year long

dataset with an average correlation of 0.44. These correla-

tions and associated dating adjustments were present in the

early andmiddle portions of the floating chronology because

the outer portion of this dataset contained tree rings from

Cabin 2 that could not be crossdated. We added 1582 to the

calendar dates for all tree rings and created another chron-

ology to test against the master reference chronology. We

again found a statistically significant correlation (r = 0.36,

n = 173, t = 5.05, p\ 0.0001) that suggested the Ulmus

tree-ring data were now absolutely crossdated, spanning

1583 to 1805 (Fig. 5) (common period of overlap is 1633 to

1805 = 173 years). For the Ulmus cores collected from

Cabin 1, the date for the outermost unmeasured rings was

therefore 1806. No cutting dates could be obtained forUlmus

from Cabin 2 because they could not be crossdated com-

pletely. However, the outermost dates based on ring counts

(e.g., 1800 for C2E005, 1801 for C2W003, and 1804 FOR

C2W006) provide a terminus pro quem and support the

structure being built a few years after 1804.

Having established that both Fraxinus and Ulmus cross-

date with Quercus tree-ring data, we next tested the corre-

lation between the Fraxinus andUlmus tree-ring series from

the two cabins. We found a statistically significant correla-

tion (r = 0.35, n = 149, t = 4.53, p\ 0.0001) that suggests

the two species do in fact crossdate against each other, but the

relationship is weaker than both relationships with the

Quercus tree-ring data.

Cutting dates

Most of the Fraxinus cores from both cabins clearly show

that the outermost measured tree ring dates to the year

1805. Careful inspection, however, revealed that most

Fraxinus cores have what could be a complete ring for

1806 (by convention, the outermost ring is never measured

because it could represent an incomplete ring), indicating

that the latest harvest of trees used to build both cabins

occurred sometime in the dormant season of 1806–1807

(Table 2). For most Fraxinus cores, the likely season for

harvest is late winter or early spring of 1807 because John

Harding purchased the property in late February of 1807

and likely began cutting trees immediately or soon after for

use in Cabin 1. Four Fraxinus cores have clear bark re-

tained, and these four cores provide convincing evidence

for tree harvest during the dormant season of 1806–1807.

Some Fraxinus cores suggest tree harvesting before the

dormant season of 1806–1807, but it is more likely that the

outermost ring or a partial outermost ring was trimmed

during construction or decayed off, despite these logs

having what appeared to be smooth surfaces.

The Ulmus cores provide supporting evidence for har-

vest during the dormant season of 1806-1807 although

most Ulmus cores have outermost ring dates that slightly

precede 1807 (Table 2). One Ulmus core (C1W005) had

bark but the outermost two rings were indistinct and we

could not tell conclusively if any wood had formed for

1806 yet. Based on these results, we can confirm that both

structures were built from Fraxinus and Ulmus trees that

were likely harvested in late winter or early spring of 1807,

before trees had come out of their winter dormant period.

110 Veget Hist Archaeobot (2016) 25:105–115

123

Discussion

The use of multiple species is uncommon for building log

structures in the southeastern U.S. as early settlers built

structures usually from a single tree species. Morgan

(1990) noted that 20 % of the log structures he studied in

east Tennessee were built from two species, usually pine

and tulip poplar. However, of nearly 50 historic structures

we have analysed over the last 15 years, the Harding

Cabins are the first we have encountered with two or more

species. The choice to use these species was likely based

on local availability and suggests Fraxinus and Ulmus were

locally abundant. Neither Morgan (1990) nor Rehder

(2012) mention the use of these two species to build log

structures, nor could we find any reference to using these

species as building materials in log structures in the

southeastern U.S.

