You are on page 1of 31

Society for American Archaeology

A Multivariate Study of Rock Art Anthropomorphs at Writing-on-Stone, Southern Alberta


Author(s): Martin P. R. Magne and Michael A. Klassen
Reviewed work(s):
Source: American Antiquity, Vol. 56, No. 3 (Jul., 1991), pp. 389-418
Published by: Society for American Archaeology
Stable URL: http://www.jstor.org/stable/280892 .
Accessed: 03/02/2013 13:09

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

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
A MULTIVARIATE STUDY OF ROCK ART ANTHROPOMORPHS AT
WRITING-ON-STONE, SOUTHERN ALBERTA

Martin P. R. Magne and Michael A. Klassen

The rock art of Writing-On-Stone has been the focus of detailed interpretation relying on concepts of group
migration. Indeed, results of previous research at Writing-On-Stone have been pivotal in reconstructions of
northwestern Plains group distributions. It is apparent that many anthropomorphic petroglyphs and pictographs
contain elements and co-associations that allow chronological orderingfrom prehistoric through protohistoric and
historic times. Clusteranalyses describedin this paper offer a more objectivemeans of assessing changes in
anthropomorphdepictionsthroughtime, leading to a critical evaluationof prior conceptsconcerningstylistic
changesand Nativegroupdistributions.In particular,the analysesdemonstratethat rock-artanthropomorphs at
Writing-On-Stoneare more stronglyinterrelatedthan previouslythought. We believethereis no need to invoke
a Shoshoneanpresencein this area to accountfor differencesin late prehistoricand historical-periodrock-art
anthropomorph styles.
El arte rupestre de Writing-On-Stone ha sido el centro de una detallada interpretaci6n basada en conceptos
de migraci6n de grupos. Sin duda, los resultados de anteriores investigaciones en Writing-On-Stone han sido
fundamentales para las reconstrucciones de las distribuciones de grupos en los llanos del noroeste de Norteamerica.
Es evidenteque muchospetroglifosy pictografiasantropomorfosposeen elementosy asociacionescomunesque
permiten un ordenamiento cronol6gico que va desde los tiempos prehist6ricos a los hist6ricos, pasando por los
protohist6ricos. Los andlisis de agrupamiento que se describen en elpresente articulo ofrecen un medio mds objetivo
para evaluar los cambios de las representaciones antropomorfas a traves del tiempo, lo que conduce a una
reconsideracioncriticade conceptospreviosrelacionadoscon los cambiosestilisticosy las distribucionesde grupos
aborigenes. En especial, los andlisis demuestran que lasfiguras antropomorfas del arte rupestre de Writing-On-
Stone estdn mds intimamente relacionadas entre si de los que se habia pensado. Creemos que no es necesario
invocar la presencia de los indios Shosone en esta regi6n para dar cuenta de las diferencias de los estilos antro-
pomorfos del arte rupestre prehist6rico.

Most rock-art research, apart from conservation studies, has taken the form of subjective analysis
of styles, motifs, and their distributions. Even though rock art offers insight into the less concrete
and material aspects of prehistoric cultures, our appreciation of its meaning stands to benefit from
a more objective and systematic approach. An ideal testing ground for new analytic methods is the
rock art of Writing-On-Stone, since it is abundant, spatially restricted, and has been subject to
previous analysis. The rock art is found on approximately 10 km of sandstone cliffs along the Milk
River and several major
or ttributary coulees in and around Writing-On-Stone Provincial Park (Figure
1). Over 60 separate sites are known from this area, including one, DgOv-2, with 26 panels. The
rock art of this area of southern Alberta, both in content and extent, is one of the most important
archaeological records on Plains. Petroglyphs and pictographs along these cliffs
record the prehistoric/historical-period cultural transition of southern Alberta and provide an ex-
cellent source for Plains Native Indian cultural reconstruction. The rock-art figures found at Writing-
On-Stone include anthropomorphs, zoomorphs, and material object motifs, as well as abstract and
geometric designs. By examining patterns of association among depictions of anthropomorphs,
historical objects, and horses at Writing-On-Stone relative chronologies of some of the motifs have
been established (Dempsey 1973; Dewdney 1964; Habgood 1967; Keyser 1977a).

MartinP. R. Magne and MichaelA. Klassen,ArchaeologicalSurveyofAlberta,ProvincialMuseumofAlberta,


12845 - 102 Ave., Edmonton, Alberta, Canada TSN OM6

AmericanAntiquity, 56(3), 1991, pp. 389-418.


Copyright? 1991 by the Society for AmericanArchaeology

389

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
390 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

Of particular interest are the anthropomorphs, being the most common motif depicted at Writing-
On-Stone (Keyser 1977a). A number of anthropomorph categories have been identified previously
and their possible stylistic and historical associations, including identification with specific Plains
Native Indian groups, have been discussed (Dempsey 1973; Habgood 1967; Keyser 1977a). Ac-
cording to the prevalent view, displacement of cultural groups introduced new anthropomorph
categories at Writing-On-Stone. Classic examples of most categories are distinguishable readily, yet
examination of all anthropomorphs at Writing-On-Stone reveals a great deal of variation. A dem-
onstration of overlapping variation between categories would suggest an evolution of anthropomorph
depictions within a single cultural tradition, without need to invoke displacement of one group by
another. Confirming Native group distributions on the northwestern Plains during the late prehistoric
and protohistoric is still a major challenge for archaeology, and rock-art research is a valuable tool
with which to address this challenge.
We describe here an objective analysis of anthropomorphs performed by assigning a presence or
absence value to the attributes of each, then using cluster analyses to test how anthropomorphs
group according to the attributes, evaluating at the same time the validity of categories defined
previously. The results of this analysis will be useful for comparisons of rock-art motifs and rela-
tionships at Writing-On-Stone and other sites throughout the northwestern Plains.

ANTHROPOMORPH CATEGORIES, AGES, AND FUNCTION


At least 46 of the 60 Writing-On-Stone area sites have anthropomorphs depicted, often with
several categories occurring on the same panel. Most anthropomorphs at Writing-on-Stone are not
executed randomly, but are composed of highly conventionalized design elements. The presence of
these elements has allowed some insightful inferences to be made concerning their origins and
interrelations. In the most comprehensive study of Writing-On-Stone to date, Keyser (1977a, 1977b)
described seven anthropomorph categories (see Figure 2): (1) shield-bearing (or warriors), (2) "clas-
sic" V-neck (or pointed shoulder, V-shoulder), (3) "other" V-necks, (4) rectangular (or square
shoulder), (5) hourglass (or X-shaped), (6) stick, and (7) "unique" representations. This classification
scheme is based partially on earlier work at Writing-On-Stone (Dempsey 1973; Dewdney 1964;
Habgood 1967) and classifications used at other sites throughout western North America. The last
two categories rarely are depicted and are not considered further in this paper.
Shield-bearing, V-neck, rectangular, and hourglass anthropomorphs generally are easy to distin-
guish, but all the motifs share certain characteristics. Most are petroglyphs and are executed as
outlines via incisions or "scratches" in the sandstone. Occasionally further detail has been added
to the shields with designs and to the torsos by way of "heartlines" or "lifelines" (Habgood 1967;
Keyser 1977a), genitals, or other decorations. A few pictograph examples have the shields or torsos
partially or entirely filled in with paint. Shield-bearing anthropomorphs (holding large body shields,
as opposed to smaller equestrian shields) are known to occur at several other Alberta sites (Habgood
1967; Leechman et al. 1955:40; Wormington and Forbis 1965:133) and throughout the northwestern
Plains (Conner and Conner 1971; Gebhard 1966; Keyser 1975, 1977a, 1984; Keyser and Knight
1976; Schuster 1987; Sundstrom 1984; Wedel 1969; Wellman 1979) and the Great Basin (Castleton
1978; Gebhard 1966; Keyser 1975; Schaafsma 1971; Schuster 1987). V-neck anthropomorphs also
have been found throughout the northwestern Plains (Conner 1980; Conner and Conner 1971;
Keyser 1977a, 1984; Over 1941; Schuster 1987; Sundstrom 1984), including several other Alberta
sites (Habgood 1967; Leechman et al. 1955:40; Wormington and Forbis 1965:133).
The distributions of rectangular and hourglass anthropomorphs are not as well documented.
Keyser (1977a) states that rectangulars are found throughout North America, but simple outline
figures similar to those at Writing-On-Stone do not appear to be common. In addition to Writing-
On-Stone, some rectangular examples are known from Wyoming (Schuster 1987), Montana (Conner
and Conner 1971), Saskatchewan (Jones and Jones 1982), South Dakota (Keyser 1984; Sundstrom
1984), and the Canadian Shield (Habgood 1967). Hourglass figures are common in Canadian Shield
rock art (Dewdney 1964) and rare overall on the northwestern Plains, occurring at Writing-On-
Stone, at a few sites in South Dakota (Over 1941:56; Sundstrom 1984:71, 114), Montana (Conner

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCKARTANTHROPOMORPHS
INSOUTHERN
ALBERTA 391

* 81 Rock art site and number \


Park road
-- Edge of sandstone cliff
--- Park boundary (1988) Edmonton

0 .5 1 KM \Calgary
STUDY
AREA

Figure 1. Locationsof rock-artsites in Writing-On-StoneProvincialPark. Rock art can be foundwithin 10


km west of the park;locationsof severalsites discussedin the text are not shown.

1980; Conner and Conner 1971; Keyser 1977a), and at a site in New Mexico (Schaafsma 1972:7).
Rectangular and hourglass anthropomorphs also appear as designs on cultural artifacts associated
with the Blackfoot Nation (Conner and Conner 1971:41; Dempsey 1973).
A relative chronology has been proposed for the various anthropomorph categories at Writing-
On-Stone. Most researchers consider shield-bearing anthropomorphs to be the oldest, followed
roughly by V-necks, then rectangulars, and finally hourglass anthropomorphs (Dempsey 1973;
Dewdney 1964; Habgood 1967; Keyser 1977a). Several of the different categories at Writing-On-
Stone also have been used to infer cultural affiliations. Keyser (1975, 1977a) has identified shield
bearers as late prehistoric in age, citing as evidence their general lack of historical associations (i.e.,
guns and horses). Keyser (1977a:55) also has concluded that the shield-bearer motif is Shoshonean
in origin, based "on the almost exact coincidence of its distribution with the known range of the
Plains Shoshone during the late prehistoric period and early historic period." Other evidence cited
is the presence of Intermontane pottery at three shield-bearing anthropomorph sites in Montana
(Homed Owl Cave, Pictograph Cave, Empty Gulch [Keyser 1975:210-211]) and the apparent
depiction of Shoshonean horse armor on a panel at Writing-On-Stone (Keyser 1977a:43-44). "Clas-

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
392 AMERICAN
ANTIQUITY [Vol. 56, No. 3, 1991

"CLASSIC"V-NECK

BARM
(5)
VNCK
HRTL (1)
(11)
FNGR CARM
(13) GNTL (6)
(10)
BELB
FEET
(14)
(12)

"OTHER"V-NECK
WEAP
(19)2)
XNCK
| -----/.

