Professional Documents
Culture Documents
net/publication/236099834
A Comparison of Scan and Focal Sampling for the Description of Wild Primate
Activity, Diet and Intragroup Spatial Relationships
CITATIONS READS
10 7,615
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Sarie Van Belle on 23 December 2014.
Key Words
Alouatta · Activity budget · Instantaneous focal sampling · Continuous focal
sampling · Group scan · Proximity · Nearest neighbor
Abstract
We used data collected during two concurrent studies of black howler monkeys
(Alouatta pigra) in Palenque National Park, Mexico, to compare systematically three
methods of behavioral data collection [group activity scan sampling (group scans), in-
stantaneous focal individual sampling (instantaneous focals) and continuous focal indi-
vidual sampling (continuous focals)] and three methods of proximity data collection
[group proximity scan sampling (group proximity scans), focal individual proximity scan
sampling (focal proximity scans) and instantaneous focal individual nearest neighbor
sampling (focal nearest neighbor samples)]. We conducted pairwise comparisons of
data among methods using Pearson correlations and one-sample t tests. A series of
Kruskal-Wallis tests were performed to compare the activity and proximity patterns of
adult males, adult females and juveniles described by each method. The three behav-
ioral data collection methods generally provided similar information about activity and
diet. However, important differences for both activity and proximity data existed among
methods. Instantaneous focals overestimated the percentage of time spent in social in-
teractions, while group scans overestimated time spent moving and underestimated
time spent feeding. Group proximity scans and focal proximity scans provided similar
spatial data, while focal nearest neighbor samples were more appropriate for determin-
ing the influence of one individual on another at any given moment. These biases sug-
gest the importance of deliberate method selection during project design and highlight
the need for taking methods into account when comparing studies.
Copyright © 2013 S. Karger AG, Basel
Focal individual observations and scan sampling are used almost universally by
primatologists to describe non-human primate behavior in a variety of wild and cap-
tive contexts. Scan samples are traditionally used to survey activity budget, diet, or
spatial arrangements of all or a subgroup of individuals in a social group simultane-
ously at specified time intervals. In contrast, focal individual samples are primarily
used to collect detailed data describing behaviors that might be overlooked by meth-
ods that do not require the observer to follow an individual in all situations. Altmann
[1974] was the first to outline primate focal and scan sampling in detail. According
to her, data describing complex behaviors (e.g. social interactions) are difficult to re-
cord with scan samples due to the challenge of locating and identifying the activity of
each individual in a group in a short amount of time. However, she notes that research
requiring information regarding easily recognizable behaviors and/or group syn-
chrony of behaviors can benefit greatly from scan sampling. In contrast, focal data
may be more useful for detecting behavioral differences between individuals in a so-
cial group or for describing patterns of subtle social interactions. Focal samples can
be recorded continuously or instantaneously, and how well the two methods corre-
spond depends on the interval between instantaneous records of behavior during a
focal sample, the type of behavioral data being recorded, and the frequency at which
individuals of the species being observed shift behaviors [Altmann, 1974].
Despite the general acceptance of focal and scan sampling for behavioral data
collection, little work has been done formally to compare them. Although the litera-
ture commonly cited [Altmann, 1974; Martin and Bateson, 1993; Lehner, 1996; Pat-
erson, 2001] by authors employing these methods describes them in great detail and
suggests ideal contexts for each, it does not provide systematic comparisons to sup-
port these suggestions and to guide researchers more effectively in project design. In
fact, comparisons of methods for collecting activity and diet data are rare in the pri-
mate literature in general [Dunbar, 1976; Rhine and Flanigon, 1978; Rhine and Ender,
1983; Bernstein, 1991; Fragaszy et al., 1992; Kawanaka, 1996; Zinner et al., 1997; Rose,
2000; Stevenson and Quinones, 2004; Hernández-Lloreda and Colmenares, 2006; Ed-
wards et al., 2010; Gilby et al., 2010], and studies comparing methods for describing
spatial proximity patterns among individuals are completely absent (but see Haddadi
et al. [2011] for a discussion on temporal and spatial criteria for accurately construct-
ing social networks). Because data describing interindividual relationships may de-
pend heavily on proximity data, especially in primates with low frequencies of social
interaction [Di Fiore and Fleischer, 2005; Van Belle et al., 2008; Takuda et al., 2012],
a thorough understanding of the techniques available for estimating spatial relation-
ships among primate individuals is critical. Without systematic comparisons of these
techniques, such an understanding is impossible.
The aim of this paper is to evaluate several commonly used primate behavioral
sampling techniques in the field. Specifically, we compare the use of group activity
scan sampling (hereafter group scans), continuous focal individual sampling (here-
after continuous focals), and instantaneous focal individual sampling (hereafter in-
stantaneous focals) in describing activity and diet in two groups of wild black howler
monkeys (Alouatta pigra) in Palenque National Park, Mexico. We also compare the
use of group proximity scan sampling (hereafter group proximity scans), focal indi-
vidual proximity scan sampling (hereafter focal proximity scans), and instantaneous
Methods
Data Collection
Behavioral data collection from the two research projects used three observation methods,
each of which was associated with its own protocol for recording proximity among group mem-
bers. For the purpose of the comparisons reported in this paper, we grouped activities into five
categories: feeding, howling, moving, resting, and engaging in social interaction. During feeding
bouts, we identified the food item and analyzed data from five food categories (young and mature
leaves, young and mature fruits, and stems) for the comparison among data collection methods.
Fig. 1. Composition of the two study groups during the study period, excluding infants. Black
circles indicate dispersal of group members. AM = Adult male; AF = adult female; JM = juvenile
male; JF = juvenile female. MIL and DOS were considered adults by June 2011. Both groups in-
cluded infants that were excluded from analyses.
Categories were also used to describe the distance between individuals (except infants) for each
proximity method. This study considers two of these spatial categories: 0–1 and >1–5 m apart.
We limited ourselves to these categories because both research teams used boundaries of 1 and
5 m. We did not include a category for greater distances because the different proximity methods
did not have consistent standards of identifying individuals >5 m apart.
(1) Group Scans and Group Proximity Scans. Instantaneous scan samples were conducted at
15-min intervals to record individual activity data. Three minutes were allotted for the location
of all non-infant group members. The first observed behavior lasting for ≥5 s was recorded for
each group member to avoid a potential bias towards more readily detected activities (e.g. mov-
ing) [Fragaszy et al., 1992; Kortsjens et al., 2005]. A total of 2,241 group scans were completed
(Balam: 1,174 group scans; Motiepa: 1,067 group scans). During each scan sample, the distance
categories between all dyads of non-infant group members were recorded as well. To prevent bi-
asing our results in favor of the most visible group members, we only included scan samples for
which complete data on all non-infant group members were recorded, representing a total of 820
group proximity scans (Balam: 413 group scans, 48.6% of all Balam scans; Motiepa: 407 group
scans, 60.1% of all Motiepa scans).
(2) Continuous Focals and Focal Proximity Scans. Thirty-minute focal individual samples
were conducted to record continuously (to the nearest second) the activity of adults in each
group. Each individual was sampled once before any individual was sampled a second time on
a given day. When focal subjects were out of view for >10 min, the sample was ended, and
focal samples lasting <20 min were discarded. We analyzed 360 focal hours of behavior (Balam:
192.3 h; Motiepa: 167.8 h). Instantaneous proximity scan samples were carried out during the
first 5 s of each focal sample as well as at 15 and 30 min to record the identity and distance cate-
gory of all neighbors of the focal individual. We included only scan samples for which we could
identify all non-infant neighbors, which resulted in 1,691 scan samples (Balam: 893 scan samples;
Motiepa: 798 scan samples).
