68
J . M. LAMBERT A N D M. B . D A L E
tion. Obviously the samples must be sufficiently large for each to
accommodate more than one species: otherwise there can be no interrelationships for analysis. The minimal size of sample site necessary to
give an unbiased sample of interrelations can only be determined if the
interrelations and distribution of the species are already known ; but
the bias will be limited if an arbitrary, rather than subjective, sample
size is fixed.
In the absence of complete coverage, the spatial relationships will be
limited by the intersample distance. Again, without foreknowledge of
the situation, an arbitrary spacing must be fixed. Taking the area as a
whole, it is usually preferable to have ‘equal information about spatial
relations, rather than varying amounts, so the intersample diKtance
should remain constant. This leads to the use of systematic sampling,
usually on some form of grid, rather than random sampling.
Such .considerations clearly preclude the subjective assessment of
species-group area and distribution implicit in any use of “stand”
samples as “representative”, and which will inevitably result in the
introduction of an unknown subjective bias. This should not, however,
be taken to imply that all sites which lie on the grid should necessarily
be recorded, provided some external criterion is used to reject unsuitable sites. For instance, in an investigation specifically concerned with
variations in forest vegetation, newly felled sites lying within the region
under study might well be excluded as irrelevant to the immediate
problem in hand; or, in a mountainous area, some sites might have to be
omitted because they were completely inaccessible.
D. THE NATURE O F T H E MEASURES
The final decision to be made concerns the nature of the information
about each species to be collected at each site. A wide variety of
measures have been proposed, ranging from a simple presence-orabsence system to the complex (‘indices of importance” used by Curtis
and McIntosh (1951). For instance, measures of density, dry weight,
leaf area, percentage cover, vitality and so on have all been adopted
at one time or another in various phytosociological studies. Given that
the prime requirement is to obtain unbiased information as efficiently as
possible, the different types of measure must now be looked at in this light.
In the first place, there is little to be said on theoretical grounds in
favour of composite, mixed quantitative measures. Examples of these
are Curtis’s “Importance Value” (a sum of non-additive numbers, i.e.
relative density, relative frequency and relative cover) and the various
“cover-abundance” scales much used in continental work. The latter,
in fact, are little more than mere subjective estimations, given
respectability by being displayed in a pseudo-quantitative form.
J . M. LAMBERT A N D M. B . D A L E
tion. Obviously the samples must be sufficiently large for each to
accommodate more than one species: otherwise there can be no interrelationships for analysis. The minimal size of sample site necessary to
give an unbiased sample of interrelations can only be determined if the
interrelations and distribution of the species are already known ; but
the bias will be limited if an arbitrary, rather than subjective, sample
size is fixed.
In the absence of complete coverage, the spatial relationships will be
limited by the intersample distance. Again, without foreknowledge of
the situation, an arbitrary spacing must be fixed. Taking the area as a
whole, it is usually preferable to have ‘equal information about spatial
relations, rather than varying amounts, so the intersample diKtance
should remain constant. This leads to the use of systematic sampling,
usually on some form of grid, rather than random sampling.
Such .considerