5%

7%

9%

11%

13%

15%

17%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

Effects of Surface Polish

Any number of post-discardal or use-relatedprocesses can induce changes to surfacetexture. Here we show how various polishinggrits can change mean reflectance values. Theintensity of spectral features can also bealtered.

0%

10%

20%

30%

40%

50%

60%

14.00 24.00 34.00 44.00 54.00 64.00 74.00

Grit Size (um)

Mean

Refl

ect

an

ce

32%

34%

36%

38%

40%

42%

44%

46%

48%

350 850 1350 1850 2350Wavelength (nm)

Refl

ect

an

ce

Within Sample Variability of Spectra

This graph shows the spectra taken at five separatelocations on a sample of Arkansas Novaculite. Eachcolor represents a series of four spectra taken at 90degree rotations at the same point. Note thesimilarity in shape is retained, but that there is amultiplicative effect ranging from 2.5% reflectance ata single point to about 6% for the whole sample. Thisbaseline shift must be corrected before analysis. Weopted to use the first derivative transformation toaccomplish this.

References

Calosero, B. 1991. Macroscopic and Petrographic Identification of theRock Types for Stone Tools in Central Connecticut. Unpublished Ph.D.dissertation, University of Connecticut.

Hunt, G. R. and Salisbury J. W. 1970. Visible and near-infrared spectra ofminerals and rocks: I silicate minerals. Modern Geology 1:283-300.

-0.0006

-0.0004

-0.0002

0

0.0002

0.0004

0.0006

350 850 1350 1850 2350

Wavelength (nm)

Fir

st D

eri

vati

ve

First Derivative Transformationof Arkansas Novaculite

This graph shows first derivative transformation forthe repetitive measurements of the ArkansasNovaculite (data from graph on left). The similarity inthe graphs show that features are retained despite thebaseline shift caused by changes in the probe-to-surface geometry.

Chert NBurlington Chert, MO. 34Conklin (Limerock) Jasper, Conklin, RI. 10Flint Ridge Flint, Licking Co., OH. 25Upper Mercer Chert, East-Central, OH. 23Knife River Flint, ND. 30Arizona Chert, AZ. 5Mexican Jasper 11Antelope Jasper, ID 4Maastrichtian Chert, Netherlands 15Arizona Chert, AZ. 5Vera Cruz Jasper, Reading Prong, PA 12Arkansas Novaculite, AK. 45N=12 N=221

The cherts used in this project wecollected by one of the authors(MJH). The sample was neverintended for such a project and doesnot represent the diversity ofsamples needed from a quarry. Insome cases the measurements weretaken on only one or two handsamples from each locality. In othercases a larger sample was available.

IntroductionLithic material is ubiquitous and durable,

sometimes comprising the only remains found atarchaeological sites. The provenance of this lithic materialholds the key to many archaeological questions.Traditional lithic sourcing methods have many drawbacksincluding extensive sample preparation time, cost anddestructiveness. These drawbacks have severely limitedsample sizes in lithic sourcing studies. This increases theprobability of underestimating the true variability of asource and is probably the most commen lament amonglithic sourcing experts.

We present here a method for sourcing lithicmaterials using diffuse reflectance spectroscopy (DSR) inthe visible and near-infrared portions of the spectrum. Toour knowledge this method has not been utilized in lithicsourcing. Although, Long et al. (2001), reported promisingresults using fourier-transformed spectroscopy in the mid-infrared.

LabSpec Pro FR 350 nm - 2500 nmA depth gauge on theprobe keeps the tip 0.5mm off the surfacewhich (JDO) found toprovide a maximumsignal to noise ratio.

A hemisphericmicroscope stagewas used toposition probe tipperpendicular tosample.

Advantages to using DSRin lithic sourcing

-Non-destructive, samples need no specialpreparation, sample is not altered.

-Cheap. Instruments are fraction of the cost ofother sourcing methods.