(3) Instantaneous Focals and Focal Nearest Neighbor Samples. Twenty-minute focal indi-
vidual samples with activity recorded instantaneously at 2-min intervals were used to describe
the behavior of the focal individual. If the focal individual was not visible for more than two time
Data Analyses
For each data collection procedure, the average percentage of time spent monthly in each
activity and consuming each food type was calculated for all focal individuals (juveniles not in-
cluded in continuous focals). Using these average monthly percentages for each individual, we
then calculated the monthly percentage of time spent in each activity and consuming each food
type for the group overall. To assess how well data collected using each method corresponded,
we compared monthly group averages from each method in a pairwise fashion. For pairwise com-
parisons including continuous focals, monthly means from both methods considered were cal-
culated only from adult individuals.
Because monthly averages for activity and diet approximated a normal distribution, we used
parametric tests to compare them. We first determined a Pearson correlation coefficient for the
average monthly percentage of time spent in each activity and consuming each food type for each
pair of methods [Fragaszy et al., 1992]. Significant correlations indicated high similarity in the
overall patterns described by each method. p values were corrected for multiple comparisons us-
ing a sequential Bonferroni procedure for α = 0.05 [Holm, 1979; Rice, 1989]. Because activity and
diet data were considered separately for each pairwise comparison of methods, p values were ad-
justed for the five tests (one for each activity or diet category) performed on each of these datasets.
Because it is possible for two datasets to be correlated but to report distinct values for each
data point, we also calculated the difference between the monthly average percentage of time
spent in each activity per pair of methods and used one-sample t tests to determine whether these
differences varied significantly from zero [Gilby et al., 2010]. If the methods reported similar val-
ues, we would expect the differences among monthly average values not to differ from zero. Be-
cause we wanted to detect failures to reject the null hypothesis for these tests, p values were not
corrected with sequential Bonferroni corrections to avoid inflating the rate of false negatives.
Because each group of howlers showed the same activity and diet patterns, data from both groups
were combined for both sets of tests.
Similarly, for each procedure, the average percentage of time spent at 0–1 m and at >1–5 m
was calculated per dyad of individuals per month for the two study groups. Using these averages,
we then calculated monthly averages for adult male-adult male dyads, adult female-adult female
dyads, adult male-adult female dyads, adult male-juvenile dyads, and adult female-juvenile dyads
in each group. As with the activity data, the monthly averages approximated a normal distribu-
tion and were compared across methods in a pairwise fashion using Pearson correlations and
one-sample t tests. Because both distance categories were considered part of the same proximity
dataset, correlation p values were adjusted using a sequential Bonferroni procedure to reflect the
ten tests (one for each age-sex dyad at each distance) performed for each pairwise comparison.
p values for t tests were not corrected.
To further assess similarities and differences among the data collection procedures, we ex-
amined whether adult males, adult females and juveniles differed in the monthly mean percent-
age of time spent in each activity and consuming each food type using a series of Kruskal-Wallis
tests for each method separately. The more similar the methods, the fewer differences we would
expect in the behavior patterns reported by each method. For example, if group scans report
males howling significantly more time on average than females, both continuous and instanta-
neous focals should also report males howling significantly more frequently than females. Simi-
larly for the proximity data, the average time spent at each distance category was compared for
Activity
Feeding 0.908 0.000 0.837 0.000 0.965 0.000
Howling 0.517 0.070 0.340 0.256 0.859 0.000
Moving 0.652 0.016 0.526 0.065 0.608 0.027
Resting 0.838 0.000 0.811 0.001 0.898 0.000
Social interaction 0.753 0.003 0.760 0.003 0.893 0.000
Food type
Young leaves 0.768 0.002 0.508 0.076 0.892 0.000
Mature leaves 0.700 0.008 0.594 0.032 0.927 0.000
Young fruits 0.380 0.201 0.977 0.000 0.480 0.096
Mature fruits 0.893 0.000 0.703 0.007 0.816 0.001
Stems 0.781 0.002 0.613 0.026 0.927 0.000
n = 13 for each test (6- and 7-monthly averages for Balam and Motiepa, respectively). Bold
p values indicate significant correlations between two data collection methods after sequential
Bonferroni corrections with the initial α value set at 0.050.
male-male, female-female, male-female, male-juvenile, and female-juvenile dyads for each meth-
od separately. Small sample sizes and a poor approximation of the normal distribution required
the use of non-parametric Kruskal-Wallis tests for these analyses. Data for these tests were com-
bined for both groups when there were no significant differences between groups, and a sequen-
tial Bonferroni correction was applied to all p values (activity and diet: five categories with three
age-sex pairwise comparisons each; proximity: two categories with five dyad comparisons each).
