Professional Documents
Culture Documents
Stones and Shafts Redux: The Metric Discrimination of Chipped-Stone Dart and Arrow Points
Author(s): Michael J. Shott
Source: American Antiquity, Vol. 62, No. 1 (Jan., 1997), pp. 86-101
Published by: Society for American Archaeology
Stable URL: http://www.jstor.org/stable/282380 .
Accessed: 20/06/2014 16:50
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Society for American Archaeology is collaborating with JSTOR to digitize, preserve and extend access to
American Antiquity.
http://www.jstor.org
Michael J. Shott
MAany of the chipped-stonebifaces so common in the archaeological recordfunctioned as the haftedpoints of darts or arrows.
For archaeologists, these artifacts possess two salient properties: (1) theyformed only part of a larger apparatus, but, (2)
because perishables decompose, they ordinarily are the only part preserved. Consequently,the identity of that apparatus-
i.e., whetherdart or arrow-is not readily apparent. For various reasons, we may wish to know if stone bifacesfunctioned as
dart or arrow points. Often we rely on reasonable assumptions, but Thomass (1978) discriminantanalysis isis a more reliable
way to distinguish the possibilities. This study extends Thomass approach by increasing the dart sample and the rate of suc-
cessful classification. Shoulder width is the most importantdiscriminating variable.An independenttest on a set of arrows
,lso strengthensconfidence in the results.
AJuchosde los bifaces liticos encontradosen el registro arqueologicofuncionaron como puntas enmangadas de dardos ofle-
chas. Talesartefactos poseen dos propiedades salientes: (1) conformaronsolamenteparte del instrumentomds grande; pero,
(2) porque lo perecedero se descompone, a menudoson la iunicaparte preservada. Por lo tanto, la identidad del instrumento,
ya sea dardo oflecha, no se puede precisar. Por varias razones, quisieramossaber si los bifaces funcionaron como dardos o
flechas. A menudo, contamos con suposiciones justas, pero el andlisis discriminante de Thomas (1978) es un modo mds
seguro para distinguir las opciones. Este estudio amplifica su estrategia mediante una muestraperteneciente de dardos mds
grande y una tasa de su clasificacion mds alta. El ancho del hombrosurge como variable mds importante.Una prueba inde-
pendente en una colecion deflechas procedente de la Gran Cuenca de los Estados Unidos tambienfortalece la confianza en
los resultados.
From the earliest accounts in New Spain to plete and wholesale? Although such questions
Hollywood's Golden Age, few items are as may seem unenlightening(Larralde1990:100) or
central to their tradition-bound popular even tiresome (Corliss 1980), they bear on
image as Native Americans'bows and arrows.Yet important theoretical issues. Conventional
archaeologistsbelieve thatthe earliestAmericans assumptionsof the bow and arrow'ssuperioreffi-
did not use them, the bow and arrow being a ciency in hunting are doubtful (Larralde 1990;
comparatively recent innovation that replaced Shott 1993:435-438); other explanations invoke
earlier dart technology and spread across North political incentives and consequences (Blitz
America from an arctic source. Most locate the 1988; Maschner 1991; Shott 1996) or, from Old
replacement in the first millennium A.D. (Beck Worldperspectives,the social dimensions of dart
i995; Blitz 1988; Christenson 1986; Hall 1977; (Cundy 1989:17-18; Rose 1960:238-242; Testart
Shott 1993; Troeng 1993). Everyone knows what 1988:11) and arrow technology (Edmonds and
bows and arrows are; darts are weapons with Thomas 1987; Vinnicombe 1972:201-202).
longer shafts, usually with foreshafts as well, Darts may not have been abandoned at once
propelledby a device known as an atlatl, sling, or when the bow and arrowappeared(Chatterset al.
spear-thrower.The last term in particular is a 1995; Heizer 1938; Larralde 1990:6; Massey
misnomer (Perkins 1992:65), and "atlatl"is used 1961; Shott 1996), but the timing of their proba-
here for consistency. The broad archaeological bly gradual abandonmentbears on the persis-
consensus begs importantquestions: how, why, tence of atlatls in ritual contexts and the rate at
and exactly when did the bow and arrow replace which their formalattributeswere alteredin sym-
earliertechnology,and was the replacementcom- bolic representation(Hall 1977).
