Protohistoric arrowhead variability

Protohistoric Arrow Head Variability in the Greater Southwest

Mark Harlan

The General Problem

Southwestern archaeologists and ethnohistorians have tended to view the period between the terminus of the “great cultures” (Anasazi,
Mogollon and Hohokam) and the arrival of the Spanish expeditions in the region as a kind of Dark Age. They take at face value the early
explorers’ description of a despoblado, with a narrow band of settlements along the Rio Grande, the islands of Hopi and Zuni and the O’odam
groups in Arizona standing in stark contrast to an empty wasteland.

Regional archaeology has appeared to confirm the Spaniards’ observations. For three generations of scholarship, archaeologists have focused
on major ruins and developed a view of the Southwest based on their rise and fall. While they have recognized lesser manifestations, especially
the many limited artifact scatters that often lack diagnostics, they have generally assumed these sites to be Archaic on the basis of one or two
diagnostic points or relegated them to an amorphous “unknown lithic scatter” category if no diagnostics are found. The dominant view has been
that most or all such locations represent a hunting and gathering lifeway that was superseded by the food-producing, pueblo- and
mound-building societies that rose to replace them. To the extent that these artifact scatters clearly appear to postdate the Archaic, they are
seen as limited use sites left by logistical expeditions coming out of the agricultural settlements they surrounded. Sites lacking diagnostic points
are usually ignored by those not interested in the Archaic.

A recent change of views (Seymour 2002, 2004a, 2004b, 2006, 2007, In Press A, In Press B, Harlan and Seymour 2005) clearly demonstrates that
the despoblado is a misconception, resulting from the failure of both the Spanish explorers and most archaeologists to perceive the mobile
peoples that filled areas empty of settlements. This paper is not the place to address critical questions concerning the ethnic identity of these
nomads or the temporal depth of their presence. The purpose here is, rather, to examine the range of variability of one of their key artifact
types, arrowheads. Protohistoric arrowheads comprise a distinctive artifact category; one of only a few lithic items providing evidence for a
mobile group presence on sites that might otherwise be attributed to Archaic peoples or to hunting parties from prehistoric farming settlements.

Despite controversy over whether smaller points are invariably associated with use of the bow and arrow while larger points indicate use of the
atlatl and dart (Butler 1975, Mallouf 1987:57, Ahler 1971:107, Flenniken 1989), the general association of points weighing less than 3.5 grams with
arrows and longer points with darts is abundantly clear (Justice 2002:15-16), assertions to the contrary (Browne 1938:359; Evans 1957:83) not
withstanding. It is also well demonstrated that arrows began to supplant darts or javelins in North America, including the Southwest, by about
A.D. 500 but that proliferation of arrowheads in both numbers and variety postdates A.D. 1300 (Justice 2002:44-45; Bamford 1986:42-44). This later
period, which may mark the beginnings of the protohistoric, also saw development of distinct regional and sub-regional variability in arrowhead
forms. The present study seeks to better understand that variability and suggest approaches to exploiting its distribution. The focus here is on
methodology and on a characterization of the variability but the ultimate goal is to achieve an improved picture of the protohistoric groups who
occupied the southern Southwest.

The Bow and Arrow Weapon System

Before concentrating on the formal variability of a collection of arrowheads from the southern Southwest, it is useful to remind ourselves that
arrowheads were actually a relatively minor part of a complex weapons system and that inferences based on their form are rather like
reconstructions of extinct shark species from their teeth alone, without the relative certainty genetic descent provides when dealing with
biological species. In both cases, the vagaries of preservation force us to deal with what time has left for us, not with the far more important parts
of a complex system that has vanished.

Viewed over the long term, the change-over from atlatl and dart to bow and arrow is not just a stylistic shift providing useful temporal
information. It marks a set of fundamental improvements in weapons system efficiency (using this term in the sense of Bleed 1986 739-740, but
recognizing that the bow and arrow system is both maintainable and reliable; see also Nelson 1991:67). The progression has continued, first with
development of the crossbow and then with improvements in the firearms that have gradually replaced all other projectile weapons in both
hunting and warfare. At each step, the systems are characterized by smaller projectiles that travel at higher velocities and allow those adept in
their use to consistently hit and damage targets at greater distances (Higgins 1978:100; Bergman 1983:239). Whether employed in hunting or in
warfare, the advantages of this capability are obvious.
A further advantage of this trend in projectile technology may be less readily apparent. At each stage, the learning curve for employment of the
technologies has been reduced. We have little historical or ethnographic information on the relative skill sets required for atlatl and dart versus
bow and arrow, but anecdotal evidence seems to indicate that the learning curve for the bow is less steep, with a better chance for more people
to achieve proficiency. Still, Medieval armies that depended on massed archers had behind them systems of learning that favored life-long
devotion to archery practice, a critical element in their success (Pope 1918:103, 122-125; Browne 1940:212).

The trend toward a reduced learning curve has extended beyond the use of spears, bows and arrows and into modern times. In contrast to
yeomen who spent their lives in archery practice, the musketeers who served in later armies could be recruited and trained on a more or less ad
hoc basis (Pope 1918:103; Hackett 1983:25-33). This is not just a function of the inherent inaccuracy of their weapons and employment of mass
formations to compensate for it. The riflemen who replaced musketeers in the Nineteenth century got more training in drill formations than in
actual use of their weapons and even modern conscript armies have generally found that a couple of weeks spent on the firing range is sufficient
for training most recruits, followed up with a day or two of refresher each year. Since the basic requirement for all of these weapons systems is
to accurately place a projectile on target at considerable distance, the information from warfare probably extends to hunting, with the caveat
that the skill needed to find the game far exceeds the skill needed to hit it.

