Accepted Manuscript

Title: Interpreting abundance indices: some zooarchaeological implications of Martu

foraging
Authors: Brian F. Codding, Douglas W. Bird, Rebecca Bliege Bird
PII:

S0305-4403(10)00260-8

DOI:

10.1016/j.jas.2010.07.020

Reference:

YJASC 2575

To appear in:

Journal of Archaeological Science

Received Date: 22 February 2010

Revised Date:

19 July 2010

Accepted Date: 20 July 2010

Please cite this article as: Codding, B.F., Bird, D.W., Bird, R.B. Interpreting abundance indices: some
zooarchaeological implications of Martu foraging, Journal of Archaeological Science (2010), doi:
10.1016/j.jas.2010.07.020
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

Interpreting abundance indices: some zooarchaeological implications of Martu foraging

Brian F. Coddinga*, Douglas W. Birda and Rebecca Bliege Birda

Department of Anthropology, Stanford University, 450 Serra Mall, Building 50, Stanford, CA 94305.

*Corresponding Author: Brian F. Codding, email: bcodding@stanford.edu, phone: 650-823-6492.

DO NOT CITE WITHOUT PERMISSION

Codding et al. Abundance Indices

Abstract: Indices of taxonomic abundance are commonly used by zooarchaeologists to examine resource
intensification, overexploitation and gender-divisions in foraging labor. The original formulation of
abundance indices developed a clear interpretive framework by linking the measure with foraging models
from behavioral ecology. However, using the same basic tenets of behavioral ecology, archaeologists
disagree about how to interpret variability in abundance index values: some suggest that high proportions
of large prey remains represent higher overall foraging efficiency, while others argue the opposite. To
help solve this problem, we use quantitative observational data with Martu hunters in Australias Western
Desert to examine how foraging decisions and outcomes best predict variation in the abundance index
values that result. We show that variation in the proportional remains of large to small game is best
predicted by hunting bout success with larger prey and the time spent foraging for smaller prey. A
declining abundance index results from decreasing hunting success with larger prey, increasing time
invested in hunting smaller prey, or both; any of which result in a lower overall return rate than if large
prey were acquired reliably. We also demonstrate that where large prey acquisition is stochastic, high
index values are correlated positively with mens proportional caloric contribution of large unreliable
game, while low index values are correlated with womens proportional foraging time for small reliable
game. We discuss these results with reference to evidence of resource intensification and gender-specific
foraging.

Keywords: human behavioral ecology, zooarchaeology, ethnoarchaeology, resource intensification,

gender division of labor, Western Australia.

Codding et al. Abundance Indices

1 Introduction

In the last few decades, zooarchaeological studies have made important contributions to our
understanding of prehistoric human behavior and human-prey dynamics, providing new insights into
broad-spectrum revolutions (e.g., Stiner 2001; Stutz et al. 2009), anthropogenic resource depression (e.g.,
Broughton 1994; Cannon 2003), human caused extinction events (Jones et al. 2008a; Nagaoka 2002), and
variability in gender-divisions of foraging labor (e.g., Kuhn and Stiner 2006; McGuire and Hildebrandt
2005). Many of these studies rely on indices of taxonomic abundance to evaluate diachronic variability in
prehistoric foraging efficiency and resource choice (see review in Lupo 2007).
Abundance indices (AI) measure the relative profitability of a foraging strategy by calculating the
proportion of higher ranking taxa relative to lower ranking taxa in a zooarchaeological assemblage. The
initial approach was developed by Bayham (1979) as a means to evaluate predictions of the encountercontingent prey choice model (PCM) from behavioral ecology (see Charnov and Orians 1973; MacArthur
and Pianka 1966; Schoener 1971; Stevens and Krebs 1986; see also Bettinger 1991, 2009). The basic
PCM assumes that a foragers main goal is to maximize the rate at which energy is acquired. To achieve
this goal, a forager searching for homogenously distributed resources within a patch should always pursue
the resource with the highest post-encounter profitability when encountered and should only pursue less
profitable resources when the opportunity costs of handling them are less than what could be gained by
continuing to search for more profitable prey. Thus, when the encounter rates with the highest ranking
prey are frequent enough, foragers should only take that resource. However, when the encounter rates
with the highest ranking prey decline, the model predicts that foragers should respond by widening their
diet breadth to include lower ranked resources, which reduces a foragers overall efficiency. When
resources are not distributed homogenously, this same logic can and has (e.g., Hildebrandt and Jones
1992; see also Broughton 1994) been applied to resource patches (see Bettinger 1991; MacArthur and
Pianka 1966) or foraging activities (Smith 1991); both of which can be thought of as the PCM operating
at a larger scale. In such cases, patches or activities are ranked by their energetic return rate and as with

Codding et al. Abundance Indices

the standard PCM, whether or not a forager should enter a patch type or be involved in a foraging activity
depends on the abundance of the highest ranking patch in the environment or the abundance of resources
within the highest ranking patch. Because neither post-encounter profitability within a patch, nor the
profitability of a resource patch or activity type overall can be observed archaeologically, Bayham (1979,
1982) argued that resource rank should scale with prey body-size (see Griffiths 1975) and that the
addition of smaller prey is a function of declining encounters with larger prey. Most applications of
abundance indices have followed this lead, and as such, the model predictions shift to be specifically
about large prey versus small prey (although, see e.g., Stiner et al. 2001; Jones et al. 2008b).
Based on this traditional framework, high proportions of large prey relative to small prey have
been used as evidence for high overall forging return rates and a divergent gender division of labor
focused on mens acquisition of larger prey (e.g., Broughton and Bayham 2003; Broughton et al. 2008;
see also Isaac 1978). Diachronic declines in the abundance of large prey then suggest declines in overall
foraging efficiency (e.g., Nagaoka 2002), widening diet breadth associated with broad spectrum
revolutions (e.g., Stutz et al. 2009) and a gender division of labor in which reductions in large game
necessitate the acquisition of smaller game by either men, women or both.
However, ethnographic and actualistic work has questioned the assumption that prey body size
and prey rank are always positively correlated. Mass capture techniques may increase post-encounter
return rates for some types of small prey, particularly fish and insects (Madsen and Krikman 1988;
Madsen and Schmitt 1998; Ugan 2005; Lupo and Schmitt 2002, 2005). Moreover, under some
circumstances, larger prey may be of lower rank than predicted due to the effects of relative prey
mobility, which can increase with prey size, and may lead to higher instances of pursuit failure (Bird et al.
2009; see also Hawkes et al. 1991; Jochim 1976; OConnell et al. 1988; Smith 1991:230-231; Stiner et al.
2000; Winterhalder 1981:95-96). If this is the case, foragers may attain higher overall return rates by
pursuing smaller prey that can be acquired more reliably. Because foragers (often men) continue to pursue
larger prey despite the acquisition risk, it may be that the actual goals of foraging vary as a function of
gender (Jochim 1988), with men focused on maximizing currencies other than the rate of resource

Codding et al. Abundance Indices

acquisition, such as social capital or prestige (Bliege Bird and Smith 2005; Hildebrandt and Mcguire
2002). Instances of such behavior represent a clear violation of one of the primary assumptions of the
PCM (e.g., Bliege Bird et al. 2001; Hawkes et al. 1991; Hill et al. 1987; see also Lee 1968).
With this critique, an alternative interpretation of abundance indices has emerged in opposition to
the traditional interpretation. This alternative view suggests that high proportions of large prey relative to
small prey represent lower overall foraging efficiency (e.g., Hildebrandt and McGuire 2002; McGuire et
al. 2007) and a gender division in foraging labor in which mens pursuit of large prey is subsidized by
womens labor focused on more reliable resources (Hildebrandt and McGuire 2002; McGuire and
Hildebrandt 2005). Accordingly, diachronic declines in the abundance of larger prey relative to smaller
prey then reflect increases in foraging efficiency and an increase in the overlap between mens and
womens resource choice, both being focused on small prey (McGuire and Hildebrandt 1994, 2005; but
see also Jarvenpa and Brumbach 2009; Kuhn and Stiner 2006; Waguespack 2005; Zeanah 2004).
That opposite interpretations exist for a single quantitative measure is a problem for
zooarchaeological analyses; if both interpretations are taken seriously, competing hypotheses about
prehistoric foraging are rendered essentially untestable. However, despite being a static material outcome
of complex and dynamic foraging decisions, abundance indices should still provide a relatively
straightforward measure of variability in the outcomes of human foraging. That is, where we can control
for the effects of differential transport and post-depositional processes, abundance indices should reflect
some aspect of the trade-off between hunting larger and smaller prey. As such, actualistic research can
provide direct evaluation of the parameters of such trade-offs and the zooarchaeological patterns they
produce.
Here we undertake an empirical exercise to examine the meaning of abundance indices. We ask
(1) how does foraging behavior predict variation in abundance indices? And (2) how do differences in
mens and womens foraging decisions explain variation in abundance indices? To answer these
questions, we draw on quantitative foraging data collected with Martu, a group of Aboriginal Australians
living in arid Western Australia. Specifically, we examine the tradeoff between hunting hill kangaroo

Codding et al. Abundance Indices

(Macropus robustus) and sand monitor lizards (Varanus gouldii), two major resources that embody the
trade-offs between hunting larger and smaller prey. First, we test predictions derived from the traditional
interpretation of abundance indices to examine how the overall rates of energetic return, total foraging
time, hunting success and total harvest size for hill kangaroo and sand monitors predict abundance index
values. Second, we test predictions derived from the traditional interpretation of abundance indices to
examine how gender-differences in foraging strategies are reflected by variability in abundance index
values. Following OConnell (1995), this approach allows predictions derived from general theory of
behavior to be tested simultaneously with observations and their archaeological correlates, thus avoiding
the problems associated with simple analogy. As such, even though Martu hunting does not represent all
variability in prehistoric or even contemporary hunting, the basic links between human foraging
decisions, prey size, mobility, foraging returns and their relative impacts on abundance indices should
provide a baseline that can help guide future analyses and resolve some of the confusion surrounding their
interpretation.

2 Contemporary Martu Foraging

Martu, sometimes written as Mardu or Mardujarra, literally means human, but is also used as a
common term of self-reference for a set of Aboriginal Australian dialect groups whose homelands are
centered around the Karlymillyi (Rudall) River and Percival Lakes in Western Australias Gibson, Little
Sandy and Great Sandy Deserts (Figure 1). Some of these groups came into contact with European
Australians in the mid-late 1960s and were subsequently settled onto missions and cattle stations
(Davenport 2005; Scelza and Bliege Bird 2008; Tonkinson 1991, 1979). After a 20-year hiatus, many
Martu returned to their homelands and today have Native Title over their traditional estates. Life in the
desert is primarily centered in one of three remote communities or outstations: Parnngurr, Punmu and
Kunawarritji. This work is focused in Parnngurr community, which is home to Martu mainly from the
Manyjiljarra, Warnman and Kartujarra dialect groups.

Codding et al. Abundance Indices

Subsistence hunting remains of central importance to Martu in these communities. Hunting

parties are sometimes planned in advance, but more frequently emerge ad hoc on a given day. Through
open discussion, sometimes occurring partially en route, individuals decide on the motor vehicle they will
take, which region they will travel to and the hunting activities they will engage in. On arrival, the party
establishes a dinner-time camp. Foragers depart on foot from this camp and will later return to process,
cook and share their acquired resources. The time they are away from this camp we define as a foraging
bout.
Once at the dinner-time camp, Martu typically choose to engage in one of a variety of hunting
activities including sand monitor (parnajalpa, Varanus gouldii), hill kangaroo (kirti-kirti, Macropus
robustus), bustard (kipara, Ardiotis australis) and perentie monitor (yaliparra, Varanus gigiantus)
hunting. As the decision to engage in one hunting activity is made prior to departure and because people
are essentially technologically and ecologically limited to that activity after the decision is made, each
activity can essentially be modeled as a resource patch (see Smith 1991). Bird et al. (2009) and Bliege
Bird and Bird (2008) discuss these hunting activities in more detail; here, we focus on the trade-offs
between two hunt types: hill kangaroo and sand monitor.
Hill kangaroo hunting occurs in and around the low lying rocky ranges that are scattered
throughout the desert (see Figure 1c). Hunters typically enter a range on foot carrying a .22 caliber bolt
action rifle and systematically walk upwind searching the range for hill kangaroo or signs of their
presence. In the suite of game commonly pursued by Martu hunters, hill kangaroo are relatively large
prey with highly variable returns. Based on a dataset of 125 hill kangaroo foraging bouts recorded from
2000 to 2009 (see data collection methods below), hill kangaroo hunters return on average with 3755
+12230 total kilocalories (total harvest size; Table 1), however, this mean is rarely realized. As reported
in Bird et al. (2009), each encounter has a 30% chance of being successful (pursuit success). Summing
over all encounters during a hunting bout, hunters return to camp with prey on average only 15% of the
time, with a median chance of success of 0% (see Table 1). Given such unreliability in acquisition, the
potential caloric rewards of a large prey harvest are realized only infrequently.

Codding et al. Abundance Indices

Sand monitor hunting occurs in the open spinifex (Triodia sp.) sand plains of the desert (see
Figure 1c). While sand monitors are relatively small (just under half a kilogram per individual), hunters
almost always acquire more than one individual, resulting in a mean harvest size of 1888 +1663
kilocalories (Table 1). An encounter with a sand monitor is successful 89% of the time (Bird et al. 2009),
with an equivalent probability of overall hunting bout success (Table 1). In the wintertime, sand monitors
are denned, and Martu frequently use controlled fire to burn off climax vegetation, which allows hunters
to more easily spot dens with the result of lower search costs and an increase in the overall return rate
(Bird et al. 2005; Bliege Bird et al. 2008). In the summer, when monitor lizards are on the surface and
encounters result in quick pursuit, overall hunting returns are lower, but hunting bout success is still
reliable (see Bird et al. 2009). Because hill kangaroo and sand monitor hunting occur in separate patches,
they are mutually exclusive activities in which search time is not shared.
The decision to hunt one or the other prey-type has serious consequences for the expected
outcome of the hunt. Differences in the probability of success create greater variability and
unpredictability in the expected harvest size and overall return rate for kangaroo compared to sand
monitor hunting. Because of the stochastic nature of hill kangaroo search and pursuit success, the amount
of time a forager spends hunting hill kangaroo does not predict total harvest size (F = 2.865, p = 0.0943;
Figure 2a). However, sand monitor hunting time does significantly predict overall harvest size (F = 131.3,
p < 0.0001; Figure 2b). A forager choosing to hunt sand monitors can more or less control the amount of
food they will acquire by adjusting the amount of time they spend foraging, but no such choice or control
is possible with hill kangaroo hunting. However, there are potential benefits to hill kangaroo hunting that,
despite being realized only rarely, may explain why hunters choose to pursue hill kangaroo at all (Bliege
Bird and Bird 2008; Bliege Bird et al. 2009). As shown in Figure 3, sand monitor hunters have a 10%
chance of returning with nothing, but also have an extremely low probability of achieving a return rate
higher than 5000 kcal/hr. Hill kangaroo hunters, on the other hand, have an 85% chance of returning
empty handed, but may occasionally achieve return rates up to 25,000 kcal/hr.

Codding et al. Abundance Indices

The trade-off between choosing to hunt hill kangaroo or sand monitor is largely one between risk
and harvest size. As these differences are driven largely by post-encounter pursuit success probabilities,
and given that these are significantly and negatively correlated with prey body size due to the effects of
relative prey mobility (Bird et al. 2009), we suggest that these trade-offs may be characteristic of large vs.
small prey in many (but certainly not all) circumstances.