Results
Activity
The average time spent feeding, resting and being social correlated strongly
among all methods (table 1). However, time spent moving was more weakly corre-
lated among all methods, and time spent howling was weakly correlated between in-
stantaneous focals and continuous focals and between instantaneous focals and group
scans (table 1). t tests indicated several differences for average monthly activity bud-
gets between group scans and both focal methods. Specifically, group scans showed
that the howler monkeys spent on average less time feeding and being social and more
time moving when compared to instantaneous focals (table 2). Group scans also
showed that the howler monkeys spent less time feeding and more time moving than
continuous focals (table 2). Instantaneous focals indicated that the howlers spent
more time engaged in social interactions compared to both continuous focals and
group scans (table 2).
Activity
Feeding –0.031 ± 0.032 –3.48 0.005 –0.013 ± 0.045 –1.11 0.288 –0.036 ± 0.024 –5.48 0.000
Howling 0.005 ± 0.014 1.50 0.161 0.005 ± 0.022 –0.74 0.472 0.004 ± 0.012 1.09 0.298
Moving 0.077 ± 0.035 7.98 0.000 0.004 ± 0.018 –0.71 0.489 0.044 ± 0.026 6.05 0.000
Resting –0.030 ± 0.046 –2.36 0.036 0.011 ± 0.051 0.79 0.444 –0.016 ± 0.039 –1.48 0.165
Social interaction –0.022 ± 0.019 –4.29 0.001 0.011 ± 0.018 2.24 0.045 0.004 ± 0.008 1.72 0.112
Food type
Young leaves 0.048 ± 0.128 1.35 0.202 –0.084 ± 0.185 –1.63 0.128 –0.01 ± 0.094 –0.52 0.615
Mature leaves –0.007 ± 0.039 –0.68 0.508 –0.045 ± 0.076 –2.16 0.052 –0.04 ± 0.049 –2.95 0.012
Young fruits –0.008 ± 0.070 –0.42 0.684 0.006 ± 0.024 0.88 0.397 –0.01 ± 0.080 –0.40 0.698
Mature fruits 0.050 ± 0.068 2.65 0.021 –0.053 ± 0.110 –1.76 0.104 –0.01 ± 0.093 –0.30 0.766
Stems –0.013 ± 0.069 –0.67 0.516 0.027 ± 0.110 0.87 0.402 0.00 ± 0.047 0.01 0.994
n = 13 for each test. Bold p values indicate significant differences between two data collection methods (α = 0.050).
Despite some differences in the actual percentage of time spent in each activity
shown by each method, all methods indicated that the howlers spent the largest per-
centage of time resting, followed by feeding, and then moving, social interactions, and
howling. There were also no differences in activity patterns between adult sex classes
regardless of the method being used (all p > 0.05), and all methods showed differences
in activity budget according to age class (fig. 2). Both group scans and instantaneous
focals showed that juveniles spent more time moving and being social and less time
howling and resting than adults. Neither method showed age differences in the time
spent feeding.
Diet
Although the amount of time spent feeding was correlated among all observation
methods, the percentage of feeding time dedicated to different food types was not.
The average percentage of total feeding time dedicated to young fruits was not strong-
ly correlated between group scans and continuous focals or between group scans and
instantaneous focals (table 1). The average percentage of total feeding time spent con-
suming young and mature leaves and stems was not correlated between the two focal
observation methods (table 1). However, the differences between the values calcu-
lated using each observation method did not differ significantly from zero in most
cases (table 2). The two exceptions were that group scans indicated more time spent
consuming mature fruits than instantaneous focals and more time spent consuming
mature leaves than continuous focals.