The replacement of personal skill by technological advancement might be referred to as the “kung fu” effect. Training a human being to fight
effectively empty-handed or without projectile weapons takes a lifetime of training, with daily attention to the art. Achieving efficiency with a
spear, even with the atlatl to assist, also appears to be extremely difficult, on the basis of the one experiment I have found referenced in the
literature (Browne 1940:212). However, groups in the Southwest seem to have seen no need for the shorter learning curve and increased range of
the bow and arrow until late in the prehistoric sequence. Since the mechanical function of the bow and arrow system seems quite obvious and
its use was clearly known by no later than AD 500 (Justice 2002:44), we may infer that increased importance of the bow and arrow technology in
the protohistoric period responded to an emerging need. Unfortunately actually elucidating this point would call for a massive distributional
study that is beyond the scope of this discussion, perhaps even exceeding the capacity of available datasets.

Given the likely importance of the bow and arrow weapons system for mobile groups in the Protohistoric period, it is unfortunate that, in
practice, we actually examine its least important component. In order of importance, the system consists of the bow (itself a compound tool
composed of stave and string) the arrow shaft (another compound tool consisting of required main shaft, optional fore shaft and optional
fletching) and finally a sharp tip, which pre-contact Native Americans could make from stone but also from bone or wood. The tip may also be
provided merely by sharpening the fore shaft. It should also be clear from the brief discussion of the kung-fu effect that the archer is actually the
most critical component of the system and the archer’s skill can largely compensate for shortcomings in the other system components, making
the specific form of any one arrowhead even less critical to system performance.
The sharp arrowhead has two basic functions – to assist the projectile in penetrating the target and to cut a large wound that will cause
hemorrhage (Browne 1940:209; Flenniken and Wilkie 1989:152; Cheshier and Kelley 2006:353). If the arrow is intended to deliver a dose of poison,
penetration alone is sufficient since poison is most often applied to the foreshaft but, for most uses, the design of the arrowhead is a compromise
between the depth and the size of the wound it causes, a fundamental distinction that Medieval archers recognized in designating their points as
broadheads or bodkins.

Broadheads and bodkins share design parameters at the front end and both types of arrowheads start with a sharp point. From there back,
design optimization can result in compromises. Penetration is maximized by keeping the surface that the point presents to the target at or near
the diameter of the arrow shaft or foreshaft, hence the relatively common practice of simply sharpening the foreshaft, an approach that works
well for small game.1 However, for larger game and human beings, the small wounds caused by this type of arrow are not immediately
debilitating, although infection may cause death in the long run and a perfectly placed bodkin, like the one said to have felled King Harold at the
Battle of Hastings, is something to be feared, even by animals with thick hides and people wearing armor.

Broadheads are more effective when penetration is not the sole concern because they are, in essence, knife blades delivered at the end of a
wooden shaft and so will cause extensive bleeding. Further sacrificing penetrating power, barbing of the arrowhead will tend to keep it in the
wound so that the mechanical effect of a moving arrow shaft can exacerbate the hemorrhage, unless the target becomes immediately still
(Browne 1940). A simple triangular point provides some barbing, but such points can be difficult to attach firmly to the shaft. Various types of
projections created by side notching or other shaping provide good anchor points for tying the point to the shaft while simultaneously barbing the
point. For hunting, a good attachment of the point to the shaft (or at least the fore shaft) is an advantage because movement as the prey runs will
enhance hemorrhaging (Flenniken and Wilkie 1989:151). For warfare, it may actually be more advantageous if the point detaches in the wound,
leading to a debilitating infection or death.

As with many mechanical devices, arrowheads can fulfill their physical function while assuming a range of actual shapes. The basic
requirements for penetrating a target or causing a large wound may tightly constrain the front end of the point’s shape, but shaping the haft end
is less constrained by physical requirements and more open to personal preference or learning traditions. The analysis discussed here has been
designed to consider both physical design constraints and personal or cultural preferences as sources of variability. A brief survey of the
literature on projectile point variability indicates that this will constitute something of a departure from past practice.

Approaches to the Study of Arrowhead Variability

A survey of the literature has found that the majority of studies addressing projectile point variability are confined to dart points (Ahler 1971;
Bamford 1986; Benfer 1970; Benfer and Benfer 1981; Bettinger 1997; Butler 1975; Curran 1993; Erkins and Bettinger 2001; Fenenga and Wheat 1940;
Gunn and Pruitt 1975; Thomas 1981; Towner and Warburton 1990) or confound dart points and arrowheads (Benfer and Benfer 1981; Bleed 1986,
1991; Cheshier and Kelley 2006; Hester 1986; Keeley 1982; Neslon 1991). Fewer studies are exclusively concerned with arrowheads (Barton and
Bergman 1982; Bergman 1997; Bergman and Newcomer 1983; Browne 1938; Higgins 1978; Mallouf 1987; Odell and Cowen 1986; Pope 1918, 1923).
This is problematic in that the design constraints on dart points and arrowheads differ considerably. At the most fundamental level, dart points
appear to function as a generalized element of a multipurpose tool. Use wear studies indicate that the dart and its point were used for butchering
game as well as taking it down, combining the design constraints of projectiles and knives (Ahler 1971). Arrowheads are more specialized. Their
small size makes them less useful for butchering and the aerodynamic requirements of arrow flight constrain the practical range of shapes and
sizes (Higgins 1978.) Interestingly, tools designed to function as single-purpose knives are relatively rare in Archaic assemblages and more
common on protohistoric sites (Seymour 2002).

The profession’s focus on dart points is readily evident in the production of projectile point typologies. Work in the Great Basin has resulted in
elaborate typologies for dart points (Bettinger and Eerkins 1997; Thomas 1981) but only a general recognition that arrowheads from that region
seem to form a single unified category (Thomas 1981). Elaborate point typologies have been developed for Texas (Justice 2002) but mainly by
amateurs anxious to name their finds and distinguish themselves among their fellow collectors (see, for example, Phelps 1987). The general lack
of attention to arrowheads as a separate class is especially evident in the many metrically based studies that have been conducted.