3 Methods

3.1 Foraging Data

We collected quantitative foraging data with a combination of (1) focal follows, where a
researcher accompanied a hunter on an entire bout to record all time allocated to search, pursuit, capture
and field processing, and (2) continuous dinner-time camp scans, where a researcher stayed at the
temporary camp and recorded the departure and return times for all foraging bouts (see Altmann 1974); in
both cases we recorded the number and whole weight of all resources each forager acquired. When two or
more hunters cooperated on a hunt, we calculated per-capita returns by dividing total harvest size (whole
weight) by the number of cooperators. Whole weights for each resource were converted into edible
weights, which were then transformed into caloric values based on published sources (Brand-Miller et
al.1993; OConnell and Marhsall 1989; see Bird et al. 2009; Bliege Bird and Bird 2008; Bliege Bird et al.
2009 for additional details).
Data used in this analysis includes 1516 adult foraging bouts recorded from 2000 to 2009. These
data were aggregated for each camp-day (n = 230), making the unit of analysis the aggregate of all
foraging bouts that occurred on a single day at a single dinner-time camp. As we are interested in the
archaeological signatures of large and small game hunting, only camp-days in which at least one forager
attempted to acquire hill kangaroo were included; days where no foragers acquired any prey (hill

Codding et al. Abundance Indices

kangaroo or otherwise) were excluded as these would have no archaeological signature. A total of 36 days
meet both criteria.
To investigate the relative effects of foraging for each hunt type (hill kangaroo and sand monitor),
four foraging measures were calculated for each camp day: (1) the overall return rate (E/T), measured as
total kilocalories acquired per minute (search, pursuit, capture, field processing and transport to the
dinner-time camp) for each hunting activity (hill kangaroo and sand monitor) (2) the total foraging time
spent (in minutes) for each hunt type, (3) whether or not any foragers were successful (success rate) in
either hunt type, and (4) the total harvest size (kilocalories) acquired in each hunting activity which
measures the total energetic yield without costs. Finally, in order to investigate the gender division of
labor in terms of the relative effects of mens and womens foraging on AI, two relative measures were
calculated: mens contribution to subsistence relative to womens (i.e., percent of total kilocalories
acquired by men) and mens foraging time relative to womens (i.e., percent of total foraging time).
An abundance index (AI) was calculated as the total number of hill kangaroo acquired for that
day over the sum number of hill kangaroo plus the total number of sand monitors. This can be thought of
as a hypothetical measure of the archaeological abundance index for each camp without any transport,
depositional or post-depositional biases. Given that this is what zooarchaeologists would choose to
measure if it were possible (see Klein and Cruz-Uribe 1984), we consider this an ideal measure for
interpreting abundance indices relative to known foraging variables. As the index is calculated from the
number of each prey acquired, it is analogous to an abundance index calculated using the minimum
number of individuals (MNI; see e.g., Grayson 1984; Klein and Cruz-Uribe 1984).

3.2 Statistical Methods

As an abundance index value is a ratio constrained to vary between 0 and 1, it cannot be used as a
dependent variable in ordinary least squares regression (OLS; see e.g., Faraway 2006; Keisnick and
McCullough 2003). Accordingly we used a generalized linear model (GLM) of the binomial family with a

10

Codding et al. Abundance Indices

logit (or logistic) link function. Multivariate models were constructed examining the relative effects of the
overall return rate, total foraging time, hunting bout success, and harvest size for each hunt type on the
abundance index values per camp-day. The value of each model was judged based on two sets of criteria,
the first follows a traditional hypothesis testing approach and the second follows an information-theoretic
approach of model selection.
Following a traditional hypothesis testing approach, models are judged to significantly improve
the prediction of the dependent variable if the alpha (or p) value is < 0.05. A significant model is judged
better than another based on comparing the models likelihood ratios ( RL2 ; Menard 2000, 2002). Similar
to R2 values in OLS, RL2 values can be interpreted as a reduction in unexplained deviance by the inclusion
of the independent variables, thus models with higher values reduce more unexplained deviance than
models with lower values. RL2 values can be calculated easily from the GLM output of most statistical
programs (see Menard 2000, 2002; see also Codding et al. 2010).
Following an information-theoretic approach of model selection, we utilized Akaikes An
Information Criterion (AIC; Akaike 1974; Burhnam and Anderson 2002) to estimate the relative value of
each model. AIC was developed as an estimate of the Kullback-Leibler distance between a particular
model and an unknown reality that actually structures the data (Burnham and Anderson 2002; Venables
and Ripley 2002). As the value represents the estimated distance between the specific model and some
true model, one model is relatively better than another if that distance is minimized, that is, lower AIC
values represent relatively better models.
Statistical analyses were run and accompanying figures produced in R (R Development Core
Team 2009). Wireframe or surface plots were created with the use of Rs Lattice library (Sarkar 2008).

4 Results

11

Codding et al. Abundance Indices

Over the 36 camp-days, the return rate for hill kangaroo hunting was highly skewed (median = 0,
mean = 35.3 + 62.2 kcals/min), with a lower median, but higher mean than sand monitor hunting (median
= 9.8, mean = 9.6 + 7.1 kcals/min, see Table 2). More time was spent sand monitor hunting (18,769
minutes) than hill kangaroo hunting (14,494 minutes) on these camp-days, resulting in 17 hill kangaroo
totaling 437,783 kcals and 217 sand monitors totaling 195,716 kcals. The daily abundance index (AI)
values that result range from 0 to 1 (median = 0, mean = 0.23 + 0.40).

4.1 How does foraging behavior predict variation in abundance indices?

Under the traditional interpretation, high abundance index values should be correlated with higher
overall return rates from acquiring a greater proportion larger prey and lower index values should
represent a lower overall return rate contributed more from acquiring a greater proportion smaller prey
(Bayham 1979). Examining the effect of return rate for hill kangaroo and sand monitor on AI confirms
this prediction ( RL2 = 0.81, p < 0.0001; Figure 4a); however, the return rate for sand monitor hunting has
a greater effect (estimate = -0.49) than the rate for hill kangaroo hunting (estimate = 0.02; see Table 3).
This suggests that even if return rates for hill kangaroo are high on a given day, foragers pursuing sand
monitors could achieve a lower rate relative to hill kangaroo hunters and the material remains of small
prey would swamp out larger prey (see Figure 4a). Although, this leaves uncertain which component of
return rate is driving this trend. Because the rate of caloric return is determined by total foraging time,
hunting success (a return rate will be 0 if the bout is unsuccessful) and caloric yield, variability in any of
these measures could be driving the effect of overall return rate on AI.
According to the traditional interpretation, the proportion of large prey to small prey should be a
function of the proportional foraging time for each resource or in each patch. The model containing the
total foraging time for each hunt type significantly predicts the abundance index value ( RL2 = 0.48, p =
0.0125; Table 3). However, sand monitor total foraging time has a greater effect. The model suggests that