Our study also firmly establishes for the first time that

three ring-porous hardwood genera commonly found in the

southeastern U.S. (Quercus, Fraxinus, and Ulmus)

demonstrate crossdating between tree genera. The corre-

lations we found in our study, however, were not as high as

those found in previous studies when comparing

chronologies created from the same species, but the cor-

relations were still statistically significant with a high de-

gree of confidence (p\ 0.0001). The composite Quercus

chronology created from the four regional chronologies did

an excellent and convincing job in crossdating the floating

Fraxinus and Ulmus tree-ring chronologies created from

both structures. Using a regional composite was desirable

because this chronology captures the overarching climate

of the region, which imparts the unique year-to-year pat-

terns found in tree rings that allow crossdating to be per-

formed. We observed, however, that using individual

Quercus chronologies was also suitable for crossdating. In

fact, the strongest correlations between the floating Frax-

inus and Ulmus tree-ring chronologies and the individual

Quercus datasets were found using only the Land between

the Lakes (KY005) chronology. For example, the Fraxinus

chronology from both structures had a correlation of 0.45

Fig. 4 Graphical

correspondence between the

Fraxinus chronology created

from tree-ring series in both

cabins (top) and the composite

reference chronology created

from all four Quercus

chronologies (bottom)

(r = 0.39, n = 149, t = 5.14,

p\ 0.0001)

Fig. 5 Graphical

correspondence between the

Ulmus chronology created from

tree-ring series in both cabins

(top) and the composite

reference chronology created

from all four Quercus

chronologies (bottom)

(r = 0.36, n = 173, t = 5.05,

p\ 0.0001)

Veget Hist Archaeobot (2016) 25:105–115 111

123

(n = 149, t = 6.11, p\ 0.0001) with the KY005 Quercus

chronology, but had a correlation of 0.39 against the re-

gional Quercus composite chronology.

Previous studies suggested that the western cabin (Cabin

2) was the one built by John Harding, while the eastern

cabin (Cabin 1) could be the original station cabin built by

Daniel Dunham in the 1780 s (or rebuilt after burning in

the 1790s). Our study shows that both cabins were con-

structed in the same year in the same season (spring of

1807) which corroborates construction by the Harding

family as John Harding purchased the land from the

Dunham family on 21 February 1807. Kilgore (1981) first

suggested that the eastern cabin was likely built by the

Dunham family while the western cabin was likely built by

the Harding family. Her conclusions were based on an

architectural analysis that relied heavily on the notch types

of the two separate pens and the tenuous geographic rela-

tionships that associated certain notch types with certain

geographic regions (Kniffen and Glassie 1966). Cabin 1 is

characterized by half-dovetail notching (Fig. 6a) and

Table 2 Inner ring and outer ring dates and types from 14 Fraxinus and 12 Ulmus logs sampled from Cabins 1 and 2

Log ID Inner ring

date

Outer

ring date

Outer ring

typeaSpecies Cutting dates

C1E002 1741 1806 B Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C1E003 1664 1803 r ?? Ulmus Near cutting date, rings counted after 1742, break at 1774, add 29 rings after break

C1E004 1677 1805 r Ulmus Outer ring appears complete, tree cut some time after fall 1805

C1E005 1636 1802 v Ulmus Near cutting date, sapwood present, outer rings eroded/missing

C1N004 1688 1806 r Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C1N006 1657 1806 r Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C1N007 1701 1806 r Fraxinus Outer ring only has earlywood, no latewood, tree cut after summer 1806

C1W004 1643 1806 r Ulmus Outer ring is partial, no obvious latewood, tree cut some time after summer 1806

C1W005 1596 1805 B Ulmus Outer ring appears complete, tree cut some time after fall 1805

C1W007 1629 1803 v Ulmus Near cutting date, sapwood present, outer rings eroded/missing

C1S004 1696 1806 r Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C1S005 1686 1801 r ?? Ulmus Near cutting date, break at 1786, added 15 rings after break

C1S007 1673 1806 r Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C1S008 1699 1806 B Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C2E004 1731 1806 B Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C2E005 1667 1800 r ?? Ulmus Crossdated out to ca. 1750, rings counted afterwards, cutting date unknown

C2E006 1626 1789 r ?? Ulmus Crossdated out to ca. 1730, rings counted afterwards, cutting date unknown

C2E007 1716 1806 r Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C2N003 1723 1804 v Fraxinus Outer surface cuts diagonally across outer rings, near cutting date only

C2N004 1739 1806 r Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C2N005 1704 1806 B Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C2W003 1615 1801 r ?? Ulmus Crossdated out to ca. 1750, rings counted afterwards, cutting date unknown