A number of researchers have explored the feasibility of using measurements, sometimes combined with attribute observations, as the basis for
categorizing projectile points but nearly all of these studies have addressed dart points, not arrowheads (Ahler 1971; Benfer and Benfer 1981;
Cheshier and Kelley 2006; Erkins and Bettinger 2001; Gunn 1975; Thomas 1981). The metrics themselves are generally linear measures (less
commonly angles) tied to the perceived key landmarks on the points. In almost all cases, those key landmarks are on the hafting end of the point
and designed to result in categories that strongly reflect variations in how the points were attached. This approach is arguably appropriate when
studying dart points, which tend to be heavily reworked before discard, so that a typology of bases is more useful than one that would require
whole points. Further, it seems likely that, as a low velocity weapon, the dart depends mainly on weight for penetration and the impression from
examining numerous illustrations is that the distinction between broadheads and bodkins may not apply to darts.

An earlier metric study of protohistoric arrowheads I conducted took off directly from the metric approaches developed for studying Great Basin
dart points, with decidedly mixed results (Harlan 2002). An indicator of potential difficulties arose when it became apparent that only a subset of
the arrowheads in the study population have any of the landmarks that assist measurement of dart points. These problems carried through the
multivariate analysis of the metric data and, despite applying a range of techniques, the metrics neither sorted the data into meaningful units nor
corresponded to units imposed by assigning the points to traditional types. The present study, another metric analysis using a subset of the same
artifacts measured previously, takes a different approach to quantification.

A Neutral Approach to the Metrics of Size and Shape

The underlying theme of the measurement approach used for this study has been to maintain neutrality, not assuming that any single part of an
arrowhead is more important than any other when searching for meaningful categories. This rules out the use of landmarks and makes
application of linear measures and angles difficult. The solution chosen here has been to divide each point into a series of horizontal and vertical
bands, measure the area of the bands, and compare the points on the basis of differences in all of the bands. The specific steps, as summarized in
figure 1, are simple: 1) convert a digital image of each point to a matrix of zeros and ones, with the ones corresponding to each pixel of the
image occupied by the point and 2) sum up the values across the rows and down the columns of the matrix to provide the individual areas of a set
of horizontal and vertical bands. The measures were then normalized to 40 horizontal and 25 vertical units so that points of varying sizes could
occupy the same multivariate space.2

Two of the many ways to examine the metrics provided by this technique are size-neutral and size dependent treatments. The data as extracted
are size dependent, since the horizontal and vertical pixel counts correspond to areas on the points3. A size neutral representation can be
achieved by converting each measurement band to its corresponding percentage of the total area of the point. The size-dependent view is most
likely to highlight variation related to the intended physical function of the points, since it captures both the distribution of area over the length of
the point and the absolute relationship between the dimensions of the point and the arrow to which it was once attached. The size-independent
view is primarily sensitive to shape, closely relating two points with the same outline even though they differ substantially in size.

Both examinations of the data are designed to address the same basic questions: 1) Can basic categories be extracted from the underlying
variability? 2) Is there correspondence between the variability as measured by this approach and the categories of traditional arrowhead
typologies? As discussed below, the size-dependent and size-independent views produce different results when examining underlying categories
in the measures themselves but similar views with respect to the correspondence between the categories from traditional typology and the
measurement data.

Results

Selection from the larger data set of 1140 arrowheads used in the earlier study cited above, combined with winnowing out problem images,
resulted in a dataset of 254 points for use in the present study. As discussed in more detail in the earlier report (Harlan 2002), the points are from
private collections made on both sides of the international border in the general vicinity of El Paso by members of the El Paso Archaeological
Society. While it would have been preferable to include all of the points from the earlier study, developing and applying a new measurement
technique proved time-consuming and a smaller subset was selected. However, blocks of the full data set were taken in without further selection
and it is likely that the sample chosen is reasonably representative of the entire group of collections.

Even this smaller set of points is too large to approach by inspection, especially given the large number of observations (40 horizontal bands and 25
vertical bands per point), so the analysis is based primarily on multivariate statistical manipulations, with other techniques applied as appropriate.
The search for underlying groups in the data relies on Principal Components analysis, followed up by K-Means cluster analysis and the
investigation of relationships between the metric data and traditional typology uses Discriminant Analysis4. These are all standard multivariate
tools familiar to many archaeologists and described in standard texts (such as Cooley and Lhones 1975 and Harmon 1967).

Figure 2 presents the results of Principal Components analysis with a plot of the first two eigenvectors from the size-dependent data on the left and
size independent data on the right. A complete lack of pattern is readily apparent in the plot based on the size-dependent data. The plot for the
size-independent data has a peculiar form because the correlation matrix derived from the size-independent data is somewhat ill-conditioned, with
the first eigenvector accounting for 93.7% of the total variance. Plots of the second and third and third and fourth eigenvectors were examined and
showed no improvement in form. This is a problem often encountered with data converted to percentages.

The lack of clustering in the Principal Components analysis indicates that underlying variation is the data set as a whole is continuous. It does not,
however, indicate that partitions do not exist, just that other methods may be needed to search for them. In line with the discussion of factors
influencing variability in the shape and size of arrowheads, two sources seem most appropriate for further discussion here: the possibilities that 1)
meaningful partitions may relate to the distinction between broad- heads and bodkins and 2) that coherent groups may be revealed by imposing
the classification of traditional typology onto the data. These alternatives are not mutually exclusive and potential interactions will be addressed.

Keeping in mind that each of the cluster analyses is just a different view of the same underlying data; they do reveal how the shape of the
points reflects the functional categories of broadhead and bodkin. The two groups called for in the first analysis are not really strongly
differentiated by either size or Length/Width ratio, recalling the thorough mixing seen in the Principal Components analysis. The four group
analysis begins to reveal something of the underlying structure of the data. Viewing the four groups as two pairs, we see a marked difference in
Length/Width ratio, coupled with an interior division of each pair into larger and smaller points, showing that within the broadhead and bodkin
categories points can be either larger or smaller. This pattern carries through to the eight group solution, except for the last pair (combining
groups 2 and 6). In this case, the pair consists of large points that are shorter and broader than the rest of the sample.