12

Codding et al. Abundance Indices

if foragers spent a large amount of time hunting sand monitors, they would drive the index down
regardless of how much time hill kangaroo hunters spent foraging. This is illustrated well in Figure 4b.
No matter how much time hill kangaroo hunters spend foraging, as long as foragers depositing remains at
the same camp spend time hunting sand monitor, the index will decline rapidly. This is essentially driven
by the trend shown in Figure 2a, which illustrates that there is no significant effect of foraging time on the
amount of hill kangaroo kilocalories acquired. This overall model (#2) has a lower RL2 and a higher AIC
(21.39) value than the previous model (#1, see Table 3), suggesting that total foraging time does not
predict variability in the abundance index as well as return rate; this is a product of the inclusion of hill
kangaroo foraging time, which does not predict variability in harvest size (Figure 2b). These results imply
that while the total foraging time for small prey may have a strong effect on AI, the effect large prey
hunting on AI might be more a product of hunting bout success or total harvest size (kcal) than foraging
time.
The traditional interpretation of abundance indices does not explicitly consider risk to be a salient
variable in understanding archaeofaunal assemblages, however, implicit in this framework is assumption
that high AI values represent successful hunts with larger prey that leading to higher overall return rates.
Thus, if this traditional interpretation is correct, then hunting success with larger prey should positively
covary with AI. The alternative interpretation suggests the opposite, specifically that high failure rates
with larger prey can lead foragers to focus more on their acquisition as a function of competitions over
prestige - a process that will lead to higher AI values (e.g., McGuire and Hildebrandt 2002). If this is true,
AI values should negatively covary with hunting success. Our results show that hunting bout success does
have a significant and positive effect on AI ( RL2 = 0.97, p < 0.0001; Table 3), supporting the implicit
assumptions of the traditional interpretation. As shown in Figure 4c, when hill kangaroo hunting bout
success is one and sand monitor success is zero, the predicted abundance index value is one; the predicted
index is zero in the opposite scenario. When both hill kangaroo and sand monitor hunters are successful,
the predicted index value falls between one and zero. While this is intuitive, the tight fit of bout success

13

Codding et al. Abundance Indices

and AI (a RL2 value of 0.97) and the low AIC value (11.6) suggests that AI tracks hunting bout success
more than return rate. However, while success rates seem to be accurate in predicting the index value, the
resolution is not precise, creating essentially a step-function from 0 to 1. Since total harvest size is a
continuous, unbound measure, it may more precisely predict changes in the index value.
If the traditional interpretation of abundance indices is correct, then AI values should increase
with hill kangaroo harvest size (total kilocalories acquired) and decrease with sand monitor harvest size.
These predictions are upheld: total harvest size has a strong effect on AI ( RL2 = 0.76, p < 0.0001) and
provides the model with the lowest AIC value (11.07; Table 3). As show in Figure 4d, AI values are only
high when the total hill kangaroo harvest size is high; however, even with a large harvest, the quantity of
hill kangaroo remains can still be swamped by a large sand monitor harvest size. As shown above (section
2), sand monitor harvest size is proportional to the amount of time spent foraging (Figure 2b), but hill
kangaroo hunting is not (Figure 2a), this implies that the quantity of sand monitor remains deposited at a
site are a function of the total amount of foraging time while the quantity of hill kangaroo remains
deposited reflects merely hunting bout success. Because sand monitors are smaller than hill kangaroo,
even a smaller total harvest will disproportionately reduce AI. Moreover, this is accentuated by the fact
that successful hill kangaroo hunts only rarely end up with more than a single carcass (23%; Table 2).
Examining Table 3, the effect of hunting bout success on AI provides the best model according to

RL2 values while the effect of harvest size on AI provides the best model according to AIC values. Taken
together, these results imply that the benefits of acquiring large prey are accurately measured by harvest
size, but the costs are better measured categorically as hunting success (i.e., success or failure) than
continuously as foraging time. Between these two variables, hunting bout success may have the single
strongest effect on AI as whether or not any large prey kilocalories will be acquired varies significantly as
a function of hunting success (see also Bliege Bird et al. 2009). While the model examining the effect of
total foraging time on AI has a low RL2 value and high AIC value, this is probably a function of the
inclusion of large prey foraging time; an examination of Figure 4 shows that the effect of total foraging

14

Codding et al. Abundance Indices

time for small prey has a strong effect on AI. Because harvest size increases linearly with the total
foraging time for small prey, and because an increase in harvest size is directly proportional to an increase
in the number of individual small prey acquired, the total time spent foraging for time small prey has a
dramatic effect on AI. This shows that the effect of small prey hunting on AI is mostly a function of the
total foraging time allocated to that activity or patch on a given day because success is nearly guaranteed
(so any time at equals success), and because the associated rewards in kilocalories (harvest size) is
proportional to the amount of time spent foraging. Thus, the effect of large prey hunting on AI is mostly a
function of hunting bout success while the effect of small prey is mostly a function of foraging time.
These results also show that while return rates may be an appropriate measure for resources characterized
by a linear relationship between foraging time (cost) and harvest size (benefit), the probability that a
hunting bout will end in failure may be more appropriate measure of foraging costs for resources where
the rewards are stochastic.

4.2 How do differences in mens and womens foraging decisions explain variation in abundance indices?

The traditional interpretation suggests that high abundance index values represent a gender
division of labor focused on mens acquisition of larger prey. If this is the case, then abundance index
values should positively covary with mens proportional foraging time and mens caloric contribution
relative to women. However, if the alternative interpretation is correct and high index values represent
mens large game hunting that is subsidized by womens work, then abundance index values should
negatively covary with mens proportional foraging time and their relative caloric contribution
(Hildebradnt and McGuire 2002; McGuire and Hildebrandt 2005).
The amount of time men spend hunting relative to women on each camp-day has a significantly
positive effect on AI (AIC = 30.36, RL2 = 0.32, p < 0.0066; see Table 4, Figure 5a). However, this is
primarily a function of womens foraging time (AIC = 31.23, RL2 = 0.27, p < 0.0156), which

15

Codding et al. Abundance Indices

independently has a much stronger effect on AI than mens foraging time (AIC = 41.94; RL2 < 0.0001, p <
0.8168; see Table 3). This suggests that variability in AI is more a function of womens time spent
hunting for small resources than mens time spent hunting larger resources. As women differentially
target sand monitors over hill kangaroo, when their foraging time increases, AI decreases significantly.
When men contribute larger harvests relative to women, AI is significantly higher (AIC =
26.04, RL2 = 0.41, p < 0.0033; Table 4, Figure 5b). Unlike foraging time, however, when mens and
womens relative contributions are examined independently of one another, this relationship is more a
function of mens harvest size (AIC = 38.33, RL2 = 0.27, p < 0.0171) than womens (AIC = 38.33, RL2 =
0.08, p < 0.1380). As men frequently target larger prey in preference to smaller prey, AI will tend to
increase when mens hunting efforts are successful in acquiring larger prey, leading to a higher caloric
contribution.
Comparing the models explanatory value based on AIC and RL2 values shows that mens
absolute caloric contribution (Model 9; Table 4) is the best gender-specific predictor of AI followed by
womens foraging time (Model 7; Table 4). These results primarily support the traditional interpretation,
suggesting that high AI values are best predicted by large harvests contributed by men and little time
spent pursuing small prey by women. However, because mens hill kangaroo hunting success is
stochastic, so too is their contribution this suggests that high AI values cannot be maintained
consistently. Our previous work shows that mens caloric contribution relative to womens is best
predicted by hunting bout success with large prey (Bliege Bird et al. 2009). When combined, these results
suggest that consistently high abundance index values, representing mens consistent acquisition of larger
prey, can result only when larger prey are predictably acquired. While this is not typically observed
ethnographically (e.g., Hawkes 1991), it still possible if the encounter rates with large prey are high
enough to mitigate pursuit failures. This is discussed below (Section 5).

5 Summary and Discussion

16

Codding et al. Abundance Indices

This examination of Martu foraging and its potential material consequences suggest some
straightforward interpretations of variability in abundance index values. First, high abundance indices
should be a function of both (1a) high frequencies of bout success with larger prey and (1b) little or no
time spent pursuing smaller prey; when combined, this should lead to a high caloric contribution by large
prey and a high overall return rate. Second, low index values then suggest that foragers either (2a) did not
attempt to pursue larger prey at all, (2b) had very low bout success with larger prey, or (2c) spent a
disproportional amount of time hunting smaller prey which swamped the contribution of larger prey; any
of these would lead to a high caloric contribution of smaller prey, and lower overall return rates than if
larger prey were reliably acquired. We discuss and explore these scenarios in greater context and detail
below.

