Geoscience Reference
In-Depth Information
for longer than is necessary reduces the effective data set. Having observed a
focal chimp for 300 hours, Kawanaka (1996) discovered that subsets of 25
hours reliably and consistently represented the same time budget (see also
Arnoldmeeks and McGlone 1986). Excessive recording of particular focal
individuals also worsens the task of finding periods of relative stationarity.
Cycling through a series of focal individuals and then checking an individual
against itself in an earlier period may circumvent this. Video recording can
sometimes relax these limitations. There may be a general warning in the find-
ing of Arnoldmeeks and McGlone (1986) that focal animal sampling of one
individual among a litter of young pigs provided a good measure of the time
budgets of other piglets in the same period for all but social behavior.
TIME SAMPLING
One-Zero Sampling
One-zero sampling consistently overestimates duration and gives unreliable
information on frequency (Dunbar 1976; Altmann 1974). This can be partly
overcome by reducing the recording interval to periods shorter than the bout
length of the behavioral categories, but the technique then often becomes
more cumbersome than continuous focal sampling. Therefore, this once
common sampling method has largely been abandoned. Using a cumulative
Poisson process, it is possible to assess, for a given interval length, the proba-
bility that the necessary conditions will not be met for accurate frequency
counts and unbiased duration estimates in one-zero sampling (Suen and Ary
1984; 1986a). Suen and Ary (1986b) also identified a post hoc correction pro-
cedure that produces results with negligible systematic errors in one-zero
duration estimates. This correction procedure requires that more than five 0
scores lie between two consecutive but not adjacent 1 scores (see also Quera
1990).
Despite its general fall from favor, one-zero sampling has some potential
virtues. It gives a measure of the probability that an observer will see at least
one example of the response in the specified time interval. Bernstein (1991)
argues that this probability may be relevant to an animal's perception: For
example, when an individual is deciding whether to spend the next 5 min in
the vicinity of another individual, the probability of attack may be more
important than the mean duration or frequency of attacks. Martin and Bate-
son (1993) also note that the onset, duration, and frequency of some behavior,
such as play, are hard to represent with continuous sampling and hard to score
Search WWH ::




Custom Search