A reasonable interpretation of the results of cluster analysis is that, for Protohistoric arrowheads from the southern Southwest, broadhead and
bodkin do not form a dichotomy but rather describe a tendency across a range of possible shapes created as design trade-offs between the need
to penetrate the target or create a large wound. We might think of the arrows in these archers’ quivers as analogous to the clubs in a golf bag.
As a dichotomy, golf clubs are designed to impart a parabolic trajectory for short shots or a flattened trajectory for long shots (discounting the
function of the putter). The well-equipped golfer does not, however, carry two clubs but rather fourteen to sixteen, allowing for fine adjustments
to the wide range of situations the game presents. Protohistoric archers may have equipped themselves in a similar fashion, allowing
themselves to choose just the right combination of arrowhead, shaft and flethcing to give them a maximum probability of success in an even
more complex set of potential situations. This leaves open the question of whether or not groups created by the traditional typologies of
arrowheads will also partition the metric data set and how these categories might or might not correspond to the functional distinction of
broadhead versus bodkin.

Before discussing any correspondence between the clusters derived from K-Means cluster analysis and the groups created by traditional
typological approaches, it is necessary to indicate how the typological groups were derived. As a first step in the process, Deni Seymour
assigned 202 of the 254 points to protohistoric types as described in sources for the archaeology of the southern Southwest (Seymour
2002:266-273).5 This procedure left 52 points not assigned to types and these were classed into generalized shape categories (Stemmed,
Basal-Notched, Side-Notched, Un-Notched and Triangular). Table 2 shows the distribution of typological categories over the four pairs of clusters
from K-Means cluster analysis calling for 8 groups. As clearly shown, there is no association.

Coupling the results of Discriminant Analysis based on typological categories with the earlier discussion of functional categories, the bottom
line appears to be that the types of the traditional typology are alternative methods of achieving the basic function of the arrowheads. Those
familiar with how these types are constructed will recognize that the alternatives relate mainly to how the points were hafted and, to a
lesser extent, how they were barbed. With the exception of the stemmed forms, the points which were not assigned to a defined type may
either represent less skilled attempts to produce a point of one of the types or they may be more expedient forms. Clearly, then, the
typological categories relate to the poorly defined realm of “style” and further discussion should focus there.

Protohistoric Arrowhead Style

This paper is not the place to enter into the debate over what “style” is, its relationship to typology or how artifacts should be classified, a
discussion with considerable breadth of coverage, time depth and staying power (Wissler 1916; Ford and Griffin 1937; Krieger 1944; Ford
1954a, 1954b; Spaulding 1953a, 1953b, 1954, 1960, 1973, 1977; Gifford 1960; Berlin 1968, Binford 1962, 1963, 1989; Abbott, Leonard and Jones
1996; Beck 1998; Bettinger 1996; Braun 1995; Carr and Neitzel 1995; Dunnell 1986, 1995; Read 1989; Sabloff and Smith 1969). However, it is
necessary to place the discussion of Protohistoric arrowhead style within the debate and to indicate where the view taken here falls with
respect to the many theoretical and methodological alternatives that have been proposed.

In general, the thrust of the present discussion falls close to the concepts formulated by Wobst (1977) and Wiesner (1982, 1983, 1985) and
rejects the approaches of Sackett (1977, 1985), especially as they guide Jelinek(1976). The discussion here does not adopt Dunnell’s
formulations or those who follow him in employing a selectionist theory of cultural variation. The aspect of Sackett’s concept of style most
firmly rejected is his assertion that “emblematic” style, the use of artifacts as a ready means of group identification is rare and unimportant.
His alternative, a concept he labeled as “isochrestic” style, recognizes that craftsmen take varying approaches to the same physical function
but his bottom line explanation is, essentially, that they just tend to do so, in the same way that they have different gestures, eating habits or
attitudes toward cross-cousins; and that they do not employ these differences to signal group membership or identity within the group
(Sackett 1985:158). Sackett provides no empirical support for his assertions, relying instead on rejections of Wiesner’s findings that are
logically inconsistent within themselves or misapplications of the data in her studies.

Jelinek (1976:19) accepts Sackett’s concept that function relates to task performance whereas style represents choices among the functional
equivalents. Like Sackett, he believes that a given attribute may be either functional or stylistic, depending on the situation in which the
analysis is conducted (Jelinek 1976:20). Based on this starting point, his expectation is that functional attributes will exhibit less variability
than stylistic attributes. In the case of Protohistoric arrowheads, the data presented above show that the biggest metric variation separates
broadheds from bodkins while the variation that appears to be stylistic is more subtle. It would then seem necessary to diverge from
Sackett’s formulation and its implications for stone tools as seen by Jelinek to achieve a better understanding of arrowheads. There is simply
too much patterning in Protohistoric arrowhead variability beyond that needed for physical function to see it as a residual category devoid of
meaning.

There are similar problems with Dunnell’s (1978) formulation, both in his own work and as his concepts have been explored by others
(Nieman 1990, 1995; Graves and Cachola-Abad 1996; Graves and Ladefoged 1995, 1996; Obrien and Holland 1990). Dunnell too has relegated
style to a residual category, defining it as the aspect of artifact variability that is not subject to selective pressure. However, it seems clear
that he is really focusing on those aspects of artifact variability where alternatives in form serve the same function and he does not
recognize the adaptive significance of the alternatives themselves. Nieman has followed up with a more formal analysis but falls into the
same trap, since he attributes stylistic variability to drift (Nieman 1995:8), confounding the diachronic causes of stylistic variation with their
potential synchronic significance.

In contrast to both Sackett and the selectionists, Wiesner and Wobst provide a workable framework for examining the aspects of arrowhead
variability that are not related to their function as broadheads or bodkins. Following their lead, we may look for the aspects of arrowhead
variability that were imparted because these craftsmen thought they made the arrowheads “look and feel” the way they should, beyond
functional design constraints. Wobst has advanced a set of concepts around the central theme that stylistic variability in artifacts comprises
those aspects of variation containing information the makers wished to communicate to themselves and those around them. Wiesner has
given us a detailed example of how these concepts map onto variability in San arrowheads, bringing us to the realm of the present study.
Both authors will be explored further, rejecting the objections Sackett raised to Wiesner’s work and ignoring the fact that Wobst’s
formulation is incompatible with Dunnell’s concept of style and with the way that Jekinek’s has applied Sackett’s definitions to stone tools.
Most fundamentally, the discussion here pays no credence to the time-honored assertion that all equations of projectile points with group
membership are fundamentally flawed at their base and cannot proceed, recognizing that this view has been widely held and continues as
gospel in some archaeological circles.