When we examined feeding patterns using each method, both groups spent the
same amount of time feeding on most food types (all p > 0.05) except young leaves
(p = 0.004) and stems (p = 0.003). Data for both groups were analyzed together for all
categories except young leaves and stems. Comparisons of feeding patterns using
Monthly rate
0.20
a a 0.60 b
0.15
0.40
0.10
0.05 0.20
0 0
GRP SCN INST FCL CONT FCL GRP SCN INST FCL CONT FCL
Moving Howling
0.25 0.05 a a
b
0.20 0.04 a
Monthly rate
Monthly rate
0.15 0.03
a a b b
0.10 0.02
a a b
a a
0.05 0.01
c b b
0 0
GRP SCN INST FCL CONT FCL GRP SCN INST FCL CONT FCL
Social interactions
0.10 c
0.08 b
Monthly rate
0.06
0.04
a a
0.02 a a a
b
0
GRP SCN INST FCL CONT FCL
Fig. 2. Monthly rates of feeding, moving, resting, howling, and engaging in social interactions by
adult males (dark gray bars), adult females (white bars), and juveniles (light gray bars) for data
collected via group scan (GRP SCN), instantaneous focals (INST FCL), and continuous focals
(CONT FCL). Statistically significant differences among the age classes within each data collec-
tion method are indicated by different letters (a, b, or c). n = 13 for each bar. Means ± SE.
these data revealed no significant differences in time spent consuming any food type
among age or adult sex classes for any method (fig. 3).
Proximity
In general, the correlations among methods for proximity data were weaker than
the correlations observed among methods for activity data, suggesting that the over-
all proximity patterns described by each method were less similar than the activity
Monthly rate
Monthly rate
0.12 0.30
a a
0.08 a 0.20
a a a
0.04 a 0.10
0 0
GRP SCN INST FCL CONT FCL GRP SCN INST FCL CONT FCL
Monthly rate
0.20 0.08 a
0.15 a a a 0.06 a a
a a
0.10 0.04 a
a
0.05 0.02
0 0
GRP SCN INST FCL CONT FCL GRP SCN INST FCL CONT FCL
Stems
0.20
a a
0.16
a a a
Monthly rate
0.12 a
a
a
0.08
0.04
0
GRP SCN INST FCL CONT FCL
Fig. 3. Monthly rates of feeding on mature and young leaves, mature and young fruits, and stems
by adult males (dark gray bars), adult females (white bars), and juveniles (light gray bars) for data
collected via group scan (GRP SCN), instantaneous focals (INST FCL), and continuous focals
(CONT FCL). Statistically significant differences among the age classes within each data collec-
tion method are indicated by different letters (a, b, or c). n = 13 for each bar. Means ± SE.
and diet patterns. However, correlations between group proximity scans and focal
proximity scans in both distance categories were higher than between focal nearest
neighbor samples and either group proximity scans or focal proximity scans (table 3).
In fact, only a few significant correlations were observed between focal nearest neigh-
bor samples and the other two methods. These correlations only occurred in the 0- to
1-meter category (table 3). Therefore, there was little agreement between focal nearest
neighbor samples and both other methods in terms of data patterns.
0–1 m
AM-AM 0.005 0.988 –0.205 0.503 0.825 0.002
AF-AF 0.622 0.041 0.692 0.009 0.945 0.000
AM-AF 0.330 0.322 0.621 0.023 0.644 0.032
AF-J 0.646 0.032 0.781 0.002 0.864 0.001
AM-J 0.510 0.109 0.729 0.005 0.787 0.004
>1 – 5 m
AM-AM –0.372 0.260 –0.341 0.729 0.683 0.021
AF-AF 0.034 0.920 0.341 0.255 0.756 0.007
AM-AF –0.716 0.013 –0.439 0.133 0.873 0.000
AF-J 0.153 0.654 –0.215 0.480 0.678 0.022
AM-J –0.126 0.710 –0.057 0.854 0.781 0.005
n = 13 for each test. Bold p values indicate significant correlations between two data collection methods
after sequential Bonferroni corrections with the initial α set at 0.050. AM = Adult male; AF = adult female;
J = juvenile.