WEAP
ACTN \-" BLEG L i I (19)
(15)
/
(18) f7
\ FEET J
(12)

RECTANGULAR
HNCK WEAP ---FNGR
(3) t // (13)
(19)

ACTN HIST
(15) (16)

Figure 2. Anthropomorphs representing typical categories proposed by Keyser (1977a) and variables employed
in analyses (variable scales).

sic" V-neck anthropomorphs as described as late prehistoric in age since they apparently
also have been
show few historical associations, and as Shoshonean in origin, due to their "close spatial and temporal
association with shield-bearing warriors" (Keyser 1977a: 5 5; see also Dempsey 1973; Habgood 1967).
Furthermore, a number of shield-bearing anthropomorphs at Writing-On-Stone and elsewhere also
have V-neck bodies (Keyser 1977a).

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
ALBERTA
INSOUTHERN
ROCKARTANTHROPOMORPHS 393
Magne and Klassen]

Keyser (1977a:26) considers it unlikely that "other" V-neck anthropomorphs "are derived from
the 'classic' V-neck anthropomorphs, since the only attribute they share in common is the V-neck."
According to Keyser (1977a), most "other" V-neck anthropomorphs are found in association with
historical indicators and thus they are historic in age. Other writers discussing Writing-On-Stone
have not made such clear distinctions between groups of V-necks, but acknowledge that these figures
seem to span from late prehistoric to historic periods (Dempsey 1973; Dewdney 1964; Habgood
1967). Rectangular anthropomorphs, which seem to appear in both late prehistoric and historic
contexts, are not clearly identified with a specific age or cultural group by Keyser (1977a) or Habgood
(1967). On the other hand, Keyser (1977a) dates all hourglass anthropomorphs to the historic period.
Keyser (1977a:56) believes much of the historic rock art at Writing-On-Stone "is probably the work
of Blackfoot, Cree, Gros Ventre and Assiniboine artists," as suggested by Native group distribution
during the historic period.
In addition to these categories, rock art at Writing-On-Stone has been classified into two functional
categories: ceremonial art and biographic art (Keyser 1977a). The characteristic motifs of ceremonial
art are considered to be mainly shield-bearing anthropomorphs, "classic" V-necks, naturalistic and
"boat-form" zoomorphs, as well as large rectangular anthropomorphs, abstracts, rakes, and bow
spears (Keyser 1977a). Apparently, these motifs rarely are depicted with historical associations,
suggesting they are for the most part prehistoric. Keyser (1977a) believes ceremonial art was personal
in nature, representing spiritual events such as vision quests. Biographic art consists of small rect-
angular and hourglass anthropomorphs, "mature-style" horses, guns, and tipis. Less frequently,
"simple" V-necks, riders, decorated halters, spears, bows, and travois are found. The large number
of guns, horses, and other historical associations indicate this art was created in historic times. The
realistic action scenes of biographic art, with "stylistic conventions" to represent various actions,
has been likened to a crude form of picture writing with defined story lines depicting actual events
in a biographical manner (Keyser 1987). Indeed, the relations between ledger art, hide painting,
and rock art in this part of the Plains could use more attention of the sort given to Kiowa art by
Young (1986). Although Keyser believes they coexisted for some time in the early historic period,
he maintains that biographic art did not evolve from ceremonial art, as evidenced by "morpho-
logical" and stylistic differences between the two forms (Keyser 1977a:54). However, the two terms
cannot be considered mutually exclusive, since certain "ceremonial" panels depict scenes such as
battles between two individual shield bearers, which must be "biographic" in the common sense.
This is clearly another issue in need of further thought.

MULTIVARIATE ANALYSES
Anthropomorph Categories and Variables
"Classic" V-necks, "other" V-necks, and rectangular anthropomorphs from Writing-On-Stone,
as defined by Keyser (1977a), have been selected for detailed analysis. These were chosen because
they closely resemble one another morphologically and may help reveal chronological and stylistic
trends. All three categories are simple in design and occur as incised or painted, outlined parallel-
sided torsos. The torsos extend into straight legs, and thin lines represent arms, necks, and other
elaborations. In all three types, dots or circles represent heads. In examining the attributes used by
Keyser (1977a) to classify the three types (Table 1), it is apparent that distinctions between each
category are ambiguous-there are no attributes or combinations of attributes that clearly define
and distinguish "classic" V-necks, "other" V-necks, and rectangulars, and no specific rules exist for
classifying them. Instead, most differences between categories have been defined on the basis of
somewhat vague qualitative judgements of attribute frequencies (see Keyser 1977a:25-27).
All anthropomorph cases that Keyser (1977a) specifically identified as belonging to a certain
category were located on copies of the original tracings he prepared in 1976. Fortunately, the single
attributes used to distinguish V-neck anthropomorphs (i.e., V-shaped shoulders) from rectangular
ones (i.e., straight shoulders) were identifiable easily in most cases. However, distinguishing "clas-
sics" from "other" V-necks was not straightforward, and when the type of shoulder was indeterminate
it was difficult to classify many figures. In the end, only panels identified by Keyser (1977a:24-25)

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
394 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

Table 1. Attributes Employed by Keyser (1977b) to Distinguish Classic V-Neck, Other V-Neck,
and Rectangular Anthropomorphs.

Attribute "Classic" V-Neck "Other" V-Neck Rectangular Body


Sides parallel parallel parallel
Legs continuation of torso continuation of torso continuation of torso
Shoulders delineated by a V a V or an X horizontal line
Torso ending horizontal line horizontal line or X horizontal line
Neck placement center of V-neck center of V-neck center of shoulder line
Hands, feet often few a few
Calves, thighs some some no
Phalli many few rare
Heads half have rare some
Faces rare no no
Heartlines many rare a few
Arms extended, bent few some, straight
Action rarely often some
View always frontal many side
Associations prehistoric/historic historic prehistoric/historic
Elaborations few some few

as depicting a single category of V-neck were used, eliminating any possibility of misclassifying
them. In other words, whenever Keyser (1977a) indicated that both types of V-neck anthropomorphs
occurred on a single panel, all V-necks on that panel were excluded from the testing. In this way,
the V-necks were not placed in the wrong groups for testing the validity of their grouping. A further
difficulty arose when it was noted that Keyser (1977a:26) classified "other" V-necks into two
subcategories in his description (i.e., "sophisticated" and "crude"), with important descriptive and
cultural distinctions between the two. However, Keyser (1977a:24) only listed one category (i.e.,
"other") in his panel-by-panel table. For the purposes of this study, both subcategories were assumed
to be listed under the single heading in the table, and were treated as one group. Rectangular
anthropomorphs were relatively easy to identify in all the tracings, but all questionable figures also
were omitted. Despite these precautions, it was at times difficult to find the same number of
anthropomorphs on panels as Keyser (1977a) indicated to be present. In some cases, fewer anthro-
pomorphs of certain types were found, while in other cases more were located. When too many
were found, those that most poorly fit Keyser's (1977a) description were eliminated until the correct
number was reached. When it was clear that no other figures could have been the ones to which
Keyser (1977a) was referring, these cases were included. For all three categories, whenever classi-
fications were in doubt, they were excluded.
In total 28 "classic" V-necks, 56 "other" V-necks, and 101 rectangular body anthropomorphs
were identified and located on the rock-art panel tracings of Writing-On-Stone. Forty-three of the
rectangular anthropomorphs are found on a single integrated panel at DgOv-81, the Battle-Scene
panel in Writing-On-Stone Provincial Park. Since all of the figures within this panel probably were
created at the same time, they are especially useful in assessing variability within the rectangular
body group. Note that in this initial study the hourglass figures at DgOv-81 and elsewhere were not
analyzed. However, future studies are planned to include these, shield-bearing anthropomorphs,
and a greater range of variation in general. The 84 V-necks and 101 rectangular anthropomorphs
derived by the above criteria represent approximately 39 and 75 percent respectively, of all those
within these types that were found at Writing-On-Stone by Keyser (1977a). Combining these groups,
185 anthropomorph cases were available for analysis.
All cases were then cataloged and coded for the presence or absence of 19 attributes, hereafter
called variables. Table 2 lists these as well as short descriptions of how they were identified. The
variables were based for the most part on the original description of the categories (Keyser 1977a)
except for those described below. For the sake of consistency previous terminology is retained even
if this terminology may be inaccurate or imprecise. Most of the individual variables are executed
with remarkable consistency among all three types of anthropomorphs and thus their presence or

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCKARTANTHROPOMORPHS
INSOUTHERN
ALBERTA 395

Table 2. Variables and Abbreviations Employed in Analyses.

Abbrevi-
List of Variables ation Description
1 V-neck VNCK V-shaped line between the shoulders
2 X-neck and/or X-waist XNCK X-shaped line between the shoulders or at waist
3 horizontal neck HNCK horizontal (straight) line between shoulders
4 straight arm SARM straight line extending from shoulder
5 bent arm BARM line extending from shoulder, bent at "elbow"
6 curved arm CARM line extending from shoulder curving to waist
7 horizontal waist HWST horizontal (straight) waist line
8 neck NECK line extending from center of shoulder line
9 head HEAD dot or circle above neck and/or shoulder line
10 genitals GNTL vertical line down from center of waist line (may terminate in a
small circle), often with dots/circles/short lines representing tes-
ticles
11 heartline HRTL vertical line into body from shoulder line, often terminating in di-
amond/other design
12 feet FEET short perpendicular extensions from end of leg
13 fingers FNGR short lines extending at an angle from end of arm
14 body elaborations BELB calves, thighs, faces, and other body details
15 action ACTN involvement in battle, hunting, riding, etc.
16 historic association HIST contact/obvious association with horses, guns, etc.
17 elaborations ELAB headdresses, fringes, "breastplates," circular "shields"
18 bent leg BLEG legs bent at knee, waist, or curved
19 weapon WEAP contact with guns, spears, bows, etc.

absence was quite easy to determine. There were cases, however, where it was difficult to assess the
presence of a variable. In these instances, a decision was made to consider that variable absent for
that case, rather than treating it as a missing value which would result in that case becoming ineligible
for analysis. The option of creating "indeterminate" categories for "possible" variables was rejected
since cluster analysis using these categories would only result in cases with indeterminate variables
grouping with other cases with the same indeterminate variables, rather than with cases with positive
variables.
Several variables also were created that were not identified specifically by Keyser (1977a) and
that did not fit into his own categories, but seemed to be of relevance (compare Tables 1 and 2).
Furthermore, method of creation (i.e., "petroglyph" or "pictograph") was not used as a variable
since we did not test the differences between media. In any case, there are a number of examples
of combination petroglyph and pictograph figures, indicating that these were not exclusive tech-
niques. Keyser (1977a:26-27, 50-52) did place some importance on the distinction between scratched
and incised petroglyphs, but erosion rates, application tools, and the abilities of various artists are
highly variable, casting doubt on the validity of this distinction. Historical associations were assumed
to be present only when the anthropomorph was in direct contact with a historical object, such as
holding a gun or riding a horse, or when the association was unambiguous, such as involvement in
a historical-period battle scene. The presence of an anthropomorph on the same panel as historical
objects was not considered to necessarily indicate a historical association, as it is possible that rock
art of various ages can be found together. Finally, only a single "legs" category was used, as we
could determine no satisfactory method for defining several. Generally, legs are simply extensions
of the torso, but in cases where waist lines or feet are not represented it was not possible to determine
if the torso did in fact extend into legs.
The coding yielded a data matrix of 18 5 cases by 19 variables for cluster analysis. For the remainder
of this paper, the abbreviations found in Table 2 will be used to signify each of the variables. A
summary of the frequencies of each of the 19 variables and the corresponding percentage of oc-
currence for all three types of anthropomorphs as classified by Keyser (1977a) can be found in
Table 3.