As indicated at the outset, the population of projectile points that provides the data for this study is from amateur collections and their
proveniences, although available to some extent, are not considered here. It is clear that these points were made by a range of communities
and were likely produced at varying times throughout the Protohistoric period (and perhaps earlier or later). Before returning to that dataset
to examine its stylistic aspects, a brief consideration of another amateur collection may provide a useful starting point, since this one is
tightly provenienced and attributable to a single event.

Robert Mallouf (1987) has studied a large collection of points from a cache found by amateurs in a cairn in northern Mexico. The cairn
contained the burial of an adolescent and possibly an adult but Mallouf, who visited the site, was unable to determine exactly what was
associated with the points because the amateurs had totally looted the cairn. The points comprise 180 of Perdiz style (there are no Perdiz
points in this study because they do not occur in the El Paso area) and a few Toyah, Fresno, Basal-Notched and Side-Notched types (the last
two used as formal categories in Mallouf’s study, not as residual categories). Like the Perdiz points, the other types are mainly made on
local jasper. Perdiz points appear around AD 1350 and were in use up into historic times, with a hiatus around 1500 (Mallouf 1987:62-63).

Mallouf does not believe that the variability seen in the points from the cairn is associated with band or tribal differences (Mallouf 1987:61)
but does recognize that Perdiz points are often attributed to the Jumano (Mallouf 1987:0064). He believes that this point style was designed to
work with cane foreshafts whereas the other types in the cache would work best with hardwood foreshafts, some of which were recovered
from the cairn (Mallouf 1987:61). He links the people who built the cairn to his Cielo Complex (Mallouf 1987:0064-65). The Cielo Complex also
is characterized by a range of distinctive tool types that are essentially the same as those that Seymour has associated with the Canutio
Complex (Mallouf 1987:65-66; Seymour 2002). Maloof links the Cielo complex to bison hunting, but not to any one ethnic group.

If these points had been found in a quiver hanging from the back of a !Kung San hunter, their variability would make perfect sense. Mallouf
is of the impression that, as a group, they are exceptionally well made. Thus the person who owned them (perhaps the adult in the cairn)
may have been an outstanding flint knapper, as some of the San are exceptionally skilled at cold hammering arrowheads (Wiesner
1983:261). In a San quiver, the distinctive point types would represent items traded from partners in other hunting communities (Wiesner
1983:261).

Extending the conformity between Wiesner results and the points from Mallouf’s cairn, we may consider the possibility that the points in the
present study were made by flint-knapping fletchers from a range of communities across the southern Southwest. Each artisan would have
possessed a distinctive tradition for giving arrows a sharp tip designed for some degree of trade-off between deeply penetrating the target
and causing extensive hemorrhage. This view guides the analysis and discussion of relationships between point types.

Figure 7 presents summary metric data for the points that are not as distinct as those in figure 5 but which still do not appear to be part of any
cluster. Clearly the Stemmed points form the most distinct group shown here. Stemmed points are a fundamentally different approach to hafting
that Mallouf (1987:61) has suggested may result from use with a cane foreshaft, while the other hafting methods in the study sample are more
appropriate for hardwood foreshafts. The three types are all small categories and small numbers may account for much of their apparent
distinctiveness. That noted, and focusing on the bar charts of average horizontal band area, the remaining categories do look generally similar,
indicating a common approach to hafting. Each of the categories consists of points that are moderately to strongly basally notched, without side
notching. The final figure (8) shows the pairing of Basal Notched and Toyah points. The similarity in the bar charts reflects the fact that Toyah is
a basally notched type, but with strong side notching. The irregularities in the histogram for the Basal Notched category reflect the fact that this
group is a true residual in comparison to the other points with basal notches.

After closely examining and regrouping the metric data in light of traditional typological categories, it is clear that the categories do track
metrically defined shapes but that they do so imperfectly. It also seems that some of the groups are more closely related than others, perhaps
indicating learning traditions that interacted without complete sharing. It would seem that shared learning relates mainly to alternative hafting
methods since the typological categories do not correspond closely to differences in the fundamental function of arrowheads as broadheads or
bodkins.

Interpreted in light of the discussion of “style” presented above, especially considering Wiesner’s study of arrowhead production among the
San, it seems reasonable to assert that the typological categories into which Seymour sorted the points do correspond to differing ways of
solving the basic problems of providing arrows with a sharp point and attaching that point to the arrow shaft. Of course, as every typologist
knows, the categories are reasonably clear at the center but fuzzy at the edges, whether the boundaries are defined by an expert applying type
descriptions or a multivariate analysis seeking communalities. The logical extension of the results is, however, a set of working hypotheses that
the point types correspond to discrete interacting communities of people who lived just before and after Spaniards intruded upon their world.

This assertion would, of course, create a correspondence between point types and people; a form that many archaeologists have been trained
to reject out of hand as mere speculation. There are two reasons for rejecting that rejection. The first is logical. Readers still following the
presentation at this point likely have some basic grounding in statistical analysis and so are trained to avoid Type I error, accepting an untrue
hypothesis on the basis of associations that are not sufficiently significant. Abhorrence of Type I error is not confined to the world of statistical
analysis. Archaeologists are taught never to accept an apparent association, like that between projectile point types and the people who
produced the points, unless given an iron-clad argument based on irrefutable evidence that those particular people really did make those
specific points, and they are willing to accept a single point found out of place as all the proof needed to reject an association based on
otherwise compelling evidence.