Michael J. Shott * Departmentof Sociology and Anthropology,Universityof NorthernIowa,CedarFalls, IA 50614-0513
86
mm
Max. Thick- Neck Published
CatalogNo.a Length Width ness Width Provenience Source Museum
349188 41.1 17.4 3.4 8.3 White Dog Cave,Arizona Guernsey & Kidder 1921 SI
11/7254 35.8 24.2 7.2 15.3 Allred Shelter,Missouri Harrington1960 NMAI
D2429 81.7 20.1 6.4 21.9 Kimberley,NW Australia PM
97179 21.8 14.0 2.9 9.8 SteamboatCave, New Mexico Cosgrove 1947 PM
97179 39.5 20.5 4.3 11.2 SteamboatCave, New Mexico Cosgrove 1947 PM
A4515-1/6 55.8 22.6 4.2 13.6 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
A4515-1/7 70.4 20.0 4.2 14.9 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
A4515-1/8 60.4 20.2 4.0 13.9 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
A4515-1/9 54.1 20.3 4.9 15.3 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
A4515-1/10 60.4 20.5 4.6 10.4 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
A4515-1/11 60.7 24.9 4.5 18.8 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
A4515.30-47 34.3 20.4 4.8 13.4 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
A4515.30-48 26.5 23.7 5.2 16.7 Cave 2, San JuanCounty,Utah Woodward1930 LACMNH
NA-7031 42.3 16.2 5.4 11.4 Point Barrow,Alaska UMUP
SA-3758 38.0 18.0 5.0 16.5 Nazca vicinity, Peru UMUP
P29-30.4G1 44.1 24.0 5.2 17.5 Sand Dune Cave, Utah Lindsayet al. 1968 MNA
P29-30.4G2 54.6 27.5 4.8 15.6 Sand Dune Cave, Utah Lindsayet al. 1968 MNA
P29-30.4G3 49.2 24.6 5.7 19.1 Sand Dune Cave, Utah Lindsayet al. 1968 MNA
P29-30.4G4 48.5 17.4 4.6 13.3 Sand Dune Cave, Utah Lindsayet al. 1968 MNA
P29-30.4G5 57.4 23.4 5.4 17.7 Sand Dune Cave, Utah Lindsayet al. 1968 MNA
P29-30.4G6 63.0 23.1 5.1 19.1 Sand Dune Cave, Utah Lindsayet al. 1968 MNA
66.56.3.3 57.2 27.4 4.0 17.9 San JuanCounty,Utah Montgomery1894 BYUb
66.55.3.3 58.8 28.0 5.5 18.2 San JuanCounty,Utah Montgomery1894 BYU
66.56.5.1 60.4 30.0 5.8 19.2 San JuanCounty,Utah Montgomery1894 BYU
66.55.3.1 52.8 27.5 5.1 17.4 San JuanCounty,Utah Montgomery1894 BYU
66.56.3.1 85.3 32.0 5.8 19.3 San JuanCounty,Utah Montgomery1894 BYU
66.55.3.2 66.5 28.4 5.0 13.8 San JuanCounty,Utah Montgomery1894 BYU
66.56.5.2 67.1 29.7 5.5 19.9 San JuanCounty,Utah Montgomery1894 BYU
95.2.147.1 48.9 22.5 5.5 12.5 San JuanCounty,Utah Montgomery1894 BYU
95.2.115.1 61.3 24.6 6.0 16.8 San JuanCounty,Utah Montgomery1894 BYU
BYU=BrighamYoungUniversityMuseum of Peoples and Culture LACMNH=LosAngeles CountyMuseumof NaturalHistory
MNA-Museum of NorthernArizona NMAI=NationalMuseum of the American Indian PM=PeabodyMuseum SBCM=San
BernardinoCountyMuseum SI=SmithsonianInstitution UMUP=UniversityMuseum,Universityof Pennsylvania.
aSee Thomas (1978:Table3) for additionaldata.
bSpecimenswere taken in the 1890s from one or more caves located between Moab and Bluff City (Montgomery 1894:227).