5.1 Traditional and Alternative Interpretations: Efficiency and Risk

Based on these results, the traditional interpretation of abundance indices appears to be mostly
correct. First, high index values do appear to be associated with relatively higher overall return rates,
however, this is only because high AI values represent large prey being acquired reliably. When hunting
success with larger prey is so high that returning from a bout successfully is nearly guaranteed, large prey
will contribute a high proportion of the total harvest and overall return rates will be high. However, if
hunting bout success with larger prey is rare, then continuing to hunt larger prey will result in frequent
failures and thus frequent return rates of zero. Because large game hunting tends to follow this all-ornothing structure, the costs and benefits of acquiring large prey are more accurately described by the
categorical measure of hunting bout success (i.e., success or failure) than the rate of energy acquired over
time. As such, it appears that the link between abundance indices and foraging models has a component
of risk that, while modeled elsewhere (e.g., Winterhalder et al. 1999), has yet to be formalized in the
traditional interpretation of abundance indices. Here the alternative interpretations of abundance indices

17

Codding et al. Abundance Indices

are correct to focus on the risk of failure with larger prey, especially because sensitivities (or
insensitivities) to such risk may be driving some important foraging decisions, particularly those that vary
by gender (Hildebrandt and McGuire 2002; McGuire and Hildebrandt 2005).
Contra the traditional interpretation, we have already shown that for Martu, a preys postencounter return rate does not correlate with prey body size (Bird et al. 2009), however, our data also
show that when hunting success is certain, larger prey may contribute significantly to overall returns
(Bliege Bird et al. 2009; section 4.1 above). Here we suggest that this is a function of stacking pursuit
success probabilities resulting from sequential encounters on a single foraging bout (see also Bliege Bird
et al. 2009; Codding et al. 2010). While pursuing large, highly mobile prey has a higher probability in
ending in failure than pursuing smaller, less mobile prey, this probability of failure is a product of low
post-encounter pursuit success (Bird et al. 2009; see also Hawkes et al. 1991; Jochim 1976; Lee 1968,
1972). Thus, the probability of overall bout success should increase with encounter rates; which may vary
depending on season or climate (Bird et al. 2009; Hawkes et al. 1991; OConnell et al. 1988), changes in
hunting methods (e.g., OConnell et al. 1988), or changes in acquisition technology (e.g., Hill and
Hawkes 1983; OConnell and Hawkes 1984; Winterhalder 1981). In essence, the risk of a failed bout may
be overcome if encounter rates with large prey are high enough that the overall probability of bout
success nears 100% as a consequence of sequential pursuits. If this is correct, high AI values represent
large prey encounter rates high enough to mitigate against the probability of bout failure. What then, does
this imply for archaeological evidence of diachronic declines in the proportion of larger prey?

5.2 Declining Abundance Indices, Broad-Spectrum Revolutions and Patch Choice

A common archaeological pattern uncovered around the world is a diachronic decline in the
proportion of large prey relative to small prey (e.g., Broughton 1994; Cannon 2003; Janestki 1997;
Nagaoka 2002; Stiner et al. 2001).These trends are interpreted to indicate a broadening of the range of

18

Codding et al. Abundance Indices

prey or the types of patches exploited as a result of anthropogenic declines in the encounter rates with
larger, high-ranked resources or the patches in which they occur.
If large and small prey occur simultaneously in a given hunt type or patch, high AI values may be
a function of high encounter rates with larger prey, leading to their successful acquisition on a reliable
basis (see section 5.1). When encounter rates with large prey decline to a point where acquisition is no
longer guaranteed, foragers would have no choice but to switch to alternative, smaller resources within
the same patch, or they would continually return home with nothing. Thus, a decline in AI values may
have less to do with foragers evaluating encounter rates with the highest ranking prey and more about
repeated failures to acquire larger, highly ranked prey. Under such conditions, diachronic declines in AI
may indeed represent prehistoric resource intensification, specifically where foraging shifts from the
reliable acquisition of larger prey to the necessary acquisition of smaller prey resulting in lower overall
return rates.
In heterogeneous environments where large and small resources are not encountered in the same
patch, these trends may be even more pronounced. Declining abundance of patches containing high
ranked prey or declining abundance of high ranked prey within those patches may lead foragers to cease
that activity or cease exploiting that patch completely (Bettinger 1991:90). If AI values were calculated
with large and small prey occurring in two different patches, the archaeological signature of such patch
depletion would result in abrupt declines in abundance index values to zero or near zero. Moreover,
continued foraging time in the patches containing larger prey will have no discernable effect on AI values
as long as some time is spent in the patch containing small prey. This results because any time spent
foraging for small prey will result in a contribution of smaller prey remains proportional to the time spent
in patch (e.g., Figure 2), and consequently a low index value (Figure 4b).
Responses to such depletions, whether within or between patches, may depend on a foragers
degree of risk-sensitivity. While many foragers may want to minimize the risk of acquisition failure, some
may want to take the risk of failure for the potential pay-off of a larger overall harvest size for Martu,
such decisions often vary independent of their effects on mean foraging efficiency (e.g., Bird et al. 2009;

19

Codding et al. Abundance Indices

Bliege Bird and Bird 2008). Since foraging decisions, particularly those involving risk, seem to vary for
men and women, there may a strong gender component to such a process.

5.3 Gender and risk

Our results show that high abundance index values represent low total foraging times by women
and high caloric contributions by mens successful acquisition of large prey; low index values indicate the
opposite. These findings closely match the traditional interpretations view of how gender-differences in
foraging are represented by abundance indices. However, we should not expect such patterns to be static.
In this regard the alternative interpretation is right in turning attention to the interactive effects of mens
and womens forging strategies and how these articulate with variability in the reliability of large game
acquisition.
Recent quantitative ethnographic work suggests that mens foraging behavior is often risk neutral
or risk prone while womens is often risk averse (Bliege Bird 2007; Bliege Bird and Bird 2008; Bliege
Bird et al. 2000, 2001, 2009; Hawkes 1990, 1991; Hill et al. 1987; Sosis 2000; Smith et al. 2000). From
this work, we hypothesize that men are often attempting to maximize harvest size regardless of the risk,
while women are often trying to minimize the risk of failure with less attention to harvest size. The way
that these goals align or conflict with one another depending on the reliability of large game hunting has
serious implications for interpreting abundance indices.
If abundance index values are high, the analysis here suggests that larger prey are being reliably
acquired, and thus, both mens and womens goals being met with the same resources even if women are
not involved in their acquisition. If mens large game acquisition were stochastic, then we would predict
that women would offset this risk by simultaneously foraging for lower risk, alternative resources to
provision self and offspring; representing a more divergent division of labor (as with Martu). However, if
the reliable acquisition of large prey were habitual with success being nearly guaranteed, we might predict
that women would either focus their efforts on hunting larger prey (Kuhn and Stiner 2006) or the post-

20

Codding et al. Abundance Indices

acquisition processing of larger prey (Jarvenpa and Brumbach 2009); representing a more convergent
division of labor. Archaeologically, it should be possible to differentiate these two scenarios. A more
convergent division of labor should be identifiable by a ubiquity of larger prey remains with the near
exclusion of all else, while a more divergent division of labor should be identifiable by the presence of
alternative resources in archaeofaunal or archaeobotanical assemblages and perhaps moderate abundance
index values.
A transition from a more convergent, to a more divergent division of labor may occur as a
function of declining success rates with large prey. If success rates with larger prey declined and men
continued to pursue them, womens foraging goals of risk minimization could only be met better by
targeting smaller, more reliable prey (Hildebrandt and McGuire 2002; Zeanah 2004). This would lead to a
diachronic decline in abundance index values as a function of decreased reliability of mens hunting
success and an increase in the amount of time women spent hunting smaller prey. In this scenario, a
switch from big game to small game, and the corresponding decline in AI, is driven by increased variance
in acquisition, not decreased overall foraging efficiency per se. In such cases, evidence for resource
intensification may be driven by differences in mens and womens foraging goals. Such a scenario may
be identified archaeologically by technology or mobility patterns which suggest a focus on larger prey,
but faunal assemblages that are dominated by smaller prey, suggesting the continued attempts to acquire
large prey despite high failure rates.