C2W004 1743 1806 r Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C2W006 1591 1804 r ?? Ulmus Crossdated out to ca. 1750, rings counted afterwards, cutting date unknown

C2S004 1711 1806 B Fraxinus Outer ring appears complete, tree cut any time from fall 1806 to spring 1807

C2S006 1583 1777 r ?? Ulmus Crossdated out to ca. 1740, rings counted afterwards, cutting date unknown

a See text for explanation of outer ring types

Fig. 6 a Half dovetail notching found on Cabin 1. b Diamond

notching found on Cabin 2. Note the precision fitting of logs in the

half dovetail notching compared to the loose workmanship of the

diamond notching (photographs by H. D. Grissino-Mayer)

112 Veget Hist Archaeobot (2016) 25:105–115

123

Kilgore (1981) noted that the half-dovetail notching was

common in North Carolina and eastern Tennessee, the re-

gion from where Daniel Dunham and his family emigrated.

In contrast, Cabin 2 is characterized by the very uncommon

diamond notching where both upper and lower sides of the

end section of the log are chamfered to produce the crown

(Gavin 1997) (Fig. 6b). This notch type can be traced to

settlers who emigrated from near the Virginia/North

Carolina border (Gavin 1997). Although John Harding

originated from Goochland County in central Virginia,

Wills (1991) noted that Harding likely learned the diamond

notching technique ‘‘in Virginia, where that method was

prevalent’’.

A major historical inconsistency exists, however: if both

cabins were constructed by the Harding family contem-

poraneously, why then is one cabin (Cabin 1) well-pre-

served while the other (Cabin 2) is in a poor state of

preservation? And why use two different notch types? The

two cabins stand in stark visual contrast to each other.

Cabin 1 is characterized by logs that are in an excellent

state of preservation. Furthermore, Cabin 1 demonstrates

unusual precision with the log notches fitting perfectly into

each other, top and bottom (Fig. 6a), suggesting a master

builder was responsible for building the cabin. The wall

construction follows a basal- to distal-end overlaying

technique, and each log was clearly chosen specifically for

each successive layer. The logs on all four walls were

meticulously hewn, producing straight sides of unusual

precision. Cabin 2 is characterized by logs that are in a

poor state of preservation, with logs fitting loosely to-

gether. The uncommon diamond notching does not require

precision fitting and could be created by someone unskilled

in cabin building or notching. Many wide gaps exist in the

notching between the logs, and the sequential size order of

each ascending log is not well organized. Any person

carefully inspecting the logs in both structures would

clearly conclude that both structures were built during

different periods.

To help resolve this puzzle, we first evaluated whether

the Fraxinus and Ulmus trees used to build both structures

could have come from the same Fraxinus and Ulmus

populations. We observed that the Fraxinus trees used to

construct Cabin 1 were considerably older (maximum

150 years, average 112 years) compared to the Fraxinus

trees used to construct Cabin 2 (maximum 102 years, av-

erage 79 years). Furthermore, the Fraxinus trees used to

build Cabin 1 were clearly larger in diameter (aver-

age = 31 cm) than the Fraxinus trees used to build Cabin

2 (average = 21 cm), based on extrapolations of the pith to

outer edge distance of all the Fraxinus cores we sampled.

The Fraxinus trees also crossdated well against each other.

The Ulmus trees cut to build both structures were generally

quite old with little difference in maximum ages (Cabin

1 = maximum 210 years, while Cabin 2 = maximum

222 years), but the Ulmus trees used in Cabin 1 were

younger (average = 145 years) compared to the Ulmus

trees in Cabin 2 (average = 168 years). However, Ulmus

specimens used to build Cabin 1 were much larger, with an

average diameter of 28.4 cm compared to the smaller di-

ameter Ulmus used to build Cabin 2 (average = 20.4 cm).

Like the Quercus, the Ulmus trees from both structures

crossdated well with each other. This also might help ex-

plain why the Ulmus tree rings in Cabin 2 were more

difficult to crossdate: the Ulmus trees cut to build Cabin 2

were both older but narrower in diameter, meaning more

rings per unit measurement. This could, however, be an

artifact of the location we chose to sample on the indi-

vidual Ulmus logs.