Montgomery(1894:227--230)reportedspecimens from several caves, including "a number"from one burial in Cave No. 1 and
six from anotherburial there. He (1894:227) identified all as "crude,stone arrow-pointswith short, wooden . . . handles,"but
Pepper's(1905) account and direct examinationleave no doubt that the specimens were dart points. Pepper (1905:127-129)
described 13 specimens from Cave No. 1 or "CaveDwelling," only nine of which were received by BYU from the Deseret
Museum,where the collection originallywas housed.
Poor preservation often forces us to seek Nusbaum (1922:126) long ago proposed such
answers from the only remaining component of a method,but Thomas (1978) was the first to use
prehistoric weapons, their stone points. Under it. He summarized earlier attempts (see also
these circumstances, we can distinguish darts Chatterset al. 1995, Christenson1986) to distin-
from arrowsonly by the size and form of the pre- guish dartsfrom arrowson the basis of attributes
served stone point. Too often the distinction is such as weight and neck width (Thomas
made by assumption (Christenson 1986; Shott 1978:461). Thomas also noted experiments
1993), which can be reasonabledependingon cir- demonstratingthat arrowspossessing large points
cumstances. But there are better ways to distin- and dartspossessing small ones could fly. But he
guish dart from arrow,and therebyto addressthe recognized that the more importantquestion was
importantquestionsposed above. an empirical one: "what existentially is, rather
than what theoretically could be" (Thomas filled with specimens acquiredbefore the advent
1978:466 [emphasisin original];see also Fenenga of rigorous standardsregardingthe provenience
1953:319). He then examinedmuseumspecimens and documentation of collections. The market
known to be dart or arrow points because they spawnedby avid collectors may have encouraged
still were hafted to dartor arrow(fore)shafts.His the fabricationof specimens not necessarilytypi-
discriminant analysis produced encouraging cal-in manufacture,size and form, decoration,
results; good metric discrimination of the two manner of use-of those in common use by the
point types was obtained, and virtually all 132 cultureto which specimens are attributed(Cundy
arrowswere identified correctlyby the resulting 1989:76; see DeBoer 1985 for examples in pot-
classification functions (Thomas 1978). In addi- tery). For instance, several University of
tion, seven of 10 dartswere identified correctly. Pennsylvaniaspecimens may consist of archaeo-
Thomas's study was a valuable start marred logical bifaces hafted to then-modem shafts or
only by the "painfully small" (1978:468) dart foreshafts (J. Cotter, personal communication
sample of 10 specimens. Fortunately,various 1994). Second, authentic archaeological speci-
museums hold darts with hafted stone points. mens can be of unknownfunction. Several from
Visits to 11 of them between 1990 and 1995 PgHb at the CanadianMuseum of Civilization,
increasedthe sample from 10 to 39. Even this fig- for example,bore stemmedchertbifaces haftedto
ure is modest, but stone points hafted to dart crude short split shafts roughlythe length of fore-
shafts or foreshafts are comparatively rare. shafts.These certainlyresembledartpoints found
Althoughmany museumshold one or a few spec- elsewhere but, unlike other specimens, their
imens, it is unlikely that the compiled data could shafts were flatterin cross section than the circu-
be substantiallyincreasedby visits to more North lar to slightly elliptical form found on virtuallyall
Americanmuseums. Metric attributesand prove- demonstrabledart shafts elsewhere. In one case
nience of these specimens are listed by museum (cat. no. 14474c) a flake was hafted laterallyin a
in Table 1.1 slot at the end opposite the biface, while another
No point and foreshaftwas found haftedto the (cat. no. 14112a) bore bifaces at both ends of the
largerdartshaft on which it presumablywas used. handle. These specimens could in fact be darts,
Strictly speaking, then, I assume that they func- althoughit is hardto imagine how they were used
tioned on dartsas opposed, for instance, to being as such. They were not included in the analysis.