This approach creates its own set of problems. Avoiding Type I error can doom us to continually committing Type II error, rejecting potentially
useful hypotheses on the basis of limited incongruities or unclear patterning, then moving off to other investigations that provide greater
certainty but often yield trivial results. The present extended discussion of arrowhead variability examined through neutral approaches to shape
and without assumptions about which variability should receive a priori privilege, is sufficient basis to proceed to the next step of considering
the possibility that the arrowhead types defined for the Protohistoric, with suitable refinements and no attempt to shoehorn divergent
specimens, may provide information on the emergence, distribution and interaction of the human communities whose lives constitute the
history of the period. The potential rewards are considerable, since we might then be able to trace the movements of peoples during the critical
time period leading up to European contact. Such studies could also help to unravel the identities of some of the “historic” peoples who are
mentioned in the sources, but not really documented by them. The present study, based as it is on un-provenienced collections, can take us no
further than this modest proposition.

Back to Function

It seems clear that stylistic studies aimed at identifying spheres of common learning are only part of the information inherent in Protohistoric
arrowheads. Perhaps the more important inferences are related to design characteristics of the bow and arrow weapons systems and the reasons
for its wide-spread adoption after A.D. 1300. Starting where little controversy should exist, it is almost certain the system was adopted for its
advantages in two key activities, hunting and warfare. Given that earlier prehistoric peoples also hunted and probably were not fundamentally
less prone to armed conflict than people in the Protohistoric, we may wonder why the bow and arrow system assumed greater importance at that
time.

In the area of hunting, the potentially increased importance of bison has not gone unnoticed by other investigators and seems a reasonable basis
for greater use of the bow and arrow. Small game in the size range of squirrels and especially rabbits can be easily taken with snares and
throwing sticks. Faunal remains from the Archaic forward attest to success of rabbit hunting without use of the bow and arrow. Deer would be
better candidates for bow hunting but, as all deer hunters know, the best bet is a good blind near a watering point or on the edge of meadow
with good browse. It is not unreasonable to believe that hunters employing the dart and atlatl successfully hunted deer and other medium-sized,
relatively thin-skinned game, again as well attested by faunal remains in the archaeological record.

Bison are another matter entirely. First, they are thick-skinned and a spear, even assisted by an atlatl, would be less effective in taking them
down, unless employed at close range (Butler 1975:109). More importantly, they are dangerous and engagement from a distance has obvious
advantages. The bow and arrow weapons system solves both problems. It offers both higher velocity and grater range, providing the opportunity
to inflict a potentially fatal wound on these large animals at distances that do not endanger the hunter.

Armed conflict is an even clearer venue for the advantages of the bow and arrow system. As historically documented to heart-wrenching effect,
the availability of projectile-launching systems that send out their lethal messengers at higher velocities, to greater ranges and at higher rates of
delivery make it exponentially easier to defend fixed positions; say, for example, one’s adobe-walled habitation. On the other side of a violent
equation, suppressive fire against those fixed positions calls for attackers with equally effective projectile weapons. When the object of violence
was theft rather than conquest, the bow and arrow would have been an effective deterrent. The wickedly serrated points that appear during the
Protohistoric, especially those that narrow at the haft, would have been capable of causing discomfort, debilitation and possibly death weeks
after the wound was inflicted. This potential may have made communities with a reputation for effective archery less desirable targets.

Like investigation of associations between projectile point types and the groups that produced the points, consideration of the changing
exigencies that led to greater use of the bow and arrow weapons system has potential to help us understand the key set of changed conditions
that followed the fall of the Southwestern great cultures and preceded the arrival of European invaders. Clearly, profound changes occurred. The
possibility that the sedentary folks that have always been our focus were not at the center of a changing world is worth considering.

NOTES

1 Interestingly, sharpened hardwood shafts were also used by a number of indigenous groups in California for their war arrows. Justice (2002:40)
attributes this to a desire to maximize the number of arrows they could afford to shoot at their enemies but it is just as likely that these people
designed their weapons to wound deeply but not to kill.
2 In practice, this proved difficult. Problems with varying image quality and had to be overcome and reformatting the data for statistical analysis
required a good deal of database manipulation, including writing a fair amount of code.
3 The photos used in this study were taken only as back-ups for the previous metric study and were not standardized in ways that would have
been most useful for the present approach. Problems separating the points from the background resulted in discarding some data and it was
necessary to adjust for camera height, which varied for groups of photos.
4 Principal Components as used in the archaeological literature actually refers to a variety of matrix reduction techniques, some of which
employ various methods for removing the unities from the principal diagonal of the correlation matrix in order to reduce the variance in the
eigenvector associated with the first eigenvalue in hopes of making the reduced-space solution easier to interpret. The problem with these
approaches is that the reduced-space solution bears no specifiable relationship to the original correlation data. The Principal Components
analyses discussed here have been applied to un-rotated correlation matrices with unities on the principal diagonal. This leaves the lion’s share
of the variance with the first eigenvector, making groups more difficult to discern, but also preserves the relationship to the underlying data.
5 The measures of length and width and size are derived from the data extracted from the photos in different ways. Length and width were
determined from a pixel count of the longest dimension and widest dimension of the point in the photo. Area is a count of the total pixels in the
photo occupied by the image of the point. In both cases, the pixel counts have been normed for camera height.
6 An independent source of typological categories avoids the possibility of a single investigator contaminating the results by applying metric
criteria that are not appropriate to the typologies. A further advantage is the fact that Seymour is more familiar with the protohistoric point types
than I am.
7 In the interest of brevity and in order to focus on results rather than methods, the discussion here omits a good deal of experimentation with the
input data, both in terms of cases and variables. It is, however useful to note that the results discussed here employ the full set of horizontal and
vertical bands while eliminating a few cases where classification into either a type or a residual category was questionable. As would be
expected, using all of the measurement bands results in an over-specified model with the raw classification matrix nearly twice as accurate as
the jackknifed classification matrix. The need for over-specification probably has implications for the nature of the process of classification, but
these implications have not been explored for the present study.