Table 4. Pairwise differences of data collection methods and corresponding one-sample t tests and p values
0–1 m
AM-AM –0.014 ± 0.035 –1.31 0.218 –0.010 ± 0.050 –0.75 0.470 0.001 ± 0.028 –0.11 0.918
AF-AF –0.127 ± 0.097 –4.32 0.002 –0.107 ± 0.138 –2.80 0.016 0.000 ± 0.062 0.01 0.993
AM-AF –0.022 ± 0.030 –2.41 0.037 –0.020 ± 0.021 –3.44 0.005 –0.005 ± 0.021 –0.82 0.430
AM-J –0.023 ± 0.031 –2.45 0.034 –0.008 ± 0.020 –1.37 0.196 –0.016 ± 0.022 –2.45 0.034
AF-J –0.036 ± 0.052 –2.29 0.045 –0.051 ± 0.050 –3.66 0.003 0.019 ± 0.044 1.47 0.171
>1 – 5 m
AM-AM –0.146 ± 0.102 –4.74 0.001 –0.113 ± 0.090 –4.52 0.001 –0.022 ± 0.065 –1.10 0.296
AF-AF –0.147 ± 0.119 –4.08 0.002 –0.126 ± 0.100 –4.53 0.001 –0.006 ± 0.078 –0.27 0.792
AM-AF –0.157 ± 0.143 –3.65 0.004 –0.148 ± 0.133 –4.01 0.002 0.003 ± 0.045 0.21 0.837
AM-J –0.129 ± 0.105 –4.05 0.000 –0.128 ± 0.082 –5.64 0.001 0.009 ± 0.058 0.51 0.623
AF-J –0.144 ± 0.065 –7.36 0.002 –0.119 ± 0.094 –4.55 0.000 –0.012 ± 0.060 –0.68 0.509
n = 13 for each test. Bold p values indicate significant differences between two data collection methods (α = 0.050).
AM = Adult male; AF = adult female; J = juvenile.
bc
0.12
0.08 c b
ac c
a a ac
0.04 a a
0
a GRP SCN FCL SCN NRST NEIGHB
>1–5 m AM-AM
0.30 a a
a AF-AF
a a
0.25 AM-AF
Mean ± SE monthly rate
a a a AM-J
a a AF-J
0.20
a J-J
0.15 b b
0.10
a a
a
0.05 a
0
b GRP SCN FCL SCN NRST NEIGHB
Fig. 4. Monthly rates of time spent at 0–1 m (a) and at >1–5 m (b) for adult male (AM)-AM dy-
ads, adult female (AF)-AF dyads, AM-AF dyads, AM-juvenile (J) dyads, AF-J dyads, and J-J dy-
ads from data collected via group proximity scans (GRP SCN), focal proximity scan (FCL SCN),
and focal nearest neighbor sampling (NRST NEIGHB). Statistically significant differences among
the age classes within each data collection method are indicated by different letters (a, b, or c).
n = 13 for each bar. Means ± SE.
Discussion
As expected with a primate species that exhibits high group cohesiveness, group
scans, continuous focals, and instantaneous focals generally provided similar infor-
mation about howler monkey activity and diet. The majority of activity and diet data
were correlated across methods, suggesting that all methods documented similar pat-
terns in behavior. For example, if group scans indicated the howlers spent a larger
percentage of time feeding in month A compared to month B, so did continuous fo-
cals and instantaneous focals. The observed discrepancies emerged mostly for catego-
ries that occurred with less frequency during the study period (e.g. howling, moving,
or feeding on young fruits). In these cases, it seems likely that the mismatches were a
product of scatter in the data from small sample sizes and not necessarily a reflection
of the quality of the method itself [Edwards et al., 2010]. Indeed, when the actual per-
centage of time spent in these weakly correlated activities was compared using t tests,
few differences were detected among methods.
However, despite the similarities in patterns among datasets, important differ-
ences did exist in the percentage of time each method showed the howlers spending
in each activity. To begin with, t tests suggested that group scans underestimate time
spent feeding and overestimate time spent moving. A bias toward moving in scan
samples has been detected in other studies [Rose, 2000; Stevenson and Quinones,
2004], and although the cohesive nature of howler monkey social groups leads to a
low interindividual variation in activity that might counteract this effect, howler
groups are naturally less cohesive during travel bouts [pers. observation]. At these
times, howler monkeys moving within or between trees are more likely to stand out.