used handheld as knives. But ethnographic Third, authentic projectiles can be launched by
sources (e.g., Nelson 1899) routinely report the atlatls (i.e., genuine darts), be launchedby hand
use of points hafted to foreshafts as detachable (i.e., spears), or be held in the hand and used by
elements of darts, and no known source reports thrusting(i.e., lances or hand-held spears). Most
tools of this natureused as handheldknives. The ethnographicspecimens lack the associateddocu-
dart foreshafts found on Table 1 specimens usu- mentation that could distinguish these alterna-
ally are tapered,have circularor slightly elliptical tives, althoughSmithsonianInstitutionspecimens
cross sections, and always are more finely made in particularoften are well documented ethno-
than the broader handles, often flatter in cross graphically (e.g., Nelson 1899). Lacking docu-
section, found on knife handles (e.g., Fowler and mentation, specimens cannot be identified
Matley 1979; Nelson 1899; note also the speci- reliablyby function,and the performancerequire-
mens from Canada'sPgHbl site describedin the ments of the variousalternativescan differ signif-
following paragraph). Also, knives rarely are icantly in ways that register in size and form of
found in sets of four or more, while dartshafted the point (Cundy 1989). Fourth, even weapons
to foreshafts often are. General cutting purposes launchedby atlatl may differ in context and man-
rarelyrequiredthe use of severalspecimensat one ner of use in ways that bear on the size and form
time, whereas dart points routinely were carried of their points. Thomas (1978:468), for instance,
in sets for the rapidrearmingof a single shaft. cautionedagainstthe facile equationof dartsused
Forseveralreasons,even the expandeddataset in terrestrialvs. marinesettings.The diversityand
demands critical treatment.First, museums are complexity of marine hunting equipmentindeed
are remarkable(e.g., Nelson 1899). Harpoonsare authentic; and (4) they were not known to be
used commonly in marinehabitats,and handheld designed for use in marine hunting.Table 1 lists
spears and lances also seem to be more common the specimens that satisfy these conditions.
there than elsewhere. Especially in the North This data set is less than ideally representative
Americanarctic, coastal settings also were popu- of the complete time and space range of dartuse.
lar among ethnographic collectors, such that Most specimens are archaeological, which is
museum collections like those in Washingtonand unavoidableconsideringthe limiteduse of dartsin
Ottawa disproportionately contain marine the ethnographicpast (Massey 1961). Most are
weapons. Thomas's point is well taken, and from the American Southwest, equally unavoid-
weapons used in marinesettings can be used only able consideringthe vagariesof organicpreserva-
with care. (For this reason, 11 Field Museum of tion; southeastern Utah is especially well
Natural History archaeological specimens from represented.But perfect representativenessis a
coastal Peruwere omittedbecause they are hafted chimera,and we must, for reasons set forth else-
as harpoon, not dart, points.) Fifth, many speci- where (Shott 1993:430-431), use all the evidence
mens are tightly bound in sinew lashing or mastic thatprudentbut criticaltreatmentmakesavailable.
(e.g., sizable collections of AustralianKimberley Not surprisingly,arrow points generally are
darts at the Field Museum and the Harvard's smaller than dartpoints. It is temptingto remove
Peabody Museum), making it difficult if not large outliers from the arrow data set, and there
impossible to measure attributes of the stem are good reasons to consider such a move.
because it is obscured by hafting. The sole "Unlocalized"stone-tippedarrows, for instance,
Kimberley point included in the study (Table 1) may not be authentic. The Menomini arrows
could be measuredin its entirety only because it reported by Skinner (1921 :Figure 57) are con-
had come loose from its mastic haft, although it spicuous metric outliers (cases 12 and 38 in
fit snugly when reinserted.In his study,Thomas Figure 1). Thomas'sMenomini specimens appar-
(1978:467) avoided this problem by x-raying ently are ethnographic,perhaps acquired for the
specimens, a measurethat, if used systematically, AmericanMuseum of NaturalHistoryby Skinner
would add many specimens to the database.2 himself (1921:19), although not necessarily the
Sixth, some specimens were found that were not ones illustratedin Skinner'smonograph.Yet the
hafted to shafts or foreshafts, but retained sinew Menomini abandonedstoneworkinglong before
lashing and/or resin adhesive on their haft ele- ethnographic study (Hoffman 1896:256, 274,
ments, e.g., Newberry Cave specimens housed at 281; Skinner 1921:323), which suggests that the
California's San Bernardino County Museum tools were fashioned for donation or sale to col-
(Davis and Smith 1981:Figures 6g,f, and 7d), lectors. Indeed, Skinner describedthe making of
Shinner Site C cache (Hattori 1982:Figure then-modem arrows using stone points retrieved
44c-f,i). These archaeological specimens were from archaeologicaldeposits and hafted to mod-
found in dry caves that also yielded abundantper- ern shafts using "instructions received in a
ishable remains of atlatls (e.g., Davis and Smith dream" (1921:323). Doubts can be entertained
1981:38-53). Unfortunately, the stone points aboutthe authenticityof such arrows,but they are
were not found attached to dart shafts or fore- provisionallyretainedfor analysis.