These findings suggest that the Fraxinus and Ulmus

trees used in Cabin 1 came from the same Fraxinus and

Ulmus tree populations used to build Cabin 2, with only

one major difference. The trees used in Cabin 1 were larger

than the trees used to build Cabin 2, a finding easily con-

firmed by visual inspection of the two structures them-

selves. We suggest that the larger trees in the original, local

Fraxinus and Ulmus stands were selectively cut first in

early 1807 and used to build Cabin 1, while the remaining

smaller diameter Fraxinus and Ulmus trees were then

harvested, also in early 1807, to build Cabin 2. If larger

diameter trees had been available, the builder of Cabin 2

would surely have used these rather than the smaller di-

ameter trees because using larger diameter trees would

mean fewer trees needed to be cut, hewn, and notched. An

alternative explanation suggests that the builder of Cabin 2

was specifically instructed to, or preferred to harvest, only

the smaller diameter trees.

Based on this mounting visible, physical, and den-

drochronological evidence, we propose a radical interpre-

tation of the construction of both cabins: the cabins were

not built by the same person or by same family. If John

Harding and his family had purposely built two cabins at

the same time in spring 1807, both cabins would likely

show similar structural features, such as similar log sizes,

only one notch type used, clear arrangement of the logs

(distal and basal log ends alternating on each side of the

cabin), and precision fitting of the logs one on top of the

other. Instead, we propose two very different groups of

people built the two cabins. We propose that Cabin 1 was

built by the John Harding family using large diameter trees

that had been harvested first from the local stands of trees,

before the trees broke dormancy in the late winter/early

spring of 1807. The cabin was meticulously built with great

attention to detail. Cabin 2 was built by another group of

people, and we suggest that the cabin was likely built by

the slaves owned by John Harding, Ben and Dicey, ‘‘who

helped [John Harding] turn his woods and fields into a

Veget Hist Archaeobot (2016) 25:105–115 113

123

working farm’’ (Wills 1991). The trees were cut after the

Harding family had harvested the larger ones to use in their

cabin, and the cabin was built by individuals who likely

were less skilled in the craft of cabin construction. These

findings would repudiate the suggestion that the second

cabin was built by John Harding sometime around 1812 as

his family grew to include three children: Amanda (b.

October 1807), William (b. September 1808), and El-

izabeth (b. April 1812) (Wills 1991).

Furthermore, we propose the two cabins were built at

different locations based primarily on the major differences

in physical appearances of the two cabins. If both cabins

had been built at the same time at the same location by the

same builders, the two cabins would likely have looked

very similar. We propose that Cabin 2 likely was built

elsewhere on the property in 1807 and perhaps served as

living quarters for the Harding family slaves. In 1819, John

Harding began making significant purchases of building

materials and internal furnishings and the main mansion of

the plantation was completed by late 1820 (Wills 1991).

We suggest that, once the Harding family relocated to the

larger mansion, Harding then gave the original cabin to a

slave family that lived and worked on the plantation. To

centralize the living arrangement, and to accommodate the

growth of extended families, the original cabin built by the

slaves was relocated to the original location of the Harding

Cabin. This suggests that the breezeway and added front

and back porches were constructed by the slave families.

The two cabins continued to serve as living quarters for the

families even after emancipation. Photographs of the two

cabins dated ca. 1880 and ca. 1890 clearly show many

individuals of an African-American extended family or

families (Fig. 7).

Conclusions

The two cabins that make up the Harding Cabin provided

a unique opportunity for the advancement of dendroar-

chaeology in the Southeast. The two cabins that collec-

tively make up the Harding Cabin on the Belle Meade

Plantation were built primarily from two tree genera,

Fraxinus and Ulmus. These are the first structures known

to be made from these two genera in the southeastern

U.S., and each cabin represents one of the few historic

structures known to be built from two or more tree spe-

cies. The individual Fraxinus and Ulmus tree-ring series

from each structure crossdated well within each species.