REFERENCES CITED

Abbott, A L, and R.D. Leonard and G.T. Jones
1996 Explaining the change from biface to flake technology: A selection application, In Darwinian Archaeologies, edited by H.G.D. Maschner. pp,
33-42. Plenum Press, New York

Ahler, S A.
1971 Projectile point form and function at the Rogers Shelter, Missouri, Archaeological Society Research Series No. 8 .

Bamford, Douglas B.
1986 Technological organization and tool curation, American Antiquity 51:38-50

Barton, R NE, and C.A. Bergman
1982 Hunters at Hengistbury: some evidence from experimental archery, World Archaeology 14:237-248

Beck, C .
1998 Projectile point types as valid chronological units, In Unit Issues in Archaeology: Measuring Time, Space and Material, edited by A.F.
Ramenofsky and A. Steffen. pp, 21-40. University of Utah Press

Benfer, Robert A, and Alice N. Benfer
1981 Automatic classification of inspectional categories: multivariate theories of Archaeological data, American Antiquity 46:381-396

Benfer, Robert A.
1970 Review of Towards a statistical overview of Archaic cultures in central and southwest Texas by Le Roy Johnson Jr, American Antiquity
35:119-120

Bergman, Christopher A, and Edward McEwen
1997 Sinew-reinforced and composite bows: technology, function and social implications, In Projectile Technology, edited by Heidi Knecht. pp, 143-
160. Plenum Press, New York and London

Bergman, Christopher A, and Mark E. Newcomer
1983 Flint Arrowhead Breakage: Examples from Ksar Akil, Lebanon, Journal of Field Archaeology 10:238-243

Berlin, Brent , and Dennis B. Breedlove and Peter H. Ronen
1968 Covert categories and folk taxonomies, American Anthropologist 70:290-

Bettinger, Robert L, and L.R. Boyd and P.J. Richardson
1996 Style, function and cultural evolutionary process, In Darwinian Archaeologies, edited by H.G.D. Maschner. pp, 133-164. Plenum Press

Binford, Lewis R.
1989 The "New Archaeology" Then and Now, In Archaeological Thought in America, edited by C.C. Lamberg-Karlovsky. pp, 50-62. Cambridge
University Press, Cambridge

1963 Red ocher caches from the Michigan area: a possible case of cultural drift, Southwestern Journal of Anthropology 19:89-107

1962 Archaeology as Anthropology, American Antiquity 28:217-225

Bleed, P .
1986 The optimal design of hunting weapons, maintainability or reliability. American Antiquity 51:737-747

Braun, D P.
1995 Style, selection and historicity, In Style, Society and Person: Archaeological and Ethnological Perspectives, edited by C. Carr and J.E.
Neitzel. pp, 123-141. Plenum Press, New York

Browne, James
1940 Projectile Points, American Antiquity 5:209-213

1938 Antiquity of the bow, American Antiquity 3:358-359

Butler, W B.
1975 The Atlatl: the Physics of Function and Performance, Plains Anthropologist 20:105-110

Carr, C , and J.E. Neitzel
1995 Style, Society and Person, In Style, Society and Person, edited by C. Carr and J.E. Neitzel. pp, -. Plenum Press, New York

Cheshier, Joseph , and Robert L. Kelley
2006 Projectile Point Shape and Durability: The Effect of Thickness: Length, American Antiquity 71:353-363

Cooley, W W, and P.R. Lohnes
1968 Multivariate Data Analysis . American Institutes for Research, Palo Alto

Dunnell, Robert C.
1978 Style and Function: A Fundamental Dichotomy, American Antiquity 43:192-202

1995 Style, Concepts and Practice, Paper Presented at the McClung Museum, University of Tennessee

1986 Methodological issues in americanist artifact classification, Advances in Archaeological Method and Theory 9:149-20

Erkins, Jelmer W, and Robert L. Bettinger
2001 Techniques for Assessing Standardization in Artifact Assemblages: Can We Scale Material Variability, American Antiquity 66:493-504

Evans, O .
1957 Probable use of stone projectile points, American Antiquity 23:83-84

Fenenga, F , and J.B. Wheat
1940 An atlatl from the Baylor Rock Shelter, Culverson County, Texas. American Antiquity 5:221-223

Flenniken, J J, and P.J. Wilkie
1989 Typology, technology and chronology of Great Basin dart points. American Anthropologist 91:149-158

Ford, James A, and James B. Griffin
1937 (A proposal) a conference on pottery nomenclature for the southeastern United States, Newsletter of the Southeastern Archaeological
Conference

Ford, James A.
1954 Comment on Spaulding's "Statistical technique for the discovery of artifact types", American Antiquity 19:390-391

1954 On the concept of types: the type concept revisited, American Anthropologist 56:42-53

Gifford, James C.
1960 The type-variety method of ceramic classification as an indicator of cultural phenomena, American Antiquity 25:341-347

Graves, M W, and C.K. Cachola-Abad
1996 Seriation as a method of chronologically ordering design traits: an example from Hawa`i, Archaeological in Oceania 31:19-32

Graves, M W, and T.N. Ladefoged
1995 The evolutionary significance of ceremonial architecture in Polynesia, In Evolutionary Archaeology: Methodological Issues, edited by P.A.
Telser. pp, 149-174. University of Arizona Press, Tucson

Gunn, Joel E, and Elton R. Pruitt
1975 Automatic classification: projectile points from West Texas, Plains Anthropologist 20:139-149

Hackett, General Sir John
1983 The Profession of Arms. McMillan Publishing Company, New York

Harlan, Mark E.
2002 Appendix B: Projectile point study, In Conquest and Concealment: After the El Paso Phase on Fort Bliss, Lone Mountain Archaeological
Services Inc. Report No. 525/528, edited by Deni J. Seymour. pp, -. United States Army Air Defense Artillery Center and Fort Bliss Directorate of the
Environment-ATZC-DOE-C 1733, On file atthe Arizona State Museum and University of New Mexico.