shafts; although perishable remains of bow and
arroware rareif not absent at these sites, the dis- Metric Analysis
embodiedpoints can be identified as dartsonly by This study uses Thomas's arrow-pointdata, so
an assumptionthatthis and an earlierstudy (Shott thereis no need to repeathis analysiswhere it per-
1993) are at pains to avoid. tained to arrows only. For completeness,
An uncritical approach easily could increase Thomas's analysis of darts alone is reproduced
the possible dart sample to 75 or more. But this below using the newly expanded data set,
analysis includes specimens only if: (1) they were althoughThomas's arrow statistics differ slightly
hafted to a shaft or foreshaft; (2) all attributes from those reported here. Thomas (1978:469)
could be measured; (3) they were undoubtedly reporteda significant differencein dartand arrow
S
H
80U *80.? 0 *38
.12 U 30 20812 i
L L
E 60 D
E 53, 67
N
G R 20
T 40
H W
I
D 10-
T
H
20
0-
132 39 132 39
la ARROW DART lb ARROW DART
112 30
110 *12
T 038 N
H 8 E o12
020 C 20 038
I
C 055 K
K 6
N W
E I I
S 4 D
S T 10
H I l
2- I
OA' 0
132 39 132 39
1c ARROW DART 1d ARROW DART
ARROWS DARTS
50-
Std. Dev = 3.91
Mean = 14.7
40 * N =132
30.
20
10
.. . . . .
10 14 18 22 26 30
2a WIDTH 2b WIDTH
this case, the first assumptionis completely valid. 1.0 for most arrowvariables(Table 2, Figure 1).
The second is unimportantif analysis is descrip- Removal of the two Menomini extreme outliers
tive only (Baxter 1993:188) but must be evaluated noted above reduces arrowskewness to below 1.0
when results are extended beyond the analyzed except for shoulder width. Normal plots and
cases. Results are robust under "mild skewness" Lilliefors tests (Norusis 1993:190) also demon-
(Baxter 1993:197; see also Klecka 1975:10) but stratethe normalityof dartvariables.Only length
can be sensitive to highly skewed distributions. among arrow variables tests as normal; again,
Technically,multivariatenormality requires nor- removing Menomini outliers produces normality
mal distributionsfor each variable around fixed in arrow variables except for shoulder width.
values of all others (Klecka 1975:10). The Thus, the assumptionof multivariatenormalityis
assumptionis evaluatedhere instead by examin- validated for darts and validated for arrow vari-
ing each variable's separate distribution. ables except shoulderwidth only with the removal
Skewness is modest for dart points, but exceeds of outliers. Figure 2 shows the distribution of
shoulderwidth for arrowsand darts.Box's M tests
Table2. Mean MetricAttributesof the for the violation of the thirdassumption,although
Arrow and Dart Samples.
it can exaggerate the probability of significant
Attribute Mean S.D. S.E. Skewness Kurtosis inequalityat large sample sizes and is sensitive to
Arrows(n = 132) skewed data (Baxter 1993:199).
Length 31.1 9.3 .8 1.0 3.1 It is importantthat any difference in results
Shoulderwidth 14.7 3.9 .3 1.7 4.4 fromThomas'sowes to the additionaldata,not the
Thickness 4.0 1.3 .1 1.4 4.1
Neck width 10.0 .2
analyticalmethods used. To ensure close similar-
2.9 .9 1.5
ity to Thomas's methods, therefore, his results
Arrows (less Menominioutliers)(n = 130) first were reproduced in relevant particulars.