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
396 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

Table 3. Variable Frequency by Keyser's (1977a) Categories.

Classic V-Neck Other V-Neck Rectangular


N==a185 n= 28 n =56 n =101
Variable 2Via #b % nc # %n # %n

VNCK 64 25 89.3 39 69.6 0 0


XNCK 10 1 3.6 9 16.1 0 0
HNCK 94 0 0 0 0 94 93.1
SARM 46 3 10.7 9 16.1 34 33.7
BARM 47 20 71.4 10 17.9 17 16.8
CARM 6 1 3.6 4 7.1 1 1.0
HWST 116 20 71.4 34 60.7 62 61.4
NECK 139 25 89.3 39 69.6 75 74.3
HEAD 90 17 60.7 22 39.3 51 50.5
GNTL 29 12 42.9 11 19.6 6 6.0
HRTL 10 6 21.4 2 3.6 2 2.0
FEET 81 17 60.7 27 48.2 37 36.6
FNGR 33 15 53.6 5 8.9 13 12.9
BELB 10 4 14.3 1 1.8 5 5.0
ACTN 71 1 3.6 23 41.1 47 46.5
HIST 80 1 3.6 20 35.7 59 58.4
ELAB 34 8 28.6 9 16.1 17 16.8
BLEG 15 2 7.1 4 7.1 9 8.9
WEAP 51 2 7.1 14 25.0 35 34.7
a ZVi = number of cases with variable.
b# = number of cases within general category with variable.
c % n = percentage of cases within general category with variable.

Cluster Analyses

Cluster analysis is a method of numerical classification which aims "to discover the pattern of
groupings in a set of data, with as few assumptions as possible about the nature of the groupings"
(Shennan 1988:194). It is based on the premise that the members of each group or cluster "should
be more similar to one another than they are to non-members [and that] with-in group similarity
should be in some sense greater than between-group similarity" (Shennan 1988:193). It should be
obvious that cluster analysis alone cannot
be used to classify anthropomorphs. The reduction of
rock art into binary numbers involves an inherent loss of stylistic and creative nuances, and cannot
be expected to account for all variation. Rather, cluster analysis is used here as a tool to assist in
validating and refining qualitative classification. Cluster analysis can be used with continuous,
interval, or binary data. Binary data, used in this analysis, simply codes the presence/absence
dichotomy of variables with mutually exclusive integers, such as 0 and 1. Cluster analysis then
compares each case with all other cases and records the number of matches for each comparison
as a coefficient of similarity in a distance matrix. This matrix is then used to determine the placement
of the cases in a hierarchical dendrogram. Jaccard's coefficient, which ignores all negative matches
when matching cases, was used in this analysis since it is the best method if large numbers of
variables occur rarely (Shennan 1988:203). A number of algorithms are available for grouping cases
into clusters based on their coefficients of similarity, and it is important that the algorithm used is
one which most closely reflects the actual distance matrix (Shennan 1988:212-232). The degree of
accuracy of the clustering pattern created by different algorithms can be tested with the cophenetic
correlation coefficient (Shennan 1988:230). Using the clustering program available from MIDAS
(Fox and Guire 1976), the cophenetic correlation showed that the average linkage algorithm produced
the best results for analyzing the anthropomorph data.
Cluster Analysis #1: All Variables. Cluster analysis of all 185 cases, using all 19 variables, resulted
in a complex dendrogram of small clusters that generally were related poorly to one another (Figure
3). This weak grouping is likely a result of the large number of absent variables and the high degree

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
'

Magne and Klassen] INSOUTHERN


ROCKARTANTHROPOMORPHS ALBERTA 397

.92

_
' A_
I' I- I I ' L.50
7_ -

IIII I
_ -! _ - -1

CLUSTER 1 4I5 2
CLUSTER CLUSTER 3

Figure 3. Complete
den drogram forall-variables cluster analysis.

of variation possible between cases. In other words, strongly related large clusters are rare as there
are few large groups of cases with very similar complexes of variables. The dendrogram does reveal

exclusivelyin Clusters 2b CLUSTERand


3. Most other variables generally occur in smaller clusters that are

few lar the dendrogroup


athroughoute inofdica
casestes ther
that are no complexes
theseThe ofdendrogram does whichreveal
variables.
ofvariation possible between cases. Inother words,relatedof
strongly large clusters, asthe
which irare expected result
a patte divis
of major ions
between cass at a fairly high disfant as it suggests core
obvious in

noneach in somplexity
clustorican be described. The degreflect ofways thetrcases, togetherwith the majordivisionsopomorph

Some further comments


Sevcategories. on the of nonhistribution
subdivisions
variablethe of
closely correspond to the cluster divisions.
and thilenehistorical,
inCluster3,1 allctive anthropomorphs of 2theshould
Cluster
have be noted. Cluster a VNCK
nearlyall
condly,
detailed anthropomorphscorrespo2b
and 3. Most other variables gener
called "classic" in smaller
V-necks
Keyscur that are);
by (19clusters
aalmost 65 percent
few variables, of all of
much the "classics"
ibetween
the used
thevariation
major
within-group clusters.udyare
is as great with
iables as very low
the frequencies
between-group bsortrand
variation.
T d he
istinc tions
b etw eenV-necks andrectangular anthropomorphs an dbetwe enhistorical and

almost 65 percent of all the "classics" used in the study are included in this group. Clusters lb and

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
398 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

All Variables Analysis Clusters

Classic L

V-Neck

Rectangular

Battle Scene
Rectangular

0 5 10 15 20 25

Numberof Cases

Figure 4. Schematic representation of all-variables cluster analysis.

lc contain simpler anthropomorphs similar to nonhistorical versions of Keyser's (1977a) "other"


V-necks. Clusters 2a and 2c contain V-neck cases, ranging from quite detailed to very simple, which
correspond to historical versions of Keyser's (1977a) "other" V-necks. Cluster 2b is composed of
mostly historical, active rectangular anthropomorphs, many of them from the Battle-Scene panel
at DgOv-81.
The distribution of battle-scene figures is a most interesting result. When a number of anthro-
pomorphs occur on a single integrated panel, it stands to reason that these figures are closely related
and should group together if cluster analysis reflects actual similarities. This assumption is validated
by 29 figures from the Battle-Scene panel (Figure 5) which grouped together in Cluster 2 and the
remaining which grouped together in Cluster 3 (Table 4). Anthropomorphs from several smaller
integrated battle scenes also grouped together. Six out of 12 battle participants at DgOv-57 (Figure
6) grouped side-by-side in Cluster 2a, while the other 6 grouped in the rest of Cluster 2. Two
combatants (one V-neck, one rectangular) from DgOv-60 grouped side-by-side in Cluster 2a (Figure
7). Four out of five of those joined together with a line at DgOv-60 also grouped together in Cluster 1c.
The major patterns observed in this analysis are as follows: Cluster la-mainly nonhistorical,
inactive detailed V-necks; Clusters lb and lc-nonhistorical, inactive simple V-necks; Cluster 2a-
historical, active V-necks; Cluster 2b-historical, active rectangular anthropomorphs; Cluster 3a-
nonhistorical, inactive rectangular anthropomorphs; Cluster 3b-historical, inactive rectangular
anthropomorphs; Cluster 4-nonhistorical, inactive, extremely simple X-necks; and Cluster 5-
nonhistorical, inactive, extremely simple V-necks. Major divisions between anthropomorphs were
based on only a few variables with VNCK, HNCK, ACTN, and HIST responsible for most of this
grouping. This could be interpreted in several ways: either all the anthropomorphs are very similar
to one another in all other respects, except for these four variables; or these four variables are
masking more-subtle distinctions between V-necks and rectangular anthropomorphs, and between
historical and nonhistorical anthropomorphs, which is suggested by traditional classification. Tra-
ditional grouping however, would suggest that even with the removal of these variables from the
clustering, V-necks and rectangulars would still group separately, as would historical and nonhis-
torical cases. To test how much similarity or dissimilarity is being masked by VNCK, HNCK,
ACTN, and HIST, it is necessary to test the cases without using these variables.
Two additional cluster analyses were therefore performed on the 185 cases to substantiate the
results of the first cluster analysis. The first of these additional analyses did not use the VNCK,
XNCK, and HNCK shoulder variables, while the second did not use the ACTN and HIST variables.
It seems valid to remove both ACTN and HIST in the same test mainly because Keyser (1977a)

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCKARTANTHROPOMORPHS
INSOUTHERN
ALBERTA 399

It
K!

K:j K - .. 7.4 11
>r ~ '6
41
~ '4 ~
6
I
Figure 5. DgOv-81 battle scene. Entire panel is 3.37 m in length.

has stated that the presence of ACTN is indicative of a historical context. Indeed, there is a high
degree of correlation between the ACTN and HIST attributes: of 80 cases with HIST, and 71 with
ACTN, 63 cases had both.
Cluster Analysis #2: No Shoulder Variables. Exclusion of the variables VNCK, XNCK, and
HNCK (the cases with these variables remained in the analysis) did not produce a simpler cluster
pattern, but it did cause some significant changes in how the cases grouped together (Table 5). Figure
8 illustrates the distribution of cases with VNCK, XNCK, HNCK, ACTN, HIST as well as Battle-
Scene (DgOv-81) cases within the derived clusters. With the exclusion of the variables VNCK,
XNCK, and HNCK, clustering divided most of the cases into two large and significant groups: those
without ACTN and HIST into Cluster 1, and those with ACTN and HIST into Cluster 2. V-necks
and rectangulars are no longer distinguished strongly. Within the divisions marked by ACTN and
HIST, the cases sorted mainly on the basis of their levels of detail.
Of particular interest is that most V-neck and rectangular cases, no longer discriminated by the
VNCK and HNCK variables, alternated throughout the dendrogram. This indicates that the VNCK
and HNCK variables are not closely associated with a complex of variables that also distinguish

Table 4. Variable Frequency by Cluster, All-Variables Analysis.

