Geoscience Reference
In-Depth Information
to the blurring of boundaries between studies of ecology, ethology, and socio-
biology. Today, the basic questions posed by those concerned with the contex-
tual interpretation of animal behavior still echo Niko Tinbergen's (1963) four
factors: causation, development, survival value, and evolution. The classic
sequence starts with observation to document the behavioral repertoire in con-
text, and this forms a basis for experimental study of causation, development,
and survival value and an attempt to reconstruct the evolution of particular
units of behavior (McFarland 1981). The approach rests on parsimony and
reductionism. Of course, there are different forms of reductionism, and the
biologist's variant, as described by Konrad Lorenz in
The Foundations of Ethol-
ogy
(1981), may deviate from the physicist's method of general reduction.
Here, we consider some limitations of the rationales behind four concepts
widely used in the study of social dynamics: statistical rationality, matrix facil-
ities, lag sequential and nested analysis, and concept of uncertainty measures
(Markov chain analysis).
STATISTICAL RATIONALITY
A society is the product of social flux among its members and between them
and outsiders. Therefore, for many purposes the description of a society
requires quantification of the interactive components of behavior between
members of, for example, a reproductive unit, group, colony, or population;
these may be continuously or instantaneously recorded. In the analysis of
observational data, insufficient attention to methodology can result in a rift
between statistical significance and biological relevance, whereas the corre-
spondence between these two can be fostered by attention to sampling tactics
and statistical sensitivity. An inadequately sensitive statistical method or model
may fail to reveal the properties of an interactive behavior, even if an appro-
priate sampling method was used.
The familiarity of two types of error has rendered neither rare. Type I error
involves mistaken rejection of a null hypothesis (i.e., a false positive); type II
error is failure to reject a false null hypothesis (i.e., ignorance) (Lindgren
1968). As a general statement, scientific validity depends partly on the sensi-
tivity of the tests involved (decreasing the probability of type I errors) and on
the number of hypotheses tested (the accumulating building blocks of the
argument adding to its objectivity and decreasing the probability of type II
errors). However, there is a tradeoff between these kinds of error; decreasing
type I error can increase the risk of type II error (Shrader-Frachette and Mc-
Coy 1992). There is also a risk that even when the hypothesis has been tightly