Length 30.6 8.3 .7 .3 -.2 Thomas (1978:469) apparently used stepwise
Shoulderwidth 14.4 3.4 .3 1.2 2.3 variable entry, a practice that Baxter (1994)
Thickness 3.9 1.1 .1 .7 1.0
Neck width 9.8
recently suggested could bias results. Using the
2.6 .2 .5 -.2
same method, Thomas's result was indeed repli-
Darts(n = 39) cated in most particularsincluding Box's M sta-
Length 51.7 14.0 2.2 .1 .0 tistic, discriminant-function and
Shoulderwidth 23.1 4.6 .7 .0 -.9 classification-function coefficients, and classifi-
Thickness 5.0 1.0 .2 .1 .1 cation results.Functionconstantsdifferedslightly
Neck width 15.2 3.3 .5 -.2 -.8
from Thomas (1978:470), each being approxi-
Note. All dimensions in mm.
mately .7 higher in value. This difference did not
Or f * _ __
V"
10S 20 3o 40 10 20 30 40
3a SHOULDER
WIDTH 3b SHOULDER WIDTH
THICKNESS
VS. SHOULDERWIDTH
TWO-VARIABLESECTION
u)
z
'i
I-
3c SHOULDER WIDTH
ClassificationResults
Shoulder Thick- Neck Arrows Darts
Length Width ness Width Correct Incorrect Correct Incorrect
Four-variablesolution (Box's M=47.3, p=.000)
Coefficient .51 .67 -.30 .09 118 14 30 9
Wilk'sX .59 .57 .89 .64
(.00) (.00) (.00) (.00)
solution clearly identify shoulderwidth as much width, and 81 of the 83 Numa points (97.6 per-
more importantthan thickness. When it alone is cent) are correctlyclassified by them. One of the
entered in analysis using the same data as in the two misclassified specimens (Fowler and Matley
three- and two-variablesolutions, similar results 1979:Appendix Table 1, Figure 51e [cat. no.
occur. 14540/5]) bears a strong resemblance to Ohio
Box's M yields insignificant results. Valley bifaces identified as dartpoints in an ear-
Classificationresults by specimen and overall are lier study (Shott 1993:433).
identicalto the three-variablesolution.The classi- As a very inadequatetest for dart points, the
fication functions are: partial metrics for a hafted dart from southern
Dart: 1.40(shoulderwidth) - 16.85 Nevada (Tuohy1982:85, Figure2) can be inserted
Arrow:.89(shoulder width)- 7.22. in the one- and two-variablesolutions. Both cor-
The D-score histogram resembles those for rectly classify the specimen as a dart point.
other solutions (Figure4d). Gunnerson (1969:101) reported width for two
The two- and one-variable solutions may be hafted darts;insertedin the one-variablesolution,
the most importantfor prediction. They include these too are classified correctly.
shoulder width, which consistently is the most
important contributor to results as it was in Problematics of Distinguishing Dart from
Thomas'sanalysis.The equal-covarianceassump- Arrow Points
tion is more reasonablein these cases, and their As Thomas (1978) noted, the size difference
successful classification rate matches the three- between dart and arrow points is an empirical
variable solution and exceeds the four-variable more than a theoreticalmatter.Small points could
one. Moreover,these solutions remove length and be used on darts,large ones on arrows.The possi-
neck width and so apply to many more archaeo- bility is exemplified in an earlier study (Shott
logical specimens. They can be applied to 1993:433) and in the persistentmisclassification
unnotched specimens and to broken ones that of some darts and arrows.It also is illustratedin
retain attributesexcept for length. Together,these GreatPlains data (Larralde1990) and a hafted
by
qualities make the solutions applicable to many dart point from Danger Cave (Jennings
more archaeological specimens than do the oth- 1957:Figure163e) examinedat the Utah Museum
ers. The only reservation is that arrow-point of NaturalHistory.This specimenbearsa massive
shoulder width is not distributed normally, an impact fractureoriginatingat the tip, so neither
inconvenient fact that must be assumed not to original length nor shoulder width can be mea-
prejudiceresults. sured. However, its neck and base width- 11.6
and 16.1 mm, respectively-are relativelynarrow.
Independent Tests Indeed, Jennings (1957:182-183) would have
Cross-validationof classification resultsis prefer- classified the point as an arrow had it not been
able to the resubstitution used in this study hafted to a dartforeshaft.