la lb lc 2a 2b 2c 3a 3b 4 5
N=185 n=23 n=22 n=5 n=16 n= 15 n=3 n=43 n= 11 n=6 n=5
Variable Vi # # # # # # # # #
VNCK 64 20 21 4 12 0 2 0 0 0 5
XNCK 10 0 1 0 0 4 0 0 0 5 0
HNCK 94 2 0 0 1 44 0 40 7 0 0
SARM 46 0 5 0 7 14 0 18 2 0 0
BARM 44 22 0 0 2 14 0 3 3 0 3
CARM 6 1 1 0 4 0 0 0 0 0 0
HWST 111 18 15 5 9 17 0 38 9 5 0
NECK 138 23 21 1 16 41 0 37 0 0 0
HEAD 90 16 10 0 13 25 1 14 11 0 0
GNTL 24 14 1 5 3 1 0 3 1 1 0
HRTL 10 6 1 0 2 0 0 1 0 0 0
FEET 77 18 7 4 13 15 0 17 5 2 0
FNGR 33 16 1 0 7 3 0 4 2 0 0
BELB 9 5 0 0 0 1 0 1 2 0 1
ACTN 71 1 0 0 15 48 3 3 1 0 0
HIST 80 0 0 0 13 49 3 7 8 0 0
ELAB 34 10 4 0 6 8 0 5 1 0 0
BLEG 15 4 0 0 1 4 0 4 2 0 0
WEAP 51 4 0 0 9 36 1 1 0 0 0

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
400 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

A '1, -X' J]^ '-' At-

0 10 CM

Figure 6. DgOv-57 battle scene.

these cases. In an interesting exception to this observation a core group of "classic" V-necks, very
similar to Cluster 1a in the all-variables analysis, remained together in this test. Furthermore, most
of the battle-scene cases stayed together in two groups as well, which would be expected as these
were almost certainly all created at the same time by the same artist. Thus, the ACTN and HIST
variables were as important here as VNCK and HNCK were in producing the patterns observed in
the all-variables analysis. The result of excluding VNCK and HNCK seems to indicate that V-necks
and rectangulars are often quite similar.
Cluster Analysis #3: No Historical Associations. Using all the variables except ACTN and HIST,
the cases divided into two large clusters and several smaller ones (Figure 9). Two major clusters
were formed based mainly on the type of shoulder variable present: mostly V-necks into Cluster 1
and mostly rectangulars into Cluster 2. While nearly 84 percent of Cluster 1 cases had VNCK, all
cases of Cluster 2 had HNCK. On the other hand, historically associated cases were distributed
throughout all of the clusters (Table 6). Within clusters, the cases also sorted themselves according

Table 5. Major Variables Frequency, No Shoulder Variables


Analysis.

Cluster na VNCKb HNCKC ACTN HIST

la 34 22 8 5 2
lb 43(3) 14 26(2) 1 5
lc 9 6 3 0 0
Id 2 2 0 0 0
Total 1 88 (8) 48 37 (3) 6 7
2a 20 10 8 19 17
2b 35 (1) 4 26 (2) 35 (11) 35 (1)
2c 12(6) 1 11 (5) 9 (5) 12(6)
2d 10(2) 0 8 (2) 1 9(1)
Total 2 77 (3) 15 53 (1) 64 (36) 73 (3)
Others 20 15 4 1 0
a Total cases per cluster (DgOv-81 battle-scene cases excluded).
b Includes XNCK cases.
c Numbers in brackets are totals with DgOv-81 battle-scene cases excluded.

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCK ART ANTHROPOMORPHSIN SOUTHERN ALBERTA 401

0 10CM
i I

Figure 7. DgOv-60 combat scene.

to relative complexity. This produced the following general groups: Cluster la-detailed anthro-
pomorphs, nonhistorical "classic" V-necks; Cluster lb, Ic, Id-fairly detailed anthropomorphs
(mostly XNCK, HNCK, or indeterminate), many with ACTN and HIST; Cluster 1e-entirely simple,
mostly nonhistorical V-necks; Cluster 2a-fairly detailed and nonhistorical rectangulars; Cluster
2b-simple, historical, rectangulars; and Cluster 2c-simple, historical, rectangular anthropo-
morphs.
No group of variables, other than the shoulder variables, seems to be more important than others
in sorting the cases. However, as in the previous tests a core group of "classic" V-necks grouped
together in Cluster la, as did most of the battle-scene cases in Clusters 2a and 2b, further sub-
stantiating the uniqueness of these two groups. The large number of cases from the Battle-Scene
panel at DgOv-81 gives the misleading impression that simple rectangulars are more highly asso-
ciated with ACTN and HIST than are V-necks. If these cases are excluded, the association between
ACTN and HIST to both V-necks and rectangulars is nearly equal. This indicates that historical
and nonhistorical cases are often quite similar and will group together if they have the same shoulder

No ShouldersAnalysisClusters

1IEbbE U
Others

a _V N
.NCK *

"7$i HNCK f \\\\

BATTLE

t ^ACTN r B\ I
~
,s~
HIST L _ i\\ \

..J Numberof Cases


0 10 20 30
I t I I ! I

Figure 8. Schematic representation of no shoulders cluster analysis.

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
402 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

No Action or Historic Variables Analysis Clusters

I R3 M F2b]g
Others

V/XNCK K~/////////////,

,/~\ HNCK - NN\NN14


h\\~~~NNKx\'

BATTLE
K*K^1

^f^ ACTN U
T- h HIST U k \\ *
Number of Cases
0 10 20 30 40 50

Figure 9. Schematic representation of no ACTN, HIST cluster analysis.

variable. Complexity of the anthropomorphs was also a factor; the most detailed rectangulars
grouped with the detailed V-necks.

Summary of Cluster-Analysis Findings


The key findings of the cluster analyses are:
1. The major dimensions of variability are shoulder types and historical associations.
2. Within these two general dimensions the cases sort into subclusters of overall complexity; cases
with few variables fall at the cluster margins.
3. The most detailed anthropomorphs, many of the "classic" V-necks, generally do not have any
historical associations.

Table 6. Major Variable Frequency, No-Action or Historical-


Variables Analysis.

Cluster na VNCKb HNCKc ACTN HIST

la 28 23 3 7 5
lb 4(3) 1 0 1 2(1)
lc 11 3 6 10 9
Id 9 1 3 4 3
le 33 33 0 6 6
Total 1 85 61 12 28 25(2)
2a 24 (1) 0 23 (1) 7 (3) 6 (1)
2b 48 (2) 0 48 (2) 28 (7) 34 (6)
2c 12 (3) 0 11 (3) 3 (1) 10 (1)
Total 2 84 (4) 0 82 (4) 38 (11) 50 (8)
Others 16 13 0 5 5
a Total cases per cluster (DgOv-81 battle-scene cases excluded).
b Includes XNCK cases.
c Numbers in brackets are totals with DgOv-81 battle-scene cases excluded.

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCK ART ANTHROPOMORPHSIN SOUTHERN ALBERTA 403

DgOw-29 DgOw-29

DgOw-29

Dg-Ov-2

Figure 10. Representativeclassic anthropomorphs(variablescales).

4. Removal of VNCK, HNCK, ACTN, and HIST showed that other differences are often very
minor.
5. When the VNCK and HNCK variables were excluded from the clustering, many V-necks and
rectangulars grouped together, suggesting they are quite similar except for the single shoulder variable.
6. In the absence of historical-period material culture, a historical-period/prehistoric dichotomy
in anthropomorph form does not appear to exist.
7. More historical-period anthropomorphs are depicted in combat and battle, even discounting
those from the DgOv-81 Battle Scene.

A REVISED CLASSIFICATION
If the classification system used in the past is valid, we would have seen the "classic" V-necks,
"other" V-necks, and rectangulars cluster separately. We would also expect that historical and
nonhistorical anthropomorphs consistently would cluster separately if they were indeed the products
of different cultural groups. This consistent separation does not occur and as such, the differences
between these forms appear to have been overemphasized and misinterpreted. Classifying anthro-
pomorphs into distinct groups, based solely on historical, nonhistorical, V-neck, and horizontal-
neck criteria is unsatisfactory, as it does not explain the variation possible within, and the similarities

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
404 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

DgOv60 DgOv60 DgOv63

Figure 11. Simple cases resembling classics (variable scales).

between, these divisions. Enough similarities exist to suggest that all of the anthropomorphic rock
art we studied is related, showing a continuous development of representation, and a classification
scheme with historical meaning should reflect this relation.
The following classification recognizes that anthropomorph representation changed over time in
a continuous fashion. Retaining some of the terminology of previous classifications, the new classes
can be labeled classic and historic, with each of these groups containing V-neck and rectangular
anthropomorphs. Another class, called simple anthropomorphs, contains all those anthropomorphs
too simple to be placed temporally with confidence, they
though share many characteristics of both
previous classes. Note that we employ the term "class" rather than "category" to differentiate these
from Keyser's classification. Because of the relation between all anthropomorphs, this classification
is not intended to be inflexible. As with any classification system, there are exceptions and misfits.
The revised classes are based primarily on the clusters of the all-variables analysis (Figures 3 and
4, Table 4). The distribution of anthropomorphs by cluster according to the revised classes is shown
in Table 7. The frequency of variables for each of the new classes can be found in Table 8.
Classic anthropomorphs include the 23 detailed, nonhistorical cases found in Cluster l a of the all-
variables cluster analysis. Significantly, 16 of the 23 cases also clustered together in both of the
additional analyses, while four more were closely related. This indicates that most of these cases
are very similar beyond the VNCK, HNCK, ACTN, and HIST variables, and can be considered a
distinct group (Figure 10). Perhaps several of these on separate panels could be the products of a
single artisan, a possibility requiring additional investigation. Twenty of the 23 classic cases have
VNCK, two have HNCK, and one is missing the shoulder category. None have HIST and only one
has ACTN. Most have HWST, NECK, and FEET, and all but one of the cases share the BARM
variable, which is poorly represented in other clusters (see Table 7). HEAD, GNTL, and FNGR
are represented frequently. Nearly all of the heads, when present, are circles rather than just dots,
and more than half of all cases with HRTL are in this group. Most cases of Cluster 1a lack only
one or two of the variables listed above.

Table 7. Revised Class Frequency by Cluster, All-Variables Analysis.

Revised
Anthropo- Clusters
morph
Classes la b 1c 2a 2b 2c 3a 3b 4 5 Total
Classic 23 0 0 0 0 0 0 0 0 0 23
Simple 0 22 5 3 0 0 36 3 6 5 80
Historica 0 0 0 13 51 (2) 3 7 (0) 8 (1) 0 0 82 (39)
Total 23 22 5 16 51 3 43 11 6 5 185
a Numbers in brackets are totals with DgOv-81 battle-scene cases excluded.

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCKARTANTHROPOMORPHS
INSOUTHERN
ALBERTA 405

Table 8. Variable Frequency by Revised Classes.

