Geoscience Reference
In-Depth Information
during true instantaneous sampling. The judgment that play has at least
occurred in a previously defined interval is more likely to be accurate. Finally,
some behavior, such as copulation, may occur often in one short interval and
then not be seen for many subsequent intervals. One-zero scan sampling
reveals how long a researcher is likely to have to watch the animals to see an
instance of such behavior.
Instantaneous Scan Sampling
This technique scores what some or all individuals are doing at predetermined
time intervals. It produces a reliable estimate of the total durations of activities
through a dimensionless measure of the proportion of occasions during which
the behavior is encountered (Hansson et al. 1993). Instantaneous scans may
give poor information on frequency, although this shortcoming is minimized
if the sample interval is small enough to catch even the shortest intervening
period between recurrence (Odagiri and Matsuzawa 1994). However, as in the
case of short one-zero measures, such short bouts become more inconvenient
than focal sampling. Furthermore, the need for statistical independence of the
data points requires that sample intervals not be too short. For example, if 10
successive scans all fell within a single long bout of a particular behavior, then
they constitute a sample of 1, not 10. Such dependence would require that the
data be transformed to a proportion on an interval scale for comparison rather
than being used as discrete categorical scores (and therefore could not be ana-
lyzed in an interaction sociomatrix). As a rule, Bernstein (1991) suggests that
if the probability of scoring the same continuous state twice in succession is
less than 0.05, then the two scans can be treated as independent. In practice
this means that the interval between scans should exceed the mean duration of
a state plus 1.96 standard deviations. Even long intervals between scans can
provide a basis for accurate estimates of mean time budgets, but the increase in
variance caused by longer sampling intervals may swamp analyses of covaria-
tion and produce type II errors (Poysa 1991).
The pros and cons of each approach, and the diverse aims of investigations,
mean not only that no particular method is universally superior, but that even
for one study an eclectic approach to sampling is likely to optimize data acqui-
sition. A mix of all occurrence sampling, scan sampling, and focal individual
instantaneous sampling is possible. The results of all three are likely to be
enhanced if decisions on the durations of intervals and focal sessions are
judged from preliminary analysis of a continuous focal sample (e.g., made by
using a portable video camera in the field).