The Fraxinus and Ulmus tree-ring series also crossdated

well against the regional composite Quercus chronology

which establishes that, for the first time, crossdating exists

between hardwood genera in the southeastern U.S. These

two cabin chronologies represent the first developed from

these two species in the southeastern U.S. Once abso-

lutely crossdated, the Fraxinus chronology from both

cabins was anchored in time from 1657 to 1805 while the

Ulmus chronology was anchored in time from 1583 to

1805. The Ulmus chronology, however, is accurate only

in the early to middle portions of the chronology, because

the rings from many Ulmus cores could not be crossdated

past a certain year. Nonetheless, the Ulmus chronology is

now the oldest in Tennessee and one of the oldest overall

in the southeastern U.S. Because the Ulmus chronology

crossdated with Quercus and Fraxinus, future studies can

use Ulmus tree-ring data to date historic structures should

no nearby Quercus chronologies exist (e.g., in western

Tennessee). The high level of crossdating between tree

rings from both cabins suggests that the Fraxinus and

Ulmus trees used to build both cabins came from the

same local forest stand.

Careful inspection of the outermost rings suggests that

the Fraxinus and Ulmus trees were cut in the dormant

season of 1806/1807, most likely in late winter/early spring

of 1807 because John Harding purchased the land on 21

February 1807. This refutes the assertion that Cabin 1

possibly represented the original cabin built by Daniel

Dunham or built later after his death. Differences in visual

appearance, level of workmanship, preservation quality,

notch types, tree size, and tree ages suggest that the two

cabins likely were built by two different families. We

propose that John Harding and his family built Cabin 1

while Cabin 2 was likely built by the slaves that lived on

the property. We further propose that Cabin 2 was perhaps

built elsewhere on the property, then relocated and con-

solidated with Cabin 1 once the main mansion was com-

pleted in 1820, after which the two cabins served as living

quarters for slaves and freedmen after emancipation.

Fig. 7 Photograph dated ca. 1890 showing the joined cabins which

housed an African-American family (photograph courtesy of the Belle

Meade Plantation)

114 Veget Hist Archaeobot (2016) 25:105–115

123

Acknowledgments We thank personnel at Belle Meade Plantation

who first expressed interest in our carrying out this interesting project,

especially Board Member Paul Clements, Executive Director Alton

Kelley, and Curator John Lamb. Historic pictures, photographs, and

documents were provided by the Belle Meade Plantation for better

understanding the historical context of both cabins.

References

Bannister B, Gell EA, Hannah JW (1966) Tree-ring dates from

Arizona N-Q: Verde-Show Low-St. Johns Area. Laboratory of

Tree-Ring Research, University of Arizona, Tucson

Cook ER (1985) A Time Series Analysis Approach to Tree-Ring

Standardization. Dissertation, University of Arizona, Tucson

DeWeese GG, Bishop WJ, Grissino-Mayer HD, Parrish BK, Edwards

SM (2012) Dendroarchaeological dating of the Chief John Ross

House, Rossville, Georgia. Southeast Arch 31:221–230

Garland NA, Grissino-Mayer HD, Deagan K, Harley GL, Waters G

(2012) Dendrochronological dating of wood from the Fountain

of Youth Park Archaeological Site (8SJ31), St. Augustine,

Florida, U.S.A. Tree-Ring Res 68:69–78. doi:10.3959/2010-11.1

Gavin M (1997) The diamond notch in Middle Tennessee. Mater Cult

29:13–23

Grissino-Mayer HD (2001) Evaluating crossdating accuracy: a

manual and tutorial for the computer program COFECHA.

Tree-Ring Res 57:205–221

Grissino-Mayer HD (2009) An introduction to dendroarcheology in

the southeastern United States. Tree-Ring Res 65:4–10. doi:10.

3959/2008-19.1

Grissino-Mayer HD, Fritts HC (1997) The International Tree-Ring

Data Bank: an enhanced global database serving the global

scientific community. Holocene 7:235–238. doi:10.1177/

095968369700700212

Grissino-Mayer HD, van de Gevel SL (2007) Tell-tale trees: historical

dendroarcheology of log structures at Rocky Mount, Piney Flats,

Tennessee. Hist Archaeol 41:32–49

Grissino-Mayer HD, LaForest LB, van de Gevel SL (2009)

