THE USE O F STATISTICS IN PHYTOSOCIOLOGY
67
phenomena, there are, however, four basic requirements to be borne
in mind. First, there should be sufficient variables for it to be possible
for significant differences, if any, to appear; secondly, for practical
convenience, the variables selected should be easily categorized, and the
categories easily identified in the field; thirdly, the definition of the
categories should be unambiguous and as independent as possible of
the system under study; and lastly, to be useful for interpretation, the
categories must be referable to other phenomena outside the immediate
system.
For ((general-puqose)) survey, the most useful criteria for categorizing plant material are those which have fairly wide relevance outside
and combine external information from a number of different sources.
The erection of such categories must itself involve prior abstraction
from other independent variables, and the level of abstraction to be
used is again a, subjective decision. Thus the widespread use of species
in traditional phytosociological work stems from a general acceptance
of the species concept as an economical method of characterizing plant
material by a number of properties simultaneously at a generally
convenient level of abstraction. The use of species composition for the
plant component in fact largely fulfils the four requirements outlined
above; and for convenience -without prejudice as to the possible
use of other plant units in other circumstances -we shall similarly
use species for categorizing our plants in the rest of this contribution.
c. THE NATURE O F THE SITES
We must now consider the criteria to be used in determining the
individual sites for record. In any given region, there is an infinite
number of possible sites, and decisions must be made as to their size,
shape and distribution. In a large area, complete coverage of the surface
is of course impracticable, and some method of sampling must be
adopted. Since the ultimate aim of the investigation is to reveal a
structure in the plant cover, and not to impose it, the sampling must be
as unbiased as possible.
The scale of any pattern which emerges w i l l obviously depend on
size and spacing of the sample sites, while the precision of the pattern
will depend on the number of samples. Although a great deal of work
on the (‘correct)’ size and shape of sample adequate to represent an
area of vegetation has been published, most of this is irrelevant in the
present context, where considerations of precision of estimate are not
involved. There are, however, other considerations to be borne in mind
arising from the nature of phytosociological investigations, which are
essentially concerned with the grouping of species, with the interrelations of such groups, and in some sense with their spatial distribu-
67
phenomena, there are, however, four basic requirements to be borne
in mind. First, there should be sufficient variables for it to be possible
for significant differences, if any, to appear; secondly, for practical
convenience, the variables selected should be easily categorized, and the
categories easily identified in the field; thirdly, the definition of the
categories should be unambiguous and as independent as possible of
the system under study; and lastly, to be useful for interpretation, the
categories must be referable to other phenomena outside the immediate
system.
For ((general-puqose)) survey, the most useful criteria for categorizing plant material are those which have fairly wide relevance outside
and combine external information from a number of different sources.
The erection of such categories must itself involve prior abstraction
from other independent variables, and the level of abstraction to be
used is again a, subjective decision. Thus the widespread use of species
in traditional phytosociological work stems from a general acceptance
of the species concept as an economical method of characterizing plant
material by a number of properties simultaneously at a generally
convenient level of abstraction. The use of species composition for the
plant component in fact largely fulfils the four requirements outlined
above; and for convenience -without prejudice as to the possible
use of other plant units in other circumstances -we shall similarly
use species for categorizing our plants in the rest of this contribution.
c. THE NATURE O F THE SITES
We must now consider the criteria to be used in determining the
individual sites for record. In any given region, there is an infinite
number of possible sites, and decisions must be made as to their size,
shape and distribution. In a large area, complete coverage of the surface
is of course impracticable, and some method of sampling must be
adopted. Since the ultimate aim of the investigation is to reveal a
structure in the plant cover, and not to impose it, the sampling must be
as unbiased as possible.
The scale of any pattern which emerges w i l l obviously depend on
size and spacing of the sample sites, while the precision of the pattern
will depend on the number of samples. Although a great deal of work
on the (‘correct)’ size and shape of sample adequate to represent an
area of vegetation has been published, most of this is irrelevant in the
present context, where considerations of precision of estimate are not
involved. There are, however, other considerations to be borne in mind
arising from the nature of phytosociological investigations, which are
essentially concerned with the grouping of species, with the interrelations of such groups, and in some sense with their spatial distribu-