Classic Simple Historic


N= 185 n=23 n=82 n=80
Variable 2Vi # # #
VNCK 64 20 32 12
XNCK 10 0 6 4
HNCK 94 2 38 54
SARM 46 0 24 22
BARM 47 22 9 16
CARM 6 1 3 2
HWST 116 18 63 35
NECK 139 23 58 58
HEAD 90 16 31 43
GNTL 29 14 12 3
HRTL 10 6 2 2
FEET 81 18 37 26
FNGR 33 16 9 8
BELB 10 5 4 1
ACTN 71 1 7 63
HIST 80 0 0 80
ELAB 34 10 11 13
BLEG 15 4 6 5
WEAP 51 4 5 42

As a definition, any nonhistorical, inactive anthropomorph with bent arms, a horizontal waist, a
neck, and feet, together with at least two of the variables listed above is a classic anthropomorph.
Based on this definition, a few cases in Cluster la are marginal, but a number of other cases from
several other clusters are very similar to classic anthropomorphs. In particular, the three nonhistorical
cases from Cluster 2a share many variables (see Figures 7 and 11). This points out that the variation
is continuous and forcing all of the cases into discrete classes is not always possible. Eighteen of the
V-necks in Cluster la also were considered to be "classic" V-necks in Keyser's (1977a) study. It is
important to note that a number of Keyser's (1977a) original "classic" V-necks seem to fit better
in other V-neck groups described later. However, unlike Keyser's (1 9 7 7a) classification, our definition
of classic anthropomorphs does not exclude rectangulars, two of which are found in Cluster 1a.
Visual examination of these shows that one is related obviously to the V-necks of this group. Not
surprisingly, this figure most closely resembles a classic V-neck from the same panel (Figure 10, top

^/ \ l 40cm

Figure 12. Unusual rectangular classic from DgOv-88 (Thunderbird Cave).

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
406 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

two glyphs). The second is an odd anthropomorph from Thunderbird Cave that quantitatively fits
with this group, but qualitatively seems out of place, since it has many "rib" lines and eyes, both
not commonly found at Writing-On-Stone (Figure 12). Other rectangulars that look very similar to
the classics probably clustered separately because of the presence of SARM (see Figure 11). Although
some classic anthropomorphs appear on the same panel as historical objects, they are never in
contact or obviously involved with them. This does not indicate necessarily that all classic anthro-
pomorphs were created before the introduction of historic items, but the high degree of correspon-
dence between detailed anthropomorphs and nonhistorical contexts in all three tests is strong
evidence that these represent the predominant prehistoric tradition.
Keyser (1977a) observed that "classic" V-necks were executed with greater skill and care than
other V-necks. Classic anthropomorphs are more detailed in composition than other anthropo-
morphs, and from a purely visual stance the use of the various elements appears to be precise and
highly conventionalized. Interestingly, all classics are petroglyphic in execution and many are incised
deeply. In summary, classic anthropomorphs are nonhistorical anthropomorphs usually, but not
exclusively, with a V-neck, executed as petroglyphs with considerable detail and precision.
Historic anthropomorphs include all 80 anthropomorphs that have a historical association of
any type. This class is valid because nearly all historical cases grouped together in several discrete
clusters in the all-variables analysis. Also, the presence of historical associations for these cases is
very significant in itself; anthropomorphs with historical associations can be identified with a specific
time period, and many cases are undoubtedly contemporaneous or peri-contemporaneous. They
can only have been created within a relatively short period of 100 to 150 years, mostly during
documented historic times, which increases the likelihood that all of these anthropomorphs are
related culturally. Finally, although most historic anthropomorphs show strong intragroup similar-
ities, they also share many characteristics with nonhistoric anthropomorphs, suggesting they rep-
resent a continuation of the prehistoric tradition.
Historic anthropomorphs range from simple to elaborate and include both V-necks and rectan-
gulars (Figure 13; also see Figures 5, 6, and 14). The cases of Cluster 2 are almost entirely historic,
grouping according to their shoulder variable. Thirteen historical cases of Cluster 2a are elaborate
historical versions of Keyser's (1977a) "other" V-necks; five cases are mounted. Significantly, all
but four of the historical cases of Cluster 2b are found at sites DgOv-57 and DgOv-60. Since these
are both battle scenes, we again raise the possibility that a single individual may have been responsible
for them. All but four of the 51 very simple cases of Cluster 2b have HNCK, while the rest have
XNCK. Cluster 2b contains the largest collection of historics, due to the inclusion of 29 from DgOv-
81. Their consistent grouping points out the overall similarity of these particular simple historic
cases. Many of the non-DgOv-81 battle-scene anthropomorphs are on horseback, while one X-neck
is mounted on a "boat-form" horse. By Keyser's (I 977a) definition, X-necks are a historical-period
addition to rock art, yet this case is mounted on a boat-form horse very similar to boat-form horses
associated with shield bearers, a predominantly prehistoric motif. A further 14 historic anthropo-
morphs are found in Clusters 3a and 3b. All of the anthropomorphs with shoulders are rectangular
and all but one of the cases are from the Battle-Scene panel.
The relation between historic V-necks and historic rectangulars is strong, and both historic types
are depicted mounted on horses of the same style (Figure 13). Furthermore, in the no shoulder
variables analysis, a number of historic V-necks grouped with historic rectangulars, indicating a
basic similarity beyond the shoulder variable. Overall, most variables found in classic anthropo-
morphs also are found in historic anthropomorphs, but in lower frequencies. In general historic
anthropomorphs can be divided into two groups. The first is a group of fairly detailed V-necks,
some of which are similar to classic V-necks. The second group consists of relatively simple an-
thropomorphs, including most DgOv-81 Battle-Scene anthropomorphs, simple X-necks, and mount-
ed rectangulars.
The simple anthropomorph class contains all very simple, nonhistorical anthropomorphs that
could not be confidently placed into the classic or historic classes. This class includes all the V-
necks that grouped into Cluster 1b and 1c and all of the nonhistorical rectangulars of Cluster 3a. It
also includes the simple anthropomorphs with XNCK in Cluster 4 and the extremely simple V-

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
ROCK ART ANTHROPOMORPHSIN SOUTHERN ALBERTA 407
Magne and Klassen]

DgOv 55 DgOv 60

DgOw 27
DgOv 9

Figure 13. Representative historic anthropomorphs (variable scales).

necks of Cluster 5, plus a few cases from Clusters 2a, 2b, and 3, for a total of 82 cases (Table 7;
Figure 15; see also Figures 7 and 1 1). Nearly all simple anthropomorphs grouped together (in various
combinations and orders) in all three cluster analyses, thereby justifying this group. With VNCK,
XNCK, and HNCK excluded, all but seven of the 81 cases in Clusters lb and Ic, 3a, 4, and 5
(including some historical cases) again grouped together in three clusters. Each of the new clusters
contained cases from several of the previous clusters, and their distinctions were no longer main-
tained completely. This indicates that most of these cases are very similar, and only the shoulder
shape sets them apart. With a difference of only one variable, all of these cases can be considered
to be a related group. Simple historical and simple nonhistorical anthropomorphs are very similar.
While there is no indication of purely historical associations for simple anthropomorphs, this does
not necessarily mean all these cases are nonhistoric. Several cases in Cluster 2a without historical
associations and several more cases from Cluster 3a are candidates for inclusion in the classic
anthropomorph category and did not group into Cluster l a because of minor differences (e.g., Figures
7 and 11).
Turning to the frequency data (Table 8), it is clear that all of the variables represented in classic
anthropomorphs are represented in much lower frequencies in the simple anthropomorphs. Simple

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
408 AMERICAN
ANTIQUITY [Vol. 56, No. 3, 1991

//^ ^/

15cm

/ /

// /
Figure 14. DgOv-43 battle scene.

anthropomorphs can potentially have any of the variables, but never more than a few at one time.
Overall, a definition of these simple anthropomorph with
little more depicted than a V-neck, X-neck, or straight shoulder set on parallel sides, together with
a neck and horizontal waist in many cases. With few exceptions, they either have straight arms or
no arms at all and there is no pattern to the presence of additional variables. For the most part,
simple V-necks resemble nonhistorical versions of those called "other" V-necks by Keyser (1977a).
Any very simple, nonhistorical anthropomorph with few details beyond a shoulder, neck, and waist
is considered to be a simple anthropomorph, regardless of the shoulder shape. Simple anthropo-
morphs may date from both historic and prehistoric contexts.

AnthropomorphClass Commonalities
Keyser's (1977a) categorization of anthropomorphs at Writing-On-Stone proposed that "classic"
V-necks, "other" V-necks, and rectangular anthropomorphs were essentially discrete categories with

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCKARTANTHROPOMORPHS ALBERTA
INSOUTHERN 409

'
Ay A
-

DgOv9 DgOv9 DgOv55

Figure 15. Representative simple anthropomorphs (variable scales).

significantly different cultural and temporal origins. However, our analysis indicates that all of these
anthropomorph types are closely related. Primarily, V-neck and rectangular anthropomorphs are
essentially the same style, with only a single variable distinguishing them (a point also noted by
Sundstrom [1984:114]). Furthermore, a major distinction between "classic" and "other" V-necks
is not at all apparent. Finally, V-necks and rectangulars have occurred together in rock art from
late prehistoric through historic times. Keyser (1977a) believes that "other" V-necks, historical and
nonhistorical (classified as historic and simple anthropomorphs in this paper), bear no relation to
"classic" V-necks (essentially classic anthropomorphs), as they only have the V-neck in common.
That this is not true has been demonstrated by this analysis; there is wide variation and sharing of
attributes in the composition of all V-neck anthropomorphs. One interesting exception is that classic
V-necks have no straight arms, while simple and historic V-necks have low frequencies of bent
arms. The large number of shared variables means that a stylistic relation is highly probable, and
the assertion that none are related to classic V-necks, or that some are only crude copies (Keyser
1977a:26), has little actual basis. The argument that some simple and historic V-necks were not
executed with the same skill or care used to depict classic V-necks (Keyser 1977a:26) may have
some validity, but this has not been demonstrated adequately. This difference may have a historical
explanation, as discussed later.
Rectangular anthropomorphs are not discussed at length by Keyser (1977a), so classification
comparisons are difficult to make. He does consider all rectangulars to be a heterogeneous group,
even though he admits they show a great deal of variation, including similarities to "classic" V-
necks, and he also points out that they are associated with all types of anthropomorphs (Keyser
1977a:27). He does not make an association between the variation in the rectangular anthropo-
morphs and the corresponding variation in the V-necks at Writing-On-Stone. This covariation,
although not always present, is visually apparent in many of the anthropomorphs (Figures 11-13,
15). It is also apparent from the clustering that saw both rectangulars and V-necks grouping together,
particularly in the simple- and historic-anthropomorph clusters. In the end, the only difference
between most rectangulars and V-necks is the difference in the shoulder shape. A graphic example
of this similarity is seen in two anthropomorphs at DgOw-32, both of which have a straight shoulder,
and beneath it, a V-shaped line (Figure 16).
Despite the overall similarities between classic, historic, and simple anthropomorphs, a number
of differences do exist. Classic anthropomorphs seem to be typically more elaborate and precisely
executed than other anthropomorphs. Historic anthropomorphs, on the other hand, seem to be
more involved with action and battle scenes. While current theories hold that differences in the
depiction of historic and nonhistoric anthropomorphs at Writing-On-Stone are the result of the
displacement of one cultural group by another (Bamforth 1988; Brink 1986; Dempsey 1973; Keyser
1977a, 1977b,1979), we believe that variation in anthropomorphs is better explained in terms of
temporal developments within a single artistic tradition.