Construction history of the Rocky Mount Historic Site

(40SL386), Piney Flats, Tennessee from tree-ring and documen-

tary evidence. Southeast Arch 28:64–77

Grissino-Mayer HD, Kobziar LN, Harley GL, Russell KP, LaForest

LB, Oppermann JK (2010) The historical dendroarchaeology of

the Ximenez-Fatio House, St. Augustine, Florida, U.S.A. Tree-

Ring Res 66:61–73. doi:10.3959/2009-8.1

Grissino-Mayer HD, Maxwell JT, Harley GL, Garland NA, Holt DH,

Absher C, Beale BJ, Boehm MS, De Graauw KA, Rautio AM,

Dye AW (2013) Dendrochronology reveals the construction

history of an early 19th century farm settlement, southwestern

Virginia, USA. J Archaeol Sci 40:481–489. doi:10.1016/j.jas.

2012.05.038

Harley GL, Grissino-Mayer HD, LaForest LB, McCauley P (2011)

Dendrochronological dating of the Lund-Spathelf House, Ann

Arbor, Michigan, USA. Tree-Ring Res 67:117–121. doi:10.

3959/2010-2.1

Henderson JP, Grissino-Mayer HD, van de Gevel SL, Hart JL (2009)

The historical dendroarcheology of the Hoskins House, Tannen-

baum Historic Park, Greensboro, North Carolina, U.S.A. Tree-

Ring Res 65:37–45. doi:10.3959/2007-11.1

Holmes RL (1983) Computer-assisted quality control in tree-ring

dating and measurement. Tree-Ring Bull 43:69–78

Holmes RL (1992) User’s manual for program EDRM. Laboratory of

Tree-Ring Research, University of Arizona, Tucson

Hoobler JA (2008) A guide to historic Nashville Tennessee. History

Press, Charleston

Kilgore SJ (1981) Brief history of the Association for the Preservation

of Tennessee Antiquities and its Belle Meade Mansion Historic

Site Museum. Thesis, Middle Tennessee State University,

Murfreesboro

Kniffen FB, Glassie H (1966) Building in wood in the eastern United

States: a time-place perspective. Geogr Rev 56:40–66

Mann DF, Grissino-Mayer HD, Faulkner CH, Rehder JB (2009) From

blockhouse to hog house: the historical dendroarcheology of the

Swaggerty Blockhouse, Cocke County, Tennesse, U.S.A. Tree-

Ring Res 65:57–67. doi:10.3959/2007-4.1

Morgan J (1990) The log house in east Tennessee. University of

Tennessee Press, Knoxville

Nash SE (1999) Time, trees, and prehistory: tree-ring dating and the

development of North American archaeology, 1914–1950.

University of Utah Press, Salt Lake City

Orvis KH, Grissino-Mayer HD (2002) Standardizing the reporting of

abrasive papers used to surface tree-ring samples. Tree-Ring Res

58:47–50

Rehder JB (2012) Tennessee log buildings: a folk tradition. Univer-

sity of Tennessee Press, Knoxville

Slayton JD, Stevens MR, Grissino-Mayer HD, Faulkner CH (2009)

The historical dendroarcheology of two log structures at the

Marble Springs Historic Site, Knox County, Tennessee, U.S.A.

Tree-Ring Res 65:23–39. doi:10.3959/2007-5.1

Speer JH (2010) Fundamentals of tree-ring research. University of

Arizona Press, Tucson

Stachowiak LA, Schneider EA, Rochner ML, Collins SA, Swiney CP,

Grissino-Mayer HD, McKenzie TG (2014) Dendrochronological

dating of the historic McKenzie Home, Meigs County, Ten-

nessee, USA. Tree-Ring Res 70:31–39. doi:10.3959/1536-1098-

70.1.31

Stahle DW (1979) Tree-ring dating of historic buildings in Arkansas.

Tree-Ring Bull 39:1–28

Stokes MA, Smiley TL (1996) An introduction to tree-ring dating.

University of Arizona Press, Tucson

Warden ML (1988) The Belle Meade Plantation. Association for the

Preservation of Tennessee Antiquities, Nashville

Wills R II (1991) The history of Belle Meade: mansion, plantation,

and stud. Vanderbilt University Press, Nashville

Veget Hist Archaeobot (2016) 25:105–115 115

123