Harmon, Harry H.
1967 Modern Factor Analysis. The University of Chicago Press, Chicago and London

Hester, Thomas R.
1986 On the misuse of projectile point typology in Mesoamerica, American Antiquity 51:412-414

Higgins, George J.
1978 The Aerodymamics of an Arrow, Journal of the Franklin Institute 216:91-

Jelinek, Arthur J.
1976 Form, function and style in lithic analysis, In Cultural Change and Continuity: Essays in Honor of James Bennett Griffin, edited by C.E.
Cleland. pp, 19-34. Academic Press, New York

Justice, Noel D.
2002 Stone Age Spear and Arrow Points of the Southwestern United States. Indiana University Press, Bloomington and Indianapolis

Keeley, L H.
1982 Hafting retooling: The effects on the archaeological record, American Antiquity 47:798-809

Krieger, Alex D.
1944 The typological concept, American Antiquity 9:271-288

Levine, Frances
1999 Our Prayers Are in This Place, Our Prayers Are in This Place. University of New Mexico Press, Albuquerque

Mallouf, Robert A.
1987 Las Haciendas: A Cairn-Burial Assemblage from Northeastern Chichuahua, Mexico, Las Haciendas: Office of the State Archaeologist
Report No 35 . Texas Historical Commission

Nelson, Ben A.
1991 Ceramic frequency and use-life: a highland Maya case in cross-cultural perspective, In Ceramic Ethnoarchaeology, edited by William A
Longacre. pp, 162-181. The University of Arizona Press, Tucson

Nieman, Fraser D.
1995 Stylistic variation in evolutionary perspective: inferences from decorative diversity and interassemblage distance in Illinois Woodland ceramic
assemblages, American Antiquity 60:7-36

Niemen, Fraser D.
1990 An Evolutionary Approach to Archaeological Inference: Aspects of Architectural Variation in 17th Century Chesapeake, Ph.D. Dissertation, Yale
University. University Microfilms

O'Brien, M , and T.D. Holland
1990 Variation, selection and the archaeological record, Archaeological Method and Theory 2:31-79

Odell, George H, and Frank Cowan
1986 Experiments with Spears and Arrows on Animal Targets, Journal of Field Archaeology 13:195-212

Phelps, Alan L.
1987 Soto: A distinctive projectile point type, The Artifact 25:7-22

Pope, Saxon
1918 Yahi Archery, University of California Publications in American Archaeology and Ethnology 13:103-152

1923 A Study of Bows and Arrows, University of California Publications in American Archaeology and Ethnology 13:329-414

Read, Dwight W.
1989 Intuitive typology and automatic classification: Divergence or full circle. Journal of Anthropological Archaeology 8:1-31

Sabloff, Jeremy A, and Robert E. Smith
1969 The importance of both analytic and taxonomic classification in the type-variety system, American Antiquity 34:278-285

Sackett, James R.
1977 The meaning of style in archaeology: a general: model, American Antiquity 42:369-380

Seymour, Deni J.
2006 Apache, Spanish and Protohistoric Archaeology of Fort Bliss, Historic and Natural Resources Report 03-06 . Cultural Resources Program,
Directorate of Environment, Fort Bliss, Texas

2007 Seventeenth-Century Sobaípuri Social And Economic Relations in Northwestern New Spain, Seventeenth-Century Sobaípuri Social And
Economic Relations in Northwestern New Spain Paper in press at New
Mexico Historical Review.

2004 A Ranchería in the Gran Apachería: Evidence of Intercultural Interaction at the Cerro Rojo Site, Plains Anthropologist 49:153-192

2002 Conquest and Concealment: After the El Paso Phase on Fort Bliss, Lone Mountain Archaeological Services Inc. Report No. 525/528. United
States Army Air Defense Artillery Center and Fort Bliss Directorate of the Environment-ATZC-DOE-C 1733, On file at Arizona State Museum and
University of New Mexico

2007 A Syndetic Approach To Identification Of The Historic Mission Site Of San Cayetano Del Tumacácori, A Syndetic Approach To Identification Of
the Historic Mission Site Of San Cayetano Del Tumacácori Paper in press at International Journal of Historical Archaeology.

2004 Before the Spanish Chronicles: Early Apache in the Southern Southwest, In Ancient and Historic Lifeways in North America’s Rocky Mountains.
Proceedings of the 2003 Rocky Mountain Anthropological Conference, Estes Park, Colorado, edited by Robert H. Brunswig and William B. Butler.
National Park Service, Department of the Interior, Estes Park

Spaulding, Albert C.
1973 The concept of artifact type in archaeology, Plateau 45:149-163

1953 Statistical techniques for the discovery of artifact types, American Antiquity 18:305-313

1953 Review of "Measurement of some Prehistoric Design Developments in the Southeastern United States." American Anthropologist 55:588-591

1960 Statistical description and comparison of artifact assemblages, The Application of Quantitative Methods in Archaeology 28:60-83

1977 On growth and form in archaeology: multivariate analysis, Journal of Anthropological Research 33:1-15

1954 Reply (to Ford), American Anthropologist 56:112-114

Thomas, David H.
1981 How to Classify Projectile Points from Monitor Valley, Nevada, Journal of California and Great Basin Anthropology 3:7-43

Towner, Ronald H, and Miranda Warburton
1990 Projectile Point Rejuvenation: A Technological Analysis, Journal of Field Archaeology 17:311-320

Wiesner, Polly
1982 Beyond willow smoke and dogs' tails: a comment on Binford's analysis of hunter-gatherer settlement systems, American Antiquity 47:171-178

1985 Style or Isochrestic Variation: a Reply to Sackett, American Antiquity 50:160-166

1982 Beyond willow smoke and dogs' tails: a comment on Binford's analysis of hunter-gatherer settlement systems, American Antiquity 47:171-178

Wissler, Clark
1916 Harpoons and Darts in the Stefansson Collection, Anthropological Papers of the American Museum of Natural History 14:-

Wobst, H M.
1977 Stylistic behavior and information exchange, In Papers for the Director: Research Essays in Honor of James B. Griffin University of Michigan
Anthropological Papers 61, edited by C.E. Cleland. pp, 317-342. University of Michigan Anthropological Papers