(Baxter 1993:201). In partialredressof this defi- Specimen SA-3758 is one of the misclassified
ciency, an independentsample of arrowpoints is darts.Besides its point, SA-3758's shaft composi-
classified to test the results. Even this measureis tion, size, and design strongly
suggest continu-
less than ideal, because arrowpoints are success- ities betweendartand arrow
technology remarked
fully classified at consistentlyhigh rates;it would on in other studies (Cundy 1989; Perkins 1992).
be betterto classify an independentdartsample if With such continuity, discriminant
analysis will
such existed. supplyhigh probabilitiesin identification,but not
Nevertheless, ethnographicGreatBasin Numa certainty.
arrowshoused at the SmithsonianInstitutionfur- Dartpoints may be similarin size to the points
nish the test. Fowler and Matley (1979:Appendix of harpoons and thrown and handheld
spears. A
Table 1) reportedlength, width, and thickness,but logical next step in
analysis is to compile suffi-
not neck width, for 83 stone arrow points. The cient data on these latter classes to
attempttheir
one- and two-variablesolutions do not use neck metric discriminationfrom both arrowsand darts.
20
F 111
r 15 111
e 1111
q 11111
u 11111
e 10o 11211
n 11111
c 1111111
y 11111111 1
1 1111111111
11111111111111 11 1
Iiiiiiiiiiiiiiiill 11 1 1 I
---7 Ix .. I
tI
/I/ ouI
out -4X
-4.0 -2.0 .0 2.0 4.0 out
Class 1111111111111111111111111111111111222222222222222222222222222
Centroids 1
4a
4- +
F
r 3 + 22 2 222
e 22 2 222
q 22 2 222
u 22 2 222
e 2 - 22222 2222 222 2
n 22222 2222 222 2
c 22222 2222 222 2
y 22222 2222 222 2
1 - 22 22222 2222222222222 2
2 22222 2222222222222 2
2 22222 2222222222222 2
2 22222 2222222222222 2
X .- x----- + --
out -4.0 -2.0 .0 2.0 4.0 out
Class 111111111111111111111111111111111222222222222222222222222222
Centroids 2
20 1
1
1
F 1
r 15 1
e 11
q 111
u 1 111
e 10o 11111
n 111111 1
c 111111 1 1
y 111111 11 1
5 111111111 11
1111111111111
111111111111111 1 1
I 111111111111111111 1 1 I
x i~ i I11111211
I
--- I
71 Class
out -4.0
-
-2.0
I
.0 2.0
-I.....x
1111111111111111111111111111111111222222222222222222222222222
4.0
I
out
Centroids 1
4b
4 - 2
2
2
F 2
r 3 - 222 22 2
e 222 22 2
q 222 22 2
u 222 22 2
e 2 222 2 222 2 22
n 2 222 2 222 2 22
c 2 2 222 2 222 2 22
y 2 2 222 2 222 2 22
1 2 22 2222222 222222222 2
2 22 2222222 222222222 2
2 22 2222222 222222222 2
2 22 2222222 222222222 2
iI
out -4.0 -2.0 .0 2.0 4.0 out
Class 1111111111111111111111111111111111222222222222222222222222222
Centroids 2
20 -
1
1
F 11
r 15 - 111
e 111
q 111
u 111
e 10 - 1111
n 1111 1
c 111111 111
y 1111111 111
5 - 11111111111
111111111111
-TI7~~~ ~1111111111111111111 1
/\z~~ 1111~111111111111111 1 1 1
X | i | i - ! X
out -4.0 -2.0 .0 2.0 4.0 out
Class 1111111111111111111111111111111111222222222222222222222222222
4c Centroids 1
2
F 2
r 6- 2
e 2
q 2
u 2
e 4 2 2
n 2 2
c 2 2 2 2
y 2 2 22
2 - 2 22 2222 22 2
2 22 2222 22 2
2 22222222 222222 2222 2
2 22222222 222222 2222 2
X I I I X
out -4.0 -2.0 .0 2.0 4.0 out
Class 111111111111111111111111111111111222222222222222222222222222
Centroids 2
20 -
1
F 11
r 15 111
e 111
q 111
u 111
e 10- 1111
n 1111 1
c 11111 1 1 1
y 1111111 111
5 - 111111111111
111111111111
11111111111111111 1
7I1 x I
11111111111111111
I i
1 1
I
1
x
out -4.0 -2.0 .0 2.0 4.0 out
Class 111111111 111111111
1111111111111111 222222222222222222222222222
Centroids
4d 1
8- 2
2
2
F 2
r 6- 2
e 2
q 2
u 2
e 4 -2 2 2
n 2 2 2
c 2 2 22
y 2 2 2 2
2 - 2 22 22 2 22 2
2 22 22 2 22 2
2 2222222 222222 2222 2
2 2222222 222222 2222 2
X I I x
out -4.0 -2.0 .0 2.0 4.0 out
Class 1111111111111111111111111111111111222222222222222222222222222
Centroids 2
Figure 4. Continued.
