Geoscience Reference
In-Depth Information
exponential distribution; such a sequence is said to have Markovian properties.
Such a process is completely characterized by a set of transition rates, which are
the probabilities per unit of time of switching from one state to another. By
definition, acts within such a sequence are independent of each other and of
time. The termination rate of an act is constant (i.e., independent of the past)
only if the duration of the bouts follows an exponential distribution. This fact
allows the inference of social interaction because dependencies between the
terminations of different individuals would imply that they exchange signals
(Haccou and Meelis 1992). Indeed, most sequences are not Markovian
because the probability that most behaviors will occur depends on earlier
events in the sequence and other confounding variables. Generally, for a given
behavior, the probability of occurrence of any future behavior in the sequence
is altered by additional knowledge concerning the past (Haccou and Meelis
1992). The program
UNCERT
(Hailman and Hailman 1993) analyzes serial
dependencies in sequential events using the method of Markov chains and
expressed in terms of the uncertainty that one event will follow another. For
example, Hailman and Dzelzkalnz (1974) used Markov chain analyses to
reveal that tail-wagging by mallard duck (
Anas platyrhynchos
) punctuated a
long sequence of behavior patterns, indicating that a display is about to start
(loosely analogous to the role of a capital letter beginning a sentence).
PREDICTABILITY OF BEHAVIOR
Measures of uncertainty can be a valuable tool in assessing the predictability of
responses of individuals to others. Again, we explore an example from our
work on wood mice (Stopka and Macdonald, unpublished data). For example,
all combinations of five male and four female wood mice kept under video
surveillance in an enclosure with abundant nest boxes sometimes denned com-
munally, although there was no thermoregulatory need to huddle. It was there-
fore surprising that cohabitants were often agonistic to each other. One of the
most common categories of interaction was approach and nose-to-nose con-
tact, categorized by Bovet (1972a, 1972b) as amicable, but lag sequential
analysis revealed that although this (nose-to-nose) behavior pattern was indeed
sometimes a precursor to flank-to-flank or nasoanal contact, on other occa-
sions it was followed by attacking or chasing. Therefore, nose-to-nose encoun-
ters are defined by their context within a sequence. Sequential analysis revealed
that nose-to-nose contact was sometimes the prelude to an amicable encounter
and sometimes to an aggressive one (figure 10.10).
The crucial question is how to recognize mixed categories. Haccou and