Therefore, when observers are challenged to locate and score the activity of all indi-
viduals in a social group in a short time period during group scans, there may be a
bias toward locating moving individuals. Similarly, if individuals are moving during
a feeding bout (i.e. foraging) an observer bias toward moving could result in a bias
against feeding. Although the protocol for group scans that was employed in this
study was designed to eliminate the bias toward moving by implementing a 5-second
minimum duration for an activity to be recorded [Fragaszy et al., 1992; Kortsjens et
al., 2005], our t test results suggest that this effort was not entirely successful. There-
fore, another measure must be taken to counteract the group scan bias toward mov-
ing if studies collecting group scan data are to be compared with focal data, especial-
ly in primates with larger interindividual distances.
Our comparisons also provided evidence that instantaneous focals overestimate
social behavior when compared to the other two data collection procedures. Instan-
taneous focals assume the recorded proportions of the behaviors performed by the
Group proximity scans Focal nearest neighbor sample Focal proximity scans
focal are related to the real-time durations of those behaviors when the interval be-
tween the records is short enough. For most howler monkey behaviors, our chosen
interval of 2 min appears to be short enough since the howlers engage in those behav-
iors for at least 2 min at a time. However, with the exception of juvenile play, social
behavior in howler monkeys rarely lasts more than a few seconds. Therefore, for a
social behavior recorded for a focal individual at any given time point the proportion
of time spent in that behavior as described by instantaneous focal data will overesti-
mate the amount of real time spent in that behavior. In a species that engages in social
behavior more frequently and/or in which social bout duration is longer, this error
may be less dramatic [Rose, 2000].
Because group scans record behavior at longer intervals, we would expect it to
overestimate social behavior as well. However, there is a lower probability of social
behavior occurring on a 15-min time point than on a 2-min time point, and subtle
social behaviors may be easily overlooked by observers during scan sampling. Fur-
thermore, group scans record the behaviors of all group members at each time point
instead of that of a single focal individual. For howler monkeys, it is uncommon for
all group members to engage in social behavior at once. As a result, the social behav-
ior of a individual at a single group scan time point is balanced by the activities of all
other individuals at that time point. Additionally, because group scans provide data
for all individuals for the duration of the day instead of for the activity of a single in-
dividual during one or two focal periods (over the course of a day), it may counteract
biases toward infrequent activities that occur during a single scan.
Finally, dramatic differences occurred among methods for describing spatial as-
sociations among group members. While proximity data from scan samples of the
entire group and relative to the focal individual corresponded well, focal nearest
neighbor samples consistently provided distinct patterns and values, suggesting that
focal nearest neighbor samples are ineffective for estimating the amount of time pairs
Acknowledgments
We thank the Mexican government environmental agency for research permission to work
in Palenque National Park, granted to Dr. Alejandro Estrada. We thank the Universidad Nacional
Autónoma de México for logistical and financial support. K.R.A. was funded by a National Geo-
graphic Waitt Grant and University of Illinois Dissertation Travel Grant, and is also supported
by an National Science Foundation Graduate Research Fellowship. S.V.B. was supported by a
postdoctoral fellowship from Universidad Nacional Autónoma de México. K.R.A. is grateful
to S. Wengert, and S.V.B. is grateful to E. Pain, B. Thomas, E. Thom, P. Boone, L. Cervera, T.
Roberts, S. Schindel, and M. Groenenberg who provided support in the field. We also thank 3
anonymous reviewers for their comments.
Altmann J (1974) Observational study of behavior: Sampling methods. Behaviour 49: 227–267.
Bernstein IS (1991). An empirical comparison of focal and ad libitum scoring with commentary on in-
stantaneous scans, all occurrence and one-zero techniques. Animal Behaviour 42: 721–728.