6 Conclusion

Abundance indices are valuable zooarchaeological measures that allow researchers to understand
variability in prehistoric human behavior. Given recent disagreements regarding their interpretations, the
aim of this work is to highlight their value by providing a clearer and more accurate interpretation of the
links between such a material measure of human foraging and the actual behavior behind it. We suggest
that while static markers of dynamic interactions, abundance indices represent a limited set of behaviors:

21

Codding et al. Abundance Indices

high index values represent the reliable acquisition of larger prey and little or no time spent on smaller
prey; low index values represent limited or no hunting success with larger prey and a greater amount of
time spent hunting smaller prey. Further, if mens and womens resource choice is determined by the
trade-off between acquisition risk and harvest size, then abundance index values may also represent a
clear maker of gender-differences in foraging strategies.
While the trends shown here may be partially contingent on the particulars of Martu hunting,
there are strong reasons to think that these findings will hold elsewhere. Specifically, we suggest that due
to the common links between prey size, mobility, utility and risk, and how these factors interact with
human foraging decisions, the relative patterns we report here accurately describe variation in abundance
index values.

Acknowledgements

This work would not exist if it werent for the kindness, patience and hospitality of Martu at Parnngurr,
Punmu and Kunawarritji; thanks especially to the Taylor and Morgan families. Thanks also to Peter Veth
and Bob and Myrna Tonkinson for their continued support. An earlier version of this paper was presented
at Stanford University, UC Davis and Sacramento State University; thanks to everyone for their
thoughtful comments, especially Robert Bettinger and David Zeanah. Special thanks to Nathan Stevens
and three anonymous reviewers for thoughtful comments on previous drafts of this paper. This research is
funded by the National Science Foundation (Graduate Research Fellowship, Doctoral Dissertation
Improvement Grant BCS-0915380, and Senior Grants BCS-0127681, -0314406 and -0850664),
Stanfords Department of Anthropology and Archaeology Center.

22

Codding et al. Abundance Indices

References

Akaike, H., 1974. A New Look at the Statistical Model Identification. IEEE Transactions on Automatic
Control 19, 716-723.

Altman, J., 1974. Observational Study of Behavior: Sampling Methods. Behaviour 49, 227-267.

Bayham, F.E., 1979. Factors influencing the Archaic pattern of animal exploitation. Kiva 44, 219-235.

Bayham, F.E., 1982. A diachronic analysis of prehistoric animal exploitation at Ventana Cave.
Unpublished PhD. dissertation, Department of Anthropology, Arizona State University, Tempe, AZ.

Bettinger, R.L., 1991 Hunter-Gatherers: Archaeological and Evolutionary Theory. Plenum Press, New
York.

Bettinger, R.L., 2009 Hunter-Gatherer Foraging: Five Simple Models. Eliot Werner Publications, New
York.

Bird, D.W., Bliege Bird, R. Parker, C., 2005. Aboriginal Burning Regimes and Hunting Strategies
Australias Western Desert. Human Ecology 33, 443-464.

Bird, D.W., OConnell, J.F., 2006. Behavioral Ecology and Archaeology. Journal of Archaeological
Research 14, 143-188.

Bird, D.W., Bliege Bird, R., Codding, B.F., 2009. In Pursuit of Mobile Prey: Martu Foraging and
Archaeofaunal interpretation. American Antiquity 74, 3-29.

23

Codding et al. Abundance Indices

Bliege Bird, R., 2007. Fishing and the sexual division of labor among the Meriam. American
Anthropologist 109, 442-451.

Bliege Bird, R., Bird, D.W., 2008. Why women hunt: risk and contemporary foraging in a Western Desert
Aboriginal community. Current Anthropology 49, 655-693.

Bliege Bird, R., Smith, E.A., 2000. Turtle hunting and tombstone opening: public generosity as costly
signaling. Evolution and Human Behavior 21, 245-261.

Bliege Bird, R., Smith, E.A., Bird, D.W., 2001. The hunting handicap: Costly signaling in male foraging
strategies. Behavioral Ecology and Sociobiology 50, 9-19.

Bliege Bird, R. Bird, D.W., Codding, B.F., Parker, C.H., Jones, J.H., 2008. The fire stick farming
hypothesis: Australian Aboriginal foraging strategies, biodiversity, and anthropogenic fire mosaics.
Proceedings of the National Academy of Sciences 105, 14796-14801.

Bliege Bird, R., Codding, B.F., Bird, D.W., 2009. Determinants of gendered foraging and production
inequalities among Martu. Human Nature 20, 105-129.

Brand Miller, J., James, K.W., Maggiore, P.M.A., 1993. Tables of Composition of Australian Aboriginal
Foods. Aboriginal Studies Press, Canberra.

Broughton, J.M., 1994. Declines in foraging efficiency during the late Holocene: The archaeological
mammal evidence from San Francisco Bay, California. Journal of Anthropological Archaeology 13, 371401.

24

Codding et al. Abundance Indices

Broughton, J.M., Bayham, F.E., 2003. Showing off, Foraging Models, and the Ascendance of LargeGame Hunting in the California Middle Archaic. American Antiquity 68, 783-789.

Broughton, J.M., Byers, D.A., Bryson, R.A., Eckerle, W., Madsen, D.B., 2008. Did climatic seasonality
control late Quaternary artiodactyl densities in western North America? Quaternary Science Reviews 27,
1916-1937.

Burnham, K.P., Anderson D., 2002. Model Selection and Multi-Model Inference: A practical
information-theoretic approach, second ed. Springer, New York.

Cannon, M.D., 2003. A model of central place forager prey choice and an application to faunal remains
from the Mimbres Valley, New Mexico. Journal of Anthropological Archaeology 22, 1-25.

Charnov, E.L., Orians, G.H., 1973. Optimal Foraging: Some Theoretical Explorations. Mimeo,
Department of Biology, University of Utah, Salt Lake City.

Codding, B.F., Porcasi, J.F., Jones, T.L., 2010 Explaining prehistoric variation in the abundance of large
prey: A zooarchaeological analysis of deer and rabbit hunting along the Pecho Coast of Central
California. Journal of Anthropological Archaeology 29, 47-61.

Davenport, S., Johnson, P., Yuwali, 2005. Cleared Out: First Contact in the Western Desert. Aboriginal
Studies Press, Canberra.

Faraway, J.J., 2006. Extending the Linear Model with R: Generalized Linear, Mixed Effects and
Nonparametric Regression Models. Chapman and Hall, New York.

25

Codding et al. Abundance Indices

Grayson, D.K., 1984. Quantitative Zooarchaeology: Topics in the Analysis of Archaeological Faunas.
Academic Press, New York.

Griffiths, D., 1975. Prey Availability and the Food of Predators. Ecology 56, 1209-1214.

Hawkes, K., 1991. Showing off: Tests of an hypothesis about mens foraging goals. Ethology and
Sociobiology 12, 29-54.

Hawkes, K., 1993. Why Hunter-Gathers Work: An Ancient Version of the Problem of Public Goods.
Current Anthropology 34, 341-361.

Hawkes, K., O'Connell, J.F., Blurton Jones, N., 1991. Hunting income patterns among the Hadza: Big
game, common goods, foraging goals and the evolution of the human diet. Philosophical Transactions:
Biological Sciences 334, 243-251.