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
410 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

0 10
I
CM
!

Figure 16. Anthropomorph panel combining V-neck and rectangular glyphs.

DEVELOPMENT OF ANTHROPOMORPH DEPICTIONS


ON THE NORTHERN PLAINS

The Late Prehistoric Period


Classic anthropomorphs without historical associations probably represent the oldest V-neck and
rectangular anthropomorphs found at Writing-On-Stone. This class is probably the result of hundreds
of years of artistic tradition, indicated by the strongly conventionalized depiction of the entire
anthropomorph and in particular, anatomical details such as genitals and heartlines. These anthro-
pomorphs are precisely executed symbols, perhaps reflecting a "ceremonial" function of the art, or
simply a greater amount of time available for this pursuit. Keyser (1977a) considered the original
"classic" V-necks to be late prehistoric in age, both because of their lack of historic association and
because of their apparent association with large shield-bearing anthropomorphs. The distribution
of V-neck and shield-bearing anthropomorphs is similar (Conner and Conner 1971; Keyser 1977a;
Schuster 1987; Wellman 1979); both motifs appear together on at least 16 panels at Writing-On-
Stone and on many of the same panels elsewhere on the northwestern Plains (see Conner and Conner
1971:17; Keyser 1977a:80, 1984:7; Over 1941:32; Schuster 1987:34; Sundstrom 1984:70-73; Wor-
mington and Forbis 1965:133).
Shield-bearing figures in northwestern Plains rock art probably date from the late prehistoric and
early protohistoric periods. This is known with more certainty now that Loendorf (1990) has dated
a panel of shield-bearing warriors in Montana at A.D. 1104. Large body shields, as depicted on
shield-bearing anthropomorphs, offered protection for pedestrian warriors, but were impractical for
mounted warriors and offered little protection against bullets. Native groups of the northwestern
Plains, including member tribes of the Blackfoot Nation, were known to use large body shields
(Dempsey 1973; Glover 1962), which were abandoned at some point after the horse became common
on the northwestern Plains during the protohistoric period; shield bearers are rarely depicted with
historical objects. That at least some shield-bearing figures at Writing-On-Stone are contempora-
neous with some V-necks has been discussed by several authors (Dempsey 1973; Dewdney 1964;
Habgood 1967). At one site, DgOw-32, three shield bearers, two of them mounted, have V-neck
bodies showing through shields (Figure 17), one attacking a pedestrian V-neck figure. At DgOv-2
(Panel 2), one classic V-neck has a circle drawn in front of the torso which is probably a shield,
while one simple V-neck at DgOv-3 is similarly drawn. Another V-neck at DgOw-30 (not classified
in this study but fitting the classic V-neck description) also has a possible shield, as do two unclassified
V-necks at DgOw-29. Similar V-neck figures with shields are known from Wyoming (Schuster 1987:
33) and South Dakota (Keyser 1984:32), while V-neck figures appear as designs on shields in both
Wyoming (Schuster 1987:33) and Montana (Conner 1980; Mulloy 1958:122, 127).
Many simple anthropomorphs also may have been created at the same time as classic anthro-
pomorphs and the basic differences between these two classes are minor, generally consisting of a
variation in detail. They may represent nothing more than variation in artistic abilities, personal

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCK ART ANTHROPOMORPHSIN SOUTHERN ALBERTA 411

0 10 CM JJ ^

Figure 17. Mounted shield-bearing anthropomorphs from DgOw-32. Redrawn from Keyser (1977a:Figure
14a).

styles, or even initial sketches. The simplicity of many "other" V-necks (simple anthropomorphs)
prompted Keyser (1977a:26) to call them "crude copies of classic V-neck figures," as this fitted well
with his view of separate cultural origins for these figures. However, when he encountered both
crude and detailed shield-bearing anthropomorphs at a site in South Dakota, Keyser (1984:20)
concluded that the crude figures were preliminary sketches for the detailed figures. Presumably,
separate cultural origins were not considered to be plausible. Yet, the reasons given for the prelim-
inary sketch conclusion, regarding shield-bearing anthropomorphs, are equally valid for many V-
neck anthropomorph cases. This leads to the conclusion that crude V-necks are not necessarily the
work of different artists nor later in age, but that they are related stylistically and temporally to
detailed V-necks.
From the evidence, most shield-bearing figures and classic anthropomorphs were created in a late
prehistoric context. The possibility that many simple anthropomorphs also are the same age means
that the generalization "all late prehistoric anthropomorphs are precisely and carefully executed"
cannot be made, but classic anthropomorphs certainly exhibit these characteristics. Although the
"ceremonial" or "spiritual" significance of late prehistoric art may be high, a strong "biographical"
component is also possible considering the involvement of a number of shield-bearing warriors (a
component of Keyser's Ceremonial Art) in action scenes and the difficulty of assigning action to
pedestrian figures (discussed below). Furthermore, many shield-bearing anthropomorphs may be
autobiographical, since designs on shields probably represent an individual's "medicine sign" (Con-
ner and Conner 1971:17). Even so, classic anthropomorphs were a distinct core of precise figures
within the late prehistoric that influenced the depiction of later anthropomorphs. The beginning of
the protohistoric period, however, did not see an abrupt end to the depiction of this rock art.

Late Prehistoric/Protohistoric Transition


The beginning of the protohistoric period on the northwestern Plains is marked by the first
appearance of European trade items, especially guns and horses. The persistence of shield-bearing
figures into at least early protohistoric rock art is evident at a number of sites at Writing-On-Stone.
For example, one shield bearer is closely associated with a rifle at DgOw-29, while another at DgOv-
3 may also be holding a rifle. Rifles also are associated with shield bearers at two sites in Montana
(Keyser 1977a), while mounted shield bearers occur in Montana (Conner and Conner 1971; Conner
1980) and in northwestern South Dakota (Keyser 1984:34, 49), as well as at Writing-On-Stone. For
example, three shield-bearing figures at DgOw-32 are mounted on "crudely-drawn boat-shaped
horses" (Keyser 1977a:23), with two of them attacking pedestrian shield bearers (Figure 17). Ac-
cording to Keyser (1977a:23), the "awkwardness of these depictions suggests that the artists had

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
412 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

not mastered the style of drawing mounted men, and the crude boat-shaped horses suggest that
these animals were still a novelty in Plains Indian culture." Two of the horses at DgOw-32 also
appear to be wearing horse armor, which apparently was abandoned soon after the introduction of
horses (Dempsey 1973; Keyser 1977a).
Considering that shield bearers continued to be depicted into the protohistoric period, it is not
surprising that a continuation of classic motifs can be seen in the historic period. In particular, five
cases from Cluster 2a of the all-variables analysis grouped with the classics in the no ACTN and
HIST analysis. For example, one case with bent arms, a circular head and a heartline resembles a
classic V-neck in many ways (Figure 13). This case is crudely mounted on a "boat-form" horse,
perhaps again representing an early attempt at depicting a mounted rider, similar to the mounted
shield bearers discussed above. Close examination of this case reveals that this horse is far too
incomplete to be conclusively considered a boat-form version, and the form of the thighs is actually
reminiscent of probable late historic horses. Even so, this mounted V-neck seems to represent a
historic continuation of the classic tradition.
A historic X-neck from DgOv-43 is mounted on a horse that is more boat-form in appearance
(Figure 14). This is an incongruous combination according to Keyser's (1977a) classification, as he
considers X-necks to be a historical-period category created by a different cultural group than that
which created protohistoric boat-form horses. This X-neck, however, is involved in a battle with
two historic V-necks that clustered in Cluster 2a. Indeed, many of the detailed historic cases of
Cluster 2a resemble classic anthropomorphs, which may indicate they are all protohistoric. Another
unclassified X-neck from DgOw-30 also appears to be mounted on a crude boat-form horse. If the
boat-form horse is indeed a transitional horse style, this means that at least two X-necks, and
possibly others, were created in an early protohistoric context, at the same time as mounted shield
bearers and detailed V-necks. Thus quite probably all of these figures were created by the same
cultural group.

Protohistoric and Historic Periods


Most historic anthropomorphs are simpler in design than classic ones and exhibit differences in
their form of depiction that we believe are due to major changes within a single culture. The argument
here is that historic anthropomorphs evolved from classic anthropomorphs during the protohistoric
period, and continued to be depicted into the historic period. Many historic anthropomorphs are
simple and are involved in obvious action situations such as battles and riding horses. They depict
few anatomical details such as fingers, heartlines, and genitals, and many do not even have heads.
Elaborations, if they exist, are often headdresses, and many weapons are guns. Overall, the precision
and care with which historic anthropomorphs were created seems to have declined from that of
classic anthropomorphs, though this is a generalization. Again, many simple anthropomorphs show
great similarity to historic anthropomorphs, and it seems probable that many simple anthropo-
morphs were created in a historical context. More historic anthropomorphs are depicted with action
than nonhistorical anthropomorphs, and some of this difference may be the result of a cataloging
bias in our analyses, but it is also clear that the horse brought a great deal more mobility to the
people of the Plains. It could be argued that an anthropomorph depicted on horseback (a historical
context) does not inherently show more "action" than one holding a bow or a spear (nonhistorical
context), but we assumed it does; both may not necessarily represent activity. Indeed, many mounted
anthropomorphs could be considered quite static in form. At least eight anthropomorphs in battle
situations have no obvious historical associations, demonstrating that action is not restricted to
historical contexts. However, associations between anthropomorphs and action appear to have
increased in frequency as the culture evolved in the protohistoric and historic periods, with much
of this action involving horses. Any increase in rock-art action can thus be explained as the result
of depicting increased mobility afforded by horses, and a corresponding increase in horse-related
events, such as horse stealing, coup counting, and combat.
Many authors recognize that great changes in Native culture resulted from the introduction of
the horse and the gun to the northwestern Plains. Within a few generations patterns of movement,
hunting, and warfare were vastly altered. Ranges expanded and contact (and conflict) with other