Di Fiore A, Fleischer RC (2005). Social behavior, reproductive strategies, and population genetic structure
of Lagothrix poeppigii. International Journal of Primatology 26: 1137–1173.
Dunbar RIM (1976). Some aspects of research design and their implications in the observational study of
behaviour. Behaviour 58: 78–98.
Edwards LN, Sorkin AE, Rhodes RC III, Peterson KH (2010). Observational methods to measure behav-
iors of captive cotton-top tamarins (Saguinus oedipus). Zoo Biology 29: 416–431.
Estrada AE, Castellanos L, Ibarra A, Garcia Del Valle Y, Muñoz D, Rivera A, Franco B, Fuentes E, Jiminez
C (2002). Survey of the population of the black howler monkey, Alouatta pigra, at the Mayan site of
Palenque, Chiapas, Mexico. Primates 44: 51–58.
Fragaszy D, Boinski S, Whipple J (1992). Behavioral sampling in the field: Comparison of individual and
group sampling methods. American Journal of Primatology 26: 259–275.
Gilby IC, Pokempner AA, Wrangham RW (2010). A direct comparison of scan and focal sampling meth-
ods for measuring wild chimpanzee feeding behavior. Folia Primatologia 81: 254–264.
Haddadi H, King AJ, Wills AP, Fay D, Loew J, Morton AJ, Hailes S, Wilson AM (2011). Determining as-
sociation networks in social animals: Choosing spatial-temporal criteria and sampling rates. Behav-
ioral Ecology and Sociobiology 65: 1659–1668.
Hernández-LLoreda MV, Colmenares F (2006). The utility of generalizability theory in the study of animal
behaviour. Animal Behaviour 71: 983–988.
Holm S (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics
6: 65–70.
Kawanaka K (1996). Observation time and sampling intervals for measuring behavior and interactions of
chimpanzees in the wild. Primates 37: 185–196.
Kortsjens AH, Nijseen E, Noe R (2005). Intergroup relationships in western black-and-white colobus,
Colobus polykomos polykomos. International Journal of Primatology 26: 1267–1289.
Lehner PN (1996). Handbook of Ethological Methods, 2nd edition. Cambridge, Cambridge University
Press.
Martin P, Bateson P (1993). Measuring Behavior: An Introductory Guide, 2nd edition. Cambridge, Cam-
bridge University Press.
Milton K (1981). Food choice and digestive strategies of two sympatric primate species. The American
Naturalist 117: 496–505.
Paterson JD (2001). Primate Behavior: An Exercise Workbook. Prospect Heights, Waveland.
Pavelka MS (2011). Mechanisms of cohesion in black howler monkeys. In Origins of Altruism and Coop-
eration (Susmann RW, Cloninger CR, eds.), pp 167–178. New York, Springer.
Rhine RJ, Ender PB (1983). Comparability of methods used in the sampling of primate behavior. American
Journal of Primatology 5: 1–15.
Rhine RJ, Flanigon M (1978). An empirical comparison of one-zero, focal-animal, and instantaneous
methods of sampling spontaneous primate social behavior. Primates 19: 333–361.
Rice WR (1989). Analyzing tables of statistical tests. Evolution 43: 223–225.
Rose LM (2000). Behavioral sampling in the field: continuous focal versus focal interval sampling. Behav-
iour 137: 153–180.
Stevenson PR, Quinones MJ (2004). Sampling methods used to quantify activity patterns and diet: A com-
parative example using woolly monkeys (Lagothrix lagothricha). Field Studies of Fauna and Flora La
Macarena, Colombia 14: 21–28.
Takuda M, Boublie JP, Izar P, Strier KB (2012). Social cliques in male northern muriquis Brachyteles hy-
poxanthus. Current Zoology 58: 342–352.
Van Belle S, Estrada AE, Strier KB (2008). Social relationships among male Alouatta pigra. International
Journal of Primatology 29: 1481–1498.
Zinner D, Hindahl J, Schwibbe M (1997). Effects of temporal sampling patterns of all-occurrence record-
ing in behavioural studies: Many short sampling periods are better than a few long ones. Ethology
103: 236–246.