Only such research can determine if dart points adequate discriminatorbetween dart and arrow
are distinguishableon metric grounds from these points. Although reasonable, their studies were
other classes of points. confined to archaeological specimens of
unknown status. Van Buren (1974:24) also
Single-VariableAnalysis reported length and weight for several undocu-
Shoulder width emerges in all solutions as the mented dart points. Weighing hafted dart and
most significant discriminatorbetween dart and arrowpoints would requiretheirremovalfromthe
arrowpoints, to the extent that using a discrimi- haft, damaging the larger specimen; it is not a
nation function alone produces similar results to practicalprocedurefor the samples used here.
multiple-variablesolutions.Figure2 suggests that Corliss (1972) proposed a metric distinction
the shoulder-widththreshold between dart and between dart and arrow points based solely on
arrow lies near a value of 20 mm. Using this neck width. He did not explicitly specify a thresh-
value, 122 of 132 arrows (92.4 percent)but only old value, but his Figure 3 showed a gap between
30 of 39 darts (76.9 percent) are correctlyclassi- the classes ca. 8.5-9 mm. Beck (1995) and
fied. These results are identical to the relatively Chatterset al. (1995) adoptedsimilarapproaches,
poor four-variablesolution for darts, but better Beck explicitly using a thresholdvalue of 9 mm.
than any multivariatesolution for arrows. Darts Fawcett and Kornfeld (1980:72) located the
are the more difficult class to identify,which jus- thresholdat 10 mm.
tifies use of the one-variable function over the Unfortunately,data do not validate such an
threshold value because of the function's higher approach (cf. Corliss 1980:352). A neck width
dart-classificationrate. threshold of 9 mm correctly classifies 38 of 39
It bears emphasizing that Thomas's original dart points, but misclassifies as darts 82 of 132
data yield similar results, althoughthe following arrowpoints (62.1 percent).A thresholdvalue of
was not reportedin his study. The three-variable 8.5 mm produces identical results for darts but
solutionusing Thomas'sdatayields the same clas- misclassifies 89 arrow points (67.4 percent).
sification results as the four-variableone using Even a thresholdvalue of 10.4 mm, one standard
his data. The two-variable solution (shoulder deviation lower than Chatterset al.'s (1995:757)
width and thickness) correctly classifies eight of mean for inferred dart points, misclassifies 57
10 darts, and the one-variablesolution (shoulder arrows(43.2 percent).
width alone) correctlyclassifies nine of 10 darts, These studies were reasonable attempts to
the lone exceptionbeing the narrowestWhite Dog infer the status of archaeologicalunknowns.But
Cave specimen. width at the shoulder,not the neck, is the best sep-
Thus, shoulderwidth is by far the most signif- aratorof dartand arrowand then, contraCorliss's
icant discriminator of dart and arrow points. (1980) claim, especially in a discriminantanaly-
However, best classification results occur when sis. This study shows that other approachesare
shoulderwidth is used in a one-variablediscrimi- prone to a higher risk of error.
nant analysis, not as a simple threshold value.
Flake Points and the Antiquity of the North
Also, results closely resemble those obtained
American Arrow
using Thomas's original data. In both data sets,
classification success improves as length, neck Odell (1988) arguedthat flakes and bifaces were
width, and thicknessareremovedin order,leaving used as arrow points as early as the Archaic
only shoulder width in the solution. A carefully period,obviouslyproposinga much earlieradvent
chosen single variable can distinguish dart from for bow and arrowtechnology than most archae-
arrow,but does it best when used in a discrimi- ologists would accept. Museum collections of
nant analysis. North American materials include few flake
Otherstudies also stressed single variablesbut points to supportone partof Odell'sview andper-
did not use discriminant analysis or shoulder haps none dating to the time periods he proposed
width as the key variable.Fenenga(1953) andVan for the arrow'sintroduction.Odell (personalcom-
Buren (1974:21-22, 34) proposed weight as an munication1996) legitimatelystressedthe contin-