Hildebrandt, W.R., Jones, T.L., 1992. Evolution of Marine Mammal Hunting: A View from the California
and Oregon Coasts. Journal of Anthropological Archaeology 11: 360-401.

Hildebrandt, W.R., McGuire, K.R., 2002. The ascendance of hunting during the California Middle
Archaic: An evolutionary perspective. American Antiquity 67, 231-256.

Hill, K., Hawkes, K., 1983. Neotropical hunting among the Ache of eastern Paraguay. In R. Hames & W.
Vickers (Eds.), Adaptive Responses of Native Amazonians. Academic Press, New York, pp. 139-188.

26

Codding et al. Abundance Indices

Hill, K., Kaplan, H., Hawkes, K., Hurtado, A.M., 1987. Foraging Decisions Among Ach HunterGatherers: New Data and Implications for Optimal Foraging Models. Ethology and Sociobiology 8, 1-36.

Isaac, G., 1978. The Food Sharing Behavior of Proto-Human Hominids. Scientific American 238, 90-108.

Janetski, J., 1997. Fremont hunting and resource intensification in the eastern Great Basin. Journal of
Archaeological Science 24, 1075-1089.

Jarvenpa, R., Brumbach, H.J., 2009.Ethnoarchaeology, circumpolar toolkits, and gender inequality.
Ethnoarchaeology 1, 57-78.

Jochim, M.A., 1976. Hunter-Gatherer Subsistence and Settlement: A Predictive Model. Academic Press,
New York.

Jochim, M.A., 1988. Optimal Foraging and the Division of Labor. American Anthropologist 90, 130-136.

Jones, T.L., Porcasi, J.F., Erlandson, J.M., Dallas Jr., H., Wake, T.A., Schwaderer, R., 2008a. The
Protracted Holocene Extinction of California 's Flightless Sea Duck, Chendytes lawi and its Implications
for the Pleistocene Overkill Hypothesis. Proceedings of the National Academy of Science 105, 41054108.

Jones, T.L., Porcasi, J.F., Gaeta, J.W., Codding, B.F., 2008b. The Diablo Canyon fauna: a course-grained
record of trans-Holocene foraging from the central California coast. American Antiquity 73, 289-316.

Kieschnick, R., McCullough, B.D., 2003. Regression analysis of variates observed on (0,1): percentages,
proportions and fractions. Statistical Modeling 3, 193-213.

27

Codding et al. Abundance Indices

Klein, R.G., Cruz-Uribe, K., 1984. The Analysis of Animal Bones from Archaeological Sites. University
of Chicago Pres.

Kuhn, S., Stiner, M.C., 2006. Whats a mother to do? The division of labor among Neanderthals and
modern humans in Eurasia. Current Anthropology 47, 953-980

Lee, R.B., 1968. What hunters do for a living: or, how to make out on scarce resources. In R.B. Lee & I.
Devore (Eds.) Man the Hunter. New York: Aldine De Gruyter, pp. 30-55.

Lee, R. B., 1979. The !Kung San: Men, Women and Work in a Foraging Society. Cambridge University
Press, Cambridge.

Lupo, K.D., 2007. Evolutionary foraging models in zooarchaeological Analysis: recent applications and
future challenges. Journal of Archaeological Research 15, 143-189.

Lupo, K.D., Schmitt, D.N., 2002. Upper Paleolithic net-hunting, small prey exploitation and womens
work effort: a view from the ethnographic and ethnoarchaeological record of the Congo basin. Journal of
Archaeological Method and Theory 9, 147-179.

Lupo, K.D., Schmitt, D.N., 2005. Small prey hunting technology and zooarchaeological measures of
taxonomic diversity and abundance: Ethnoarchaeological evidence from Central African forest foragers.
Journal of Anthropological Archaeology 24, 335-353.

MacArthur, R.H., Pianka, E.R., 1966. On Optimal Use of a Pathcy Environment. The American
Naturalist 100, 603-609.

28

Codding et al. Abundance Indices

Madsen, D.B., Kirkman, J., 1988 Hunting Hoppers. American Antiquity 53, 593604.

Madsen, D.B., Schmitt, D.N., 1998 Mass Collecting and the Diet Breadth Model: A Great
Basin Example. Journal of Archaeological Science 25, 445457.

McGuire, K., Hildebrandt, W.R., 1994. The Possibilities of Women and Men: Gender and the California
Milling Stone Horizon. Journal of California and Great Basin Anthropology 16, 41-59.

McGuire, K., Hildebrandt, W.R., 2005. Re-thinking Great Basin foragers: Prestige hunting and costly
signaling during the Middle Archaic period. American Antiquity 70, 695-712.

McGuire, K., Hildebrandt, W.R., Carpenter, K.L., 2007. Costly Signaling and the Ascendance of No-CanDo Archaeology: A Reply to Codding and Jones. American Antiquity 72, 358-365.

Menard, S., 2002. Applied Logistic Regression Analysis, Second ed. Sage Publications, Thousand Oaks.

Nagaoka, L. 2002. The effects of resource depression on foraging efficiency, diet breadth, and patch use
in southern New Zealand. Journal of Anthropological Archaeology 21, 419442

OConnell, J.F., 1995. Ethnoarchaeology Needs a General Theory of Behavior. Journal of Archaeological
Research 3, 205-255.

OConnell, J. F., Hawkes, K., 1984. Food Choice and Foraging Sites among the Alyawara. Journal of
Anthropological Research 40, 504-535.

29

Codding et al. Abundance Indices

OConnell, J.F. and Marshall, B., 1989. Analysis of Kangaroo Body Part Transport Among the Alyawara
of Central Australia. Journal of Archaeological Science 16, 393-405.

OConnell, J. F., Hawkes, K., Blurton Jones, N., 1988. Hadza hunting, butchering, and bone transport and
their archaeological implications. Journal of Anthropological Research 44, 113161.

R Development Core Team 2010. R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria.

Sarkar, D. 2008. Lattice: Multivariate Data Visualization with R, Springer, New York.

Scelza, B., Bliege Bird, R., 2008. Group structure and female cooperative networks in Australias
Western Desert. Human Nature 19, 231248.

Schoener, T. W., 1971. Theory of Feeding Strategies. Annual Review of Ecology and Systematics 2, 369404.

Smith, E.A., 1991. Inujjuamiut Foraging Strategies: Evolutionary Ecology of an Arctic Hunting
Economy. Aldine, New York.

Smith, E.A., Bliege Bird, R., 2000. Turtle hunting and tombstone opening: public generosity as costly
signaling. Evolution and Human Behavior 21, 245-261.

Sosis,R., 2000. Costly signaling and torch fishing on Ifaluk atoll Evolutuion and Human Behavior 21,
223-244.

30

Codding et al. Abundance Indices

Stevens, D.W., Krebs, J.R., 1986. Foraging Theory. Princeton University Press.

Stiner, M.C., 2001. Thirty Years on the Broad Spectrum Revolution and Paleolithic Demography.
Proceedings of the National Academy of Sciences 98, 6993-6996.

Stiner, M.C., Munro, N.D., Survoell, T.A., 2000. The Tortoise and the Hare: Small-Game Use, the BroadSpectrum Revolution, and Paleolithic Demography. Current Anthropology 41, 39-73.

Stutz, A.J., Munro, N.D., Bar-Oz, G., 2009. Increasing the Resolution of the Broad Spectrum Revolution
in the Southern Levantine Epipaleolithic (19-12 ka). Journal of Human Evolution 56, 294-306.

Tonkinson, R., 1974. The Jigalong Mob: Aboriginal Victors of the Desert Crusade. Cummings, Menlo
Park.

Tonkinson, R., 1991. The Mardu Aborigines: Living the Dream in Australias Desert. Holt, Rinehart &
Winston, New York.