-Fast. The instrument takes approximately 10seconds to take 100 measurements on asample.

-Widely available in commercial labs anduniversities.

-Safe, no waste products or radioactivity areproduced or toxic substances used.

-Instrument can be used with minimal training.

-Portable, whole apparatus weighs less than 20pounds.

-Samples as small as 3mm may be measured.

-Multiple attributes are simultaneouslymeasured.

APPLICATION OF VIS/NIR SPECTRAL REFLECTANCE IN SOURCING AND RECOGNITION OF HEAT-TREATMENT IN CHERTSHUBBARD, Michael J. , Department of Anthropology, Kent State Univ, Lowry Hall, Kent, OH 44240,mhubbar1@kent.edu, WAUGH, David A., Department of Geology, Kent State Univ, Kent, OH 44242, andORTIZ, Joseph D., Geology, Kent State Univ, Lincoln and Summit Streets, Kent, OH 44242

M a t e r i a l s

What is DSR?When incident light hits a sample’s surface it

interacts differentially with each of the sample’sconstituents. Depending on the wavelength of light,specific electronic and vibrational processes of themolecules present and crystal orientations all combineto produce the recorded spectra.

DSR can be used to identify mineral spectra (e.g.Hunt and Salisbury, 1970). Source/pattern matching canbe accomplished entirely qualitatively and the exactchemical composition is unnecessary. Indeed, manyfactors which are considered confounding variables inquantitative studies may actually add information toqualitative analyses and increase materialdiscrimination.

However, there is one confounding factor whichis quite problematic in diffuse reflectance, that ofspecular reflectance. The specular (“mirror-like”)component of reflectance occurs at the air/rockinterface. Although it does contain information aboutsurface texture, it does not contain any compositionalsignatures of the specimen .

SpecularReflectance

DiffuseReflectance

Absorption

E f f e c t o f A n g l e

Here an Upper Mercer Chert blank which hadbeen ground flat with 600 grit paper was mounted on auniversal stage. Measurements were taken on thesame spot with the stage tilted sequentially in twodegree increments, first in one direction and then theother (i.e. 180° orthogonal). The nearly linear drop-offin reflectance values with increasing angle wasexpected. What appears to be the complete inversionof the whole spectrum between matched angle pairswas not. Needless to say, such inversions in thesourcing analysis could seriously distort our ability topredict matches between similar materials.

To explore the potential causes of theobserved inversions we swapped the chert blank witha BaSO4 calibration standard on the assumption thatthe chert’s crystal orientation could be the culprit. Themost obvious difference is that there is only a netdeviation of about 4-5% reflectance with the standardas compared to about 30% with the chert blank.

To illustrate how this might affect the dataanalysis a correlation matrix is provided. A typical r2

score for samples from the same source were usuallyabove 0.9. Here the production of spurious featuresand curvilinear trending due the subtle differences insurface-to-probe angle have combined to producescores well below what should be expected.

Instrument Error

The total error for each sample is a function ofinstrument noise, the effects of probe angle,and the within sample variability. Each wasmeasured separately and the total error (Et) wascalculated by the following formula:

where E i is the instrument noise, Ew is the withinsample variability and Ea is the averagevariability due to angle. The mean total error forall bandwidths was 1.85%.†

Et = Ei2 + Ew

2 + Ea2

Conclusions

Our results suggest that DSR is a useful addition to the lithic sourceanalyst’s toolbox both as a stand-alone application or in conjunction withother sourcing methods.

Potential sources of error that were identified in this study includebackscattering of underlying substances with thin or translucentsamples and unpredictable spectral effects due to surface texture andprobe angle. There is reason to believe that a few simple modifications tothe experimental set-up, sample selection and the way the data isprocessed can significantly reduce these errors. More sophisticatedpattern-matching algorithms capable of extracting and weighing“signature” features, detrending data, and removing baseline drift existand will be explored in the near future.