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCK ART ANTHROPOMORPHSIN SOUTHERN ALBERTA 413

Native groups undoubtedly increased in frequency. The horse became central to the Plains Native
Indian way of life, while warfare and the status system were modified substantially. Keyser (1979:
46) suggests that "prior to the historical period, status rested primarily on power acquired through
the vision quest rather than by a system of graded war honors" and goes to some length to point
out evidence of these changes at Writing-On-Stone. Yet he still maintains that differences in the
rock art are the result of two different cultural groups. By not linking these two arguments, Keyser
(1979) fails to make explicit the full implications of his research; that is, that the changes observed
in the historical rock art probably resulted from changes within a single culture, brought on following
the introduction of the horse and the gun, rather than the appearance of a new culture. From our
reading of Wissler (1911), we suggest that the vision quest was just one, albeit essential, way of
achieving status and power in prehistoric times. Indeed, it was a crucial step in being able to acquire
status in yet other ways, including hunting or raiding success. Furthermore, the number of individuals
achieving higher (male) status was probably lower in prehistoric times, and once the horse and gun
accelerated aggressiveness, more individuals, perhaps with less experience with a rock-art tradition,
were achieving status and could have been "leaving their marks" at Writing-On-Stone. This alone
could account for the high number of simple anthropomorphs and their apparent strong similarity
to historic anthropomorphs. The rock art of Writing-On-Stone, as a form of historical record keeping,
reflected the changes occurring within the culture. Traditional "spiritual" depictions of heartlines
and genitals apparently declined as the importance of coup and horses increased. Perhaps the rock
art no longer had spiritually oriented, introspective connotations and instead represented more
prosaic biographical events, as suggested by Keyser's (1 977a, 1979) Biographical Art category, which
he argues eventually developed a strictly storytelling function (Keyser 1987). This complete lack
of ceremonial function is not probable, considering the spiritual nature of Plains Native culture to
this day.
Regardless of the motivation behind the rock art, the changes observed do not establish a priori
that a cultural displacement occurred. An alteration in meaning or motivation in the rock art could
just as easily be the result of a within-group response to changing cultural conditions; the recurrent
depiction of horses, guns and battles was a response to a modified cultural value system. As the
emphasis shifted to a horse culture, certain material items were lost, new items were added, and
this was reflected in the rock art. For example, body shields became cumbersome and were aban-
doned, while trailing headdresses, depicted in historical-period rock art, may have been adopted
due to their spectacular effect during a high-speed attack. The greatly increased pace of Plains life
may have resulted in a corresponding decrease in the time spent on creating rock art. While some
historic anthropomorphs are deeply incised and carefully rendered, a large component is more lightly
scratched. Perhaps the emphasis changed from detailed representations to simple recordings as the
culture altered ceremonial and status values. Consequently the rock art was quickly executed by
greater numbers of less-skilled artists and became proletarianized. Even so, the presence of precise
and detailed historical-period anthropomorphs means that a universal change in precision did not
occur. Furthermore, continuation of many conventions such as the V-neck displays a considerable
continuity in style if not technique. In general then, we view all anthropomorph motifs as very
strongly related. In a chronological context, these relations are illustrated schematically in Figure
18.
One subject not yet addressed in detail is the presence of hourglass figures at Writing-On-Stone.
Hourglass anthropomorphs have been considered a historic addition to Writing-On-Stone rock art
(Keyser 1977a) and form part of Keyser's argument for the replacement of cultural groups in that
area. Reexamination of the evidence suggests that this may not be the case; in fact, hourglass
anthropomorphs may have had a minor but continuous presence in the rock art from the late
prehistoric period onwards. As many as 100 hourglass anthropomorphs are found at over 21 separate
sites at Writing-On-Stone. About 65 of these are found on the Battle-Scene panel (Figure 5). Hourglass
figures are generally simple, although a few have arms and feet, and some have additional elaborations
such as headdresses, fringes, and t-shaped heads. Of the approximately 35 hourglass anthropomorphs
outside of the Battle-Scene panel, only about 13, or less than 38 percent, have any historical
association or show any action, suggesting many are not necessarily historical-period in age. One
fringed hourglass anthropomorph at DgOw-29 appears to be closely associated with several classic

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
414 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

c Shield-bearing

I Classic
OL

0 Historic
0
a.
O Simple 1 111111111111111
11111111111111111111111111111
11 1 1111i
<
Hourglass
111111111llllllll 1111
111111111111
Ulullll lm
LATEPREHISTORIC PROTOHISTORIC HISTORIC
I I
ca. A. D. 1700 ca. A. D. 1800

TEMPORALDISTRIBUTION

I Certainor HighlyProbable ||| Probable E Uncertain


Figure 18. Chronology of anthropomorph-class developments. Although not redefined in this study, shield-
bearer and hourglass classes are shown to provide a more inclusive context.

anthropomorphs and boat-form zoomorphs. Another hourglass figure not recorded by Keyser is
closely associated with a shield bearer, and both figures on this panel are isolated from other rock
art. Although inconclusive, these associations and the number of nonhistorical hourglass anthro-
pomorphs are suggestive of a long-term presence at Writing-On-Stone. Thus the presence of nu-
merous hourglass figures at DgOv-81 is unusual but not in itself evidence of the intrusion of a new
cultural group. Indeed, one observation that could be pursued is that the hourglass figures at DgOv-
81 are mostly located to the rear of each opposing force, while the rectangular figures appear in the
central battle area. Is differential status being depicted? Or perhaps different genders are represented,
with women at the rear guarding horses and travois of the right (offensive) force, and those of the
defensive force in tipis with children. The absence of V-neck anthropomorphs at this panel may
only indicate the personal preference of the artist, since the scene probably was executed by a single
individual. However, most differences between the historic anthropomorphs of this panel and earlier
classic anthropomorphs can be explained in terms of a rapid and far-reaching alteration of Plains
Native culture during the historic period. The entire complex of historical changes to Writing-On-
Stone rock-art anthropomorphs culminated in the Battle-Scene panel. It is probably one of the last
petroglyph panels created at Writing-On-Stone.

Implications for Cultural Affiliation


The anthropomorph motifs examined in this paper represent continuous rock-art traditions. The
majority of the variation depicted is the result of chronological change and development of motifs,
reflecting ongoing changes in Plains culture. In themselves, V-necks and rectangular bodies are not
indicative of age or cultural association. Classic, simple, and historic anthropomorphs all show basic
similarities, and are undoubtedly related stylistically. Although they do show some differences, at
no point is it necessary to explain these differences in terms of a break in cultural affiliation. Rather,
the differences are a case of continuous variation through time. The overall relation of these an-
thropomorphs contradicts the traditional view that all late prehistoric rock art at Writing-On-Stone
is only affiliated with the Shoshone, while protohistoric/historical-period rock art is affiliated with
one or several different groups, such as the Blackfoot, Gros Ventre, Cree, or Assiniboine (Dempsey
1973:19-22; Habgood 1967; Keyser 1977a, 1977b, 1979). This view is critical to debates regarding

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCKARTANTHROPOMORPHS
INSOUTHERN
ALBERTA 415

late prehistoric Native group distributions authored by Brink (1986) and Bamforth (1988), among
others. The presence of related V-neck and rectangular anthropomorphs from the late prehistoric
through the historical-period suggests that the artists responsible for at least these types of anthro-
pomorphs were present in the Writing-On-Stone area during this entire time period. This does not
rule out the possibility that several different groups, including the Shoshone, may have created
various rock-art motifs at Writing-On-Stone at various times. It does, however, reduce the likelihood
that the Shoshone were responsible for all, or even any, late prehistoric rock art, particularly
anthropomorphs. This inference is even more likely given the A.D. 1100 date assignable to shield
bearers in south-central Montana, well before Shoshonean presence in that area (Loendorf 1990).
With respect to this latter recent evidence, we must point out that Loendorf s (1990:52) alternative
(and tentative) ascription of these A.D. 1100 figures to possible Athapaskan migrants is even more
dubious than a Shoshone one, although he considers it "quite" possible that shield bearers in
Montana dating between A.D. 1450 and 1750 were created by "more than one cultural group"
(Loendorf 1990:52). Northern Athapaskan culture areas have very little in the way of a rock-art
tradition and certainly no shield-bearing warrior antecedents. Leaving this major issue aside for
now, there is no question in our minds that the assignment of Athapaskans to the Avonlea "culture"
of the northern plains (Kehoe 1973; Wilcox 1988) is highly debatable, and the archaeological
evidence, including rock art, needs to be extensively reconsidered (see also Ives 1990:49-52; Magne
1987).
The ethnohistoric evidence for a Shoshonean presence in southern Alberta is based largely on
various references to the "Snake" Indians in historical accounts (Brink 1986). As Brink (1986:45)
notes, many authors "implicitly" accept that the Snake are actually the Shoshone, and this notion
has become entrenched in northwestern Plains archaeological thought (e.g., see Bamforth 1988:88-
89). The archaeological evidence for the presence of the Shoshone in southern Alberta rests almost
exclusively on Keyser's (1975, 1977a) contention that the shield-bearing anthropomorph motif is
derived from the Great Basin (Bamforth 1988:89), the original home of the Shoshone. Any other
convincing archaeological evidence for a Shoshone presence in southern Alberta is lacking (Byrne
1973:519-521; Greiser 1988; Vickers 1988) and the archival record, Brink's (1986) principal source,
is quite ambiguous as well. Indeed, there is now considerable archaeological and ethnohistorical
evidence accumulating for a Siouan identification for the Snake in southern Alberta (Byrne 1973:
520-521; Vickers 1988:25).
Even without this lack of evidence, the Shoshone are unlikely candidates for the artists responsible
for historical-period anthropomorphs at Writing-On-Stone due to their very limited use of the
Alberta plains during the protohistoric and historic periods (Bamforth 1988:91-93; Brink 1986;
Magne et al. 1987). This point is acknowledged explicitly by Keyser (1977a; 1979); hence, the need
for a cultural schism between late prehistoric and historic rock art in his classification. But if the
Shoshone were not responsible for the historical-period rock art at Writing-On-Stone and if all of
the temporal ranges of the art are closely related, then the Shoshone are responsible for none or
very little. For all of this rock art, we must look elsewhere for its creators. Directly related to this
issue are rejections of Shoshonean origin for incised shield-bearing and V-neck anthropomorphs
for sites in eastern Wyoming (Schuster 1987), southwestern South Dakota (Sundstrom 1984:108),
and Montana (Conner and Conner 1971). We agree with Schuster (1987) and Sundstrom (1984)
that the shield-bearing motif may have diffused from the Great Basin in one form or another, but
this does not lead to the conclusion that Shoshone artists must have created them all. Curiously,
the complete lack of V-necks west of the Continental Divide (Gebhard 1966; Schuster 1987) implies
that this motif could not have originated with the Shoshone in the Great Basin, but must have
originated on the northwestern Plains.
V-neck and shield-bearing anthropomorphs on historical-period hide paintings (Gebhard 1966;
Schuster 1987:35-36) prove without a doubt that tribes other than the Shoshone were creating these
motifs. If, as seems certain from historical evidence, tribes other than the Shoshone created rock
art and other forms of depictional art, and since we know that other tribes, including the Blackfoot,
were using body shields in the late prehistoric, then there is no valid reason why these tribes could
not have been depicting shield bearers in their rock art. Given this, it is not implausible that the

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
416 AMERICANANTIQUITY [Vol. 56, No. 3, 1991

incised shield-bearing anthropomorph rock-art motif also originated on the Plains, or could at least
have been depicted by tribes other than the Shoshone. For example, although Gebhard (1966)
believes the shield motif originated in the Great Basin, he also believes that this motif diffused to
other areas and was incorporated into the depictional art of various Plains tribes. The Shoshone
did not have a monopoly on the design.