Ugan, A., 2005a. Does Size Matter? Body Size, Mass Collecting, and Their Implications for
Understanding Prehistoric Foraging Behavior. American Antiquity 70, 75-89.

Ugan, A. 2005b. Climate, bone density, and resource depression: What is driving variation in large and
small game in Fremont archaeofaunas? Journal of Anthropological Archaeology 24, 227-251

Venables, W. N., Ripley, B. D. 2002. Modern Applied Statistics with S. Fourth ed. Springer, New York.

31

Codding et al. Abundance Indices

Waguespack, N. M., 2005. The Organization of Male and Female Labor in Foraging Societies:
Implications for Early Paleoindian Archaeology. American Anthropologist 107, 666-676.

Winterhalder, B., 1981. Foraging Strategies in the Boreal Forest: An Analysis of Cree Hunting and
Gathering. In Winterhalder, B., Smith, E.A. (Eds.), Hunter-Gatherer Foraging Strategies: Ethnographic
and Archaeological Analyses. University of Chicago Press, pp. 66-98.

Winterhalder, B., Lu, F., Tucker, B., 1999. Risk-Sensitive Adaptive Tactics: Models and Evidence from
Subsistence Studies in Biology and Anthropology. Journal of Archaeological Research 7, 301-348.

Zeanah, D.W., 2004. Sexual Division of Labor and central place foraging: a model for the Carson Desert
of Western Nevada. Journal of Anthropological Archaeology 23, 1-32.

32

Codding et al. Abundance Indices

Figures

Figure 1. Maps (a) of the continent situating the Martu Native Title in (b) Western Australia including all temporary
camp locations examined in this study with (c) key terrain features.

Figure 2. Total harvest size (kcal) as a function of total foraging time (min) for each recorded bout of (a) hill
kangaroo and (b) sand monitor hunting.

Figure 3. A plot of an empirical cumulative density function describing the probability of achieving a given return
rate on hill kangaroo and sand monitor hunts.

Figure 4. Surface plots showing the effects of the (a) overall return rate (E/T), (b) total foraging time (TFT), (c)
hunting bout success and (d) harvest size (kcal) for hill kangaroo and sand monitor hunting on abundance index
values.

Figure 5. Abundance index values as a function of mens (a) foraging time and (b) harvest size relative to womens
for each camp-day.

33

Table 1. Summary of foraging measures for each hunt type per bout . a
Variable
Hunt Type
N Median Mean
SD
Total foraging time (min)
Hill kangaroo
125
184.00
182.12
84.76
Sand monitor
690
191.00
185.83
76.37
Total harvest size (kcal)
Hill kangaroo
125
0.00 3755.08 12230.00
Sand monitor
690 1535.89 1888.64
1663.80
E/T (kcal/min)
Hill kangaroo
125
0.00
21.04
67.38
Sand monitor
690
8.92
10.59
8.70
Overall Success
Hill kangaroo
125
0.00
0.15
0.36
Sand monitor
690
1.00
0.89
0.31
a
data from 1516 foraging bouts recorded from 2000 to 2009 (see Section 2).

Codding et al. Abundance Indices

Table 2. Summary of the database comprising 36 camp-days. a

Sum
Median
Mean
Hill kangaroo mean return rate (kcal/min)
0
35.39
Sand monitor mean return rate (kcal/min)
9.76
9.58
Hill kangaroo total foraging time (min)
14494
275
402.61
Sand monitor total foraging time (min)
18769
365
521.36
Hill kangaroo hunting bout success
13
0
0.36
Sand monitor hunting bout success
29
1
0.86
Hill kangaroo total harvest size (kcal)
437783
0 12160.63
Sand monitor total harvest size (kcal)
195716 3662.82
5436.56
% Men's caloric contribution
0.23
0.39
% Men's foraging time
0.49
0.54

SD
62.20
7.07
399.72
461.77
0.49
0.40
20923.93
5023.98
0.40
0.27

Hill kangaroo acquired ("MNI")

17
0
0.47
0.74
Sand monitor acquired ("MNI")
217
4
6.03
6.12
Abundance Index
0
0.23
0.40
a
The summary data presented here characterizes foraging results for these 36 camp-days,
not foraging results for these hunt types overall. For such characterizations, see Table 1,
Figure 2, Figure 3, Bird et al. (2009), Bliege Bird and Bird (2008) and Bliege Bird et al.
(2009).

35

Table 3. Model results examining the effects of quantitative foraging measures on the abundance index values.
DM a
GM a
Est.
DF AIC
#
1

R2L

pb

Hill kangaroo and sand monitor mean return rates (whole model)
Intercept
Hill kangaroo mean return rate (E/T; kcal/min)
Sand monitor mean return rate (E/T; kcal/min)

0.32
0.02
-0.49

2
-

12.74
-

6.42
-

27.34
-

0.81
-

< 0.0001
0.8088
0.0499
0.0198

Hill kangaroo and sand monitor total foraging time (whole model)
Intercept
Hill kangaroo total foraging time (min)
Sand monitor total foraging time (min)

0.85
0.00
-0.01

2
-

21.39
-

17.31
-

16.45
-

0.49
-

0.0125
0.3149
0.9670
0.0306

Hill kangaroo and sand monitor hunting bout success (whole model)
Intercept
Hill kangaroo hunting bout success
Sand monitor hunting bout success

1.16
22.41
-24.72

2
-

11.60
-

1.16
-

32.60
-

0.97
-

< 0.0001
1.0000
0.9990
0.9990

Hill kangaroo and sand monitor total harvest size (whole model)
Intercept
Hill kangaroo total harvest size (kcal)
Sand monitor total harvest size (kcal)

-0.39
0.00
0.00

2
-

11.07
-

8.05
-

25.55
-

0.76
-

< 0.0001
0.7466
0.0846
0.0395

a
b

For definitions, see Menard (2002). The null deviance (D0) for each model is 33.76.
Significant values (p < 0.05) are marked with an asterisk

*
*

.
*

Codding et al. Abundance Indices

Table 4. Model results examining the effects of gender-differences in foraging on the abundance index values
DM a
GM a
pb
R2L
Est. DF AIC
#
5
6
7
8
9
10

Men's relative foraging time

Men's foraging time
Women's foraging time
Men's relative contribution
Men's contribution
Women's contribution

5.69
0.00
0.00
4.63
0.00
0.00

1
1
1
1
1
1

30.36
41.94
31.23
26.04
30.05
38.33

22.93
33.70
24.54
18.98
24.68
30.90

10.83
0.06
9.22
14.78
9.08
2.86

0.32
0.00
0.27
0.44
0.27
0.08
R

a
b

0.0066
0.8168
0.0156
0.0033
0.0171
0.1380

*
*
*
*

For definitions, see Menard (2002). The null deviance (D0) for each model is 33.76.
Significant values (p < 0.05) are marked with an asterisk

37

12000
10000

80000

8000

60000

4000

6000

40000

2000

20000

Total Harvest Size (kcal)

(b) Sand monitor

(a) Hill kangaroo

100

200

300

400

100

Total Foraging Time (min)

200

300

400

1.0
0.6
0.4
0.2

Sand Monitor

0.0

Probability

0.8

Hill Kangaroo

5000

10000

15000

20000

Overall Return Rate (kcal/hr)

25000

Hill Kangaroo E/T

Sand Monitor E/T

Hill Kangaroo TFT

Sand Monitor TFT

AI

(a)

AI

(b)

Hill Kangaroo Success

Sand Monitor Success

AI

(c)

Hill Kangaroo KCAL

AI

(d)

Sand Monitor KCAL

(a)

(b)

1.0

0.8

AI

0.6

0.4

0.2

0.0
0.2
0.4
0.6
0.8
1.0
0.0
Men's Total Foraging Time Relative to Womens

0.0

0.2
0.4
0.6
0.8
1.0
Men's Contribution (kcals) Relative to Womens