We have shown that heat-treatment of cherts can alter the spectralcharacteristics of chert. We believe that some materials andtemperatures may produce a reliably detectable signature, which fallsoutside the source material’s normal range of variation, allowing DSR todetect heat-treatment in archeological artifacts.

Non-Chert LithicsWhile an objective method for sourcing chert is sorely needed, there is something to

be said for a method that can easily and objectively differentiate between chert and non-chertartifacts. In a double-blind study, Calogero (1991) found that archaeologists in New Englandwere only able to visually discriminate between rhyolite and chert a surprising 40% of thetime. Rhyolites and cherts are sufficiently distinct in their reflectance characteristics that it isunlikely the two could be confused spectroscopically. Below are a few examples of spectra ofsome common lithic materials. Note their differences from the mean chert spectrum.

5%

10%

15%

20%

25%

30%

35%

400 900 1400 1900 2400

Wavelength (nm)

Refl

ect

an

ce

BelizeVera CruzVera CruzVera Cruz

Can You Spot the Imposter?

Belize VC1 VC2 VC3Belize 0VC1 0.054 0VC2 0.047 0.001 0VC3 0.049 0.049 0.046 0

DSR Can

One of these jasper flakes is from Cayo, Belize whilethe other three are from eastern Pennsylvania. Canyou detect the exotic? Lift the tab for the correctanswer. Note: lower scores in the RMSD matrix arethe better fit matches.

Thin flake of Knife River Chertwith various backgrounds

None12345>5

86.7%

9.0%

1.9%1.4%0.5%

0.5%0.5%

# of Misses

Search MatchThe final data set consisted of just over a

half million points. Several methods were exploredto compare these spectra. We present here just theresults from Pearson’s correlation and the RootMean Square Deviation (RMSD) on both thereflectance data and its first derivative. Regardlessof method we scored a comparison as a “hit” if itshighest score (r2 or RMSD) was with anothersample of the same material (excluding spectrafrom the same sample). Otherwise we counted thenumber of false positives before a legitimate matchoccurred.

ResultsThe derivatized data was consistently more

reliable and though correlation and the RMSDproduced approximately the same number of directmatches (82.8% versus 82.5%), the RMSD was muchmore likely to predict a match within the top fivescores. Throughout the analysis the number ofmisses also gave us an indication of problemsamples. More often than not samples which did nothave a proper match within the top few scorescould be justifiably culled from the final analysis onthe basis of obvious sampling flaws (e.g. spectrawas of cortex or an inclusion rather then the chert).

After removing 11 samples that should nothave been included in the first place, the final resultfor the RMSD on derivitized data was an 86.7%match rate and 99.0% within the top five scores.Closer examination of the misses that persistsuggests that visual inspection of the unknownspectrum with its closest matches will in manycases allow the researcher to rule out spuriousmatches, further increasing accuracy.

Distribution of incorrect matchesbased RMSD scores.

Heat Treatment

The following experiment wasdesigned to test whether significant spectralchanges occur as a result of heat-treatment.One sample of each material was buried in analuminum pan with about 10cm of sand andplaced in a oven. The temperature was raisedgradually at 50°C/hour for each batch until thedesired temperature was reached (200°C,400°C, or 600°C) and was then maintained atthis temperature for four hours. We used theexisting oven temperature gauge rather than amore accurate thermocouple. The sampleswere re-measured at the same spot beforeand after heating.

Heat-treatment Results

Most heat-treatment was probablyaccomplished in the temperature range of200-500°C. Only the white novaculite showedno apparent visual changes at anytemperature, but it should be noted that all thecolor changes that did occur could very easilyfall within the normal range of variation fortheir material type. The heat-damagedsamples would be the only truly reliable visualindicator of heat-treatment in the field.

The series of graphs to the rightshows representative spectra for eachmaterial both before and after heat-treatment.Spectral changes are readily apparent athigher temperatures. However, by subtractingone spectrum from another and rescaling, it isapparent even for what appear to be unalteredspectra that small, potentially diagnosticchanges, have occurred.

