THE USE O F STATISTICS I N PHYTOSOCIOLOBY
81
situation. In large-scale survey, it is the overall pattern of the vegetation, rather than the exact relationships of every individual site, which
provides the major interest; in such a situation, a polythetic method
might well be rejected in favour of a cruder, but quicker, monothetic
technique, since “misclassification” of a few isolated sites is unlikely
seriously to disturb the general ecological picture.
Once the groups of sites have been eventually extracted, they may
either be used directly for the generation of hypotheses as to the
ecological nature of their interrelationships, or they could be manipulated further. For instance, the groups at a given hierarchical level could
be explored for reticulate relationships, although, if the classification
has been efficient, such exploration may only confuse instead of clarify
the issue. Similarly, the groups in theory could be ordinated, but, again,
an efficient primary classification is likely to render subsequent ordination unprofitable. However, although the use of the more powerful
classificatory techniques may thus make subsequent structuring on
other systems inadvisable, the potential value of the primary results
for direct hypothesization is itself increased : by maximizing the fundamentally important similarities or differences between the sites, the
central relationships are exposed more clearly by greater elimination
of peripheral information. I n short, assuming that the most important
function of such statistical analysis is to facilitate hypothesization
rather than merely to describe, the use of the most powerful classificatory method available seems justified; once the data have been efficiently
structured in this way, the primary statistical operation may be
terminated as far as the site relationships alone are concerned.
D. INVERSE ANALYSIS
So far, we have dealt exclusively with ultimate sitelsite relationship,
defined in terms of their floristic composition; such classification of
sites, in fact, is comparable in effect to the classification of “stands” in
traditional phytosociology , since the so-called stands are basically
individual sites with their whole complement of species. However, the
interest of the data should not be regarded as lying entirely in the site
relationships, since the extraction of separate species-groups may also
provide information of value to the phytosociologist : a complementary
classification of the species, in terms of the sites in which they occur,/
may in fact reveal similarities and differences in their ecological
behaviour which might otherwise be obscured. This we may call
“inverse” analysis, to contrast it with the “normal” process of siteanalysis.
It is obvioupthat the type of information obtained by “inverse”
analysis will be different from that provided by “normal” methods. I n
81
situation. In large-scale survey, it is the overall pattern of the vegetation, rather than the exact relationships of every individual site, which
provides the major interest; in such a situation, a polythetic method
might well be rejected in favour of a cruder, but quicker, monothetic
technique, since “misclassification” of a few isolated sites is unlikely
seriously to disturb the general ecological picture.
Once the groups of sites have been eventually extracted, they may
either be used directly for the generation of hypotheses as to the
ecological nature of their interrelationships, or they could be manipulated further. For instance, the groups at a given hierarchical level could
be explored for reticulate relationships, although, if the classification
has been efficient, such exploration may only confuse instead of clarify
the issue. Similarly, the groups in theory could be ordinated, but, again,
an efficient primary classification is likely to render subsequent ordination unprofitable. However, although the use of the more powerful
classificatory techniques may thus make subsequent structuring on
other systems inadvisable, the potential value of the primary results
for direct hypothesization is itself increased : by maximizing the fundamentally important similarities or differences between the sites, the
central relationships are exposed more clearly by greater elimination
of peripheral information. I n short, assuming that the most important
function of such statistical analysis is to facilitate hypothesization
rather than merely to describe, the use of the most powerful classificatory method available seems justified; once the data have been efficiently
structured in this way, the primary statistical operation may be
terminated as far as the site relationships alone are concerned.
D. INVERSE ANALYSIS
So far, we have dealt exclusively with ultimate sitelsite relationship,
defined in terms of their floristic composition; such classification of
sites, in fact, is comparable in effect to the classification of “stands” in
traditional phytosociology , since the so-called stands are basically
individual sites with their whole complement of species. However, the
interest of the data should not be regarded as lying entirely in the site
relationships, since the extraction of separate species-groups may also
provide information of value to the phytosociologist : a complementary
classification of the species, in terms of the sites in which they occur,/
may in fact reveal similarities and differences in their ecological
behaviour which might otherwise be obscured. This we may call
“inverse” analysis, to contrast it with the “normal” process of siteanalysis.
It is obvioupthat the type of information obtained by “inverse”
analysis will be different from that provided by “normal” methods. I n