CONCLUSIONS
Cluster analysis cannot be expected to account for all variation encountered in rock art, but the
presence of repeated conventionalized design elements in Writing-On-Stone anthropomorphs has
allowed use of this method for determining general patterns of association. At the very least, cluster
analysis introduces a systematic approach to classifying rock art at Writing-On-Stone and dem-
onstrates that three principal misconceptions regarding classification of the anthropomorphs there
have been applied previously. These misconceptions are: (1) Not all V-necks are related; (2) V-neck
and rectangular anthropomorphs are unrelated; and (3) nonhistoric and historic anthropomorphs
are unrelated. In fact, all V-necks share a number of characteristics and show a continuous degree
of variation. The difference between V-neck and rectangular bodies rests primarily on a single
variable-shoulder shape-which does not justify their separation. Finally, the differences between
nonhistoric and historic anthropomorphs can be explained in terms of a temporal development in
depiction, rather than as two separate traditions.
With these misconceptions in mind, the traditional classification of "classic" V-necks, "other"
V-necks, and rectangulars lacks validity. A new classification that recognizes stylistic relations and
the historical evolution of anthropomorphs is necessary. In this system, classic anthropomorphs
are late prehistoric V-necks and rectangular figures that are detailed and carefully executed and
represent a highly conventionalized stylistic tradition. This tradition evolved further in the proto-
historic and historic periods to reflect changes in the Plains Native Indian culture resulting from
the introduction of the horse, as seen in historic anthropomorphs. More emphasis was placed on
action and event depiction, while attention paid to detail and precision declined. Yet historic
both V-neck
bothropomorphs,
anthropomorphs, an gular bodied, retained many variables in common with
classic anthropomorphs and represent a continuation of many design elements. Simple anthropo-
morphs are cases that share characteristics with both classic and historic anthropomorphs and are
probably stylistically and temporally related to both of these groups.
There is much yet to be done before a reasonably complete understanding of the Writing-On-
Stone rock art can be achieved. Finding analytically appropriate classifications and applying them
to cultural, historical, and geographic problems is only one issue. The hints of individuality perceived
in these analyses are one obvious area in need of examination, particularly in light of recent studies
of Plains ledger art and hide paintings (Feder 1980; Young 1986). Implementation of some of these
concepts in an analysis of intersite variations, similarities, and correspondences could be highly
informative. We still do not know much about what this rock art meant to the Native creators, and
we believe that terms such as "ceremonial" or "biographic" should be employed carefully until it
is clear that such distinctions actually existed in the past.
The overall relation of the Writing-On-Stone anthropomorphs studied in this paper suggests a
single culture was responsible for all of the rock art, or that several groups continuously depicted
rock art from the late prehistoric until the historic periods. Previously, it was thought that all late
prehistoric rock art in the area was created by Shoshonean artists, while historic rock art was left
by the Blackfoot or other groups. As the Shoshone did not have a major historical presence in
southern Alberta, it is unlikely this group was responsible for Writing-On-Stone rock art. This
suggestion must be confirmed with further archaeological, ethnographic, historical, and rock-art
investigations.

Acknowledgments. We would like to thank but certainly not hold responsible,Alwynne Beaudoin, Jack
Brink, Maurice Lanteigne,and Rod Vickers for commenting on previous drafts. Formal reviews by James
Keyser, Lawrence Loendorf, and one anonymous reviewer were very helpful. R. G. Matson provided feedback
concerning the multivariate analyses, Lawrence Halmrast and Bob Ward provided assistance in the field, and

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
Magne and Klassen] ROCK ART ANTHROPOMORPHSIN SOUTHERN ALBERTA 417

Mark Rasmussen helped this research get started. Figures 5 and 6 were reproduced directly from polyethylene
tracings made by James Keyser in 1976, now housed at the Archaeological Survey of Alberta. Figures 1 and 17
were drawn by Wendy Johnson. Eric Damkjar and Karie Hardie also assisted with figures. The Spanish translation
was provided by Naldo Lombardi via the Translation Bureau of Federal and Intergovernmental Affairs, Alberta
Government. This research has been supported by Alberta Culture and Multiculturalism and Alberta Recreation
and Parks.

REFERENCES CITED

Bamforth, D. B.
1988 Ecology and Human Organization on the Great Plains. Plenum Press, New York.
Brink, J.
1979 Archaeological Investigations at Writing-On-Stone. Occasional Paper No. 12. Archaeological Survey
of Alberta, Edmonton.
1986 Dog Days in Southern Alberta. Occasional Paper No. 28. Archaeological Survey of Alberta, Edmonton.
Byrne, W. J.
1973 The Archaeology and Prehistory of Southern Alberta as Reflected by Ceramics, Volume Two. Mercury
Series Paper No. 14. National Museum of Man, Archaeological Survey of Canada, Ottawa.
Castleton, K. B.
1978 The East and Northeast, Petroglyphs and Pictographs of Utah, vol. 1. Utah Museum of Natural History,
Salt Lake City.
Conner, S.
1980 Historic Period Indicators in the Rock Art of the Yellowstone. Archaeology in Montana 21(2): 1-13.
Conner, S., and B. L. Conner
1971 Rock Art of the Montana High Plains. The Art Galleries, University of California, Santa Barbara.
Dempsey, H.
1973 A History of Writing-On-Stone. Ms. on file, Alberta Recreation and Parks and the Archaeological
Survey of Alberta, Edmonton, Alberta.
Dewdney, S.
1964 Writings on Stone Along the Milk River. The Beaver Outfit 295 (winter):22-29.
Feder, N.
1980 Plains Pictographic Painting and Quilled Rosettes. A Clue to Tribal Identification. American Indian
Art 5(2):54-62.
Fox, D. J., and K. E. Guire
1976 Documentation for MIDAS. Statistical Research Laboratory, University of Michigan, Ann Arbor.
Gebhard, D.
1966 The Shield Motif in Plains Rock Art. American Antiquity 31:721-732.
Glover, R. (editor)
1962 David Thompson's Narrative 1784-1812. The Champlain Society, Toronto.
Greiser, S. T.
1988 Late Prehistoric Cultures On the Montana Plains. In Indians of the Great Plains, A.D. 500-1500,
edited by K. H. Schlesier, in preparation. Ms. in possession of authors.
Habgood, T.
1967 Petroglyphs and Pictographs in Alberta. Archaeological Society of Alberta, Newsletter Nos. 13 and 14:
1-40.
Ives, J. W.
1990 A Theory of Northern Athapaskan Prehistory. Westview Press, Boulder.
Jones, T., and L. Jones
1982 The St. Victor Petroglyphs: Description and Condition Report. Ms. on file, Saskatchewan Museum of
Natural History, Regina.
Kehoe, T. F.
1973 The Gull Lake Site: A Prehistoric Bison Drive in Southwestern Saskatchewan. Publications in Anthro-
pology and History No. 1. Milwaukee Public Museum, Milwaukee.
Keyser, J. D.
1975 A Shoshonean Origin for the Plains Shield-Bearing Warrior Motif. Plains Anthropologist 20:207-215.
1977a Writing-On-Stone: Rock Art on the Northwestern Plains. Canadian Journal of Archaeology 1:15-80.
1977b The Rock Art of Writing-On-Stone. Ms. on file, Alberta Recreation and Parks and the Archaeological
Survey of Alberta, Edmonton.
1979 The Plains Indian War Complex and the Rock Art of Writing-On-Stone, Alberta, Canada. Journal of
Field Archaeology 6:41-48.
1984 Section 1: The North Cave Hills. In The Rock Art of Western South Dakota, pp. 1-51. Special
Publication No. 9. South Dakota Archaeological Society, Sioux Falls.
1987 A Lexicon for Historic Plains Indian Rock Art: Increasing Interpretive Potential. Plains Anthropologist
32:43-71.

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
418 AMERICAN
ANTIQUITY [Vol. 56, No. 3, 1991

Keyser, J. D., and G. C. Knight


1976 The Rock Art of Western Montana. Plains Anthropologist 21:1-12.
Leechman, D., M. Hess, and R. L. Fowler
1955 Pictographs in Southwestern Alberta. Annual Report 1953-1954 Bulletin No. 136:36-53. National
Museum of Canada, Ottawa.
Loendorf, L.
1990 A Dated Rock Art Panel of Shield Bearing Warriors in South Central Montana. Plains Anthropologist
35:45-54.
Magne, M.
1987 Plains and Plateau Athapaskan Movements in Late Prehistoric and Early Historic Times. Paper pre-
sented at the 1987 Meeting of the Canadian Archaeological Association, Calgary.
Magne, M., and Contributors to the Saskatchewan-Alberta Dialogue
1987 Distributions of Native Groups in Western Canada, A.D. 1700 to A.D. 1850. In Archaeology in Alberta
1986, edited by M. Magne, pp. 220-232. Occasional Paper No. 31. Archaeological Survey of Alberta,
Edmonton.
Mulloy, W.
1958 A Preliminary Historical Outline for the Northwestern Plains. University of Wyoming Publications No.
22(1). Laramie.
Over, W. H.
1941 Indian Picture Writing in South Dakota. Archaeological Studies, Circular IV. University of South
Dakota Museum, Vermilion.
Schaafsma, P.
1971 Rock Art of Utah. Papers of the Peabody Museum of Archaeology and Ethnology Vol. 65. Harvard
University, Cambridge.
1972 Rock Art in New Mexico. University of New Mexico Press, Albuquerque.
Schuster, H. H.
1987 Tribal Identification of Wyoming Rock Art: Some Problematic Considerations. Archaeology in Mon-
tana 28(2):25-43.
Shennan, S.
1988 Quantifying Archaeology. Academic Press, San Diego.
Sundstrom, L.
1984 Section 2: The Southern Black Hills. In The Rock Art of Western South Dakota, pp. 53-142. Special
Publication No. 9. South Dakota Archaeological Society, Sioux Falls.
Vickers, J. R.
1988 Cultures of the Northwestern Plains: From the Boreal Forest Edge to Milk River. In Indians of the
Great Plains, A.D. 500-1500, edited by K. H. Schlesier, in preparation. Ms. in possession of authors.
Wedel, W. R.
1969 A Shield and Spear Petroglyph from Central Kansas: Some Possible Implications. Plains Anthropologist
14:125-129.
Wellman, K.
1979 A Survey of North American Indian Rock Art. Akademische Druck-u. Verlagsanstalt. Graz, Austria.
Wilcox, D. R.
1988 Avonlea and Southern Athapaskan Migrations. In Avonlea Yesterday and Today: Archaeology and
Prehistory, edited by L. B. Davis, pp. 273-280. Saskatoon Archaeological Society, Saskatoon.
Wissler, C.
1911 Social Organization and Ritualistic Ceremonies of the Blackfoot Indians. Anthropological Papers Vol.
7, Pt. 1, pp. 1-64, American Museum of Natural History, New York.
Wormington, H. A., and R. G. Forbis
1965 An Introduction to the Archaeology of Alberta, Canada. Proceedings No. 11. Denver Museum of Natural
History, Denver.
Young, G. A.
1986 Aesthetic Archives: The Visual Language of Plains Ledger Art. In The Arts of the North American
Indian, edited by E. L. Wade, pp. 45-62. Hudson Hills Press, New York.

Received January 30, 1990; accepted March 1, 1991

This content downloaded on Sun, 3 Feb 2013 13:09:04 PM


All use subject to JSTOR Terms and Conditions
View publication stats

You might also like