-0.2

-0.1

0

0.1

0.2

0.3

0.4

350 850 1350 1850 2350

Wavelength (nm)

Reflectance CurveFirst Derivative (100x)

This spectrum demonstrates the strong amorphous ironfeatures (400-1000 nm) characteristic of many jaspers.The absorption bands at ~1400, 1900 and 2400 nm aretypical of the -OH bonds in water and the large bandaround 2300 nm is probably associated with carbonatesand/or clay minerals. We have refrained from identifyingall the features since it was unnecessary for ourstatistical approach.

Absorption Features in Conklin Jasper

Upper Mercer Chert

200°C 400°C 600°C

Knife River Chert

200°C 400°C 600°CFlint Ridge

0%

5%

10%

15%

20%

25%

30%

35%

40%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

0%

5%

10%

15%

20%

25%

30%

35%

40%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

5%

10%

15%

20%

25%

30%

35%

40%

45%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

5%

10%

15%

20%

25%

30%

35%

40%

45%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

200°C

400°C

600°C

200°C 400°C 600°C

Porcelenite

200°C 400°C 600°CArkansas Novaculite

5%

7%

9%

11%

13%

15%

17%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

5%

7%

9%

11%

13%

15%

17%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

7%

12%

17%

22%

27%

32%

37%

42%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

7%

12%

17%

22%

27%

32%

37%

42%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

7%

12%

17%

22%

27%

32%

37%

42%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

AcknowledgementsWe thank the Kent State Department of Geology of for providing facilities toconduct this research. We especially thank Merida Keatts for help in printingthis poster. Eric Katz helped proofread this poster, all errors are solely his.

Heat-treatedUn-heated spectra

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

0%

5%

10%

15%

20%

25%

30%

35%

40%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

10%

15%

20%

25%

30%

35%

40%

45%

50%

55%

60%

450 950 1450 1950

Wavelength (nm)

Refl

ect

an

ce

0+

0-

2+

2-

4+

4-

6+

6-

8+

8-

Angle

Effect of Background onTranslucent Samples

This graph shows the effects of sample backgroundtaken on a thin translucent chert sample. Note theclay and the black cardboard increased thereflectance of the sample. Blue arrows indicateexamples of major spurious features that couldhave been assigned to the chert spectra had theeffects of background materials not been noted.

5%

10%

15%

20%

25%

30%

35%

40%

350 850 1350 1850 2350Wavelength (nm)

Refl

ect

an

ce

White Clay

Black Cardboard

~100mm of air

92%

94%

96%

98%

100%

102%

104%

350 850 1350 1850 2350

Wavelength (nm)

Refl

ect

an

ce

0°

-2°-4°-6°

-8°+2°+4°

+6°+8°

SampleAngle 0 -2 -4 -6 -8 2 4 6 80 1.000

-2 0.067 1.000-4 -0.169 0.847 1.000-6 -0.232 0.786 0.975 1.000-8 -0.235 0.791 0.974 0.992 1.0002 0.282 -0.618 -0.572 -0.499 -0.538 1.0004 0.168 -0.672 -0.597 -0.513 -0.553 0.976 1.0006 0.141 -0.625 -0.512 -0.416 -0.459 0.964 0.986 1.0008 0.263 -0.624 -0.609 -0.535 -0.574 0.955 0.964 0.959 1.000

0

0.1

0.2

0.3

0.4

0.5

0.6

350 850 1350 1850 2350

Wavelength (nm)

Re

lfe

cta

nc

e

Hixton Quartzite

Porecelanite

Rhyolite

Rose Quartz

Mahogany Obsidian

Mean Chert

T H E I N S T R U M E N T

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

3.5%

325 825 1325 1825 2325

Wavelength (nm)

Sta

nd

ard

Devia

tio

n (

%)

Sample variabilityError due to angleInstrument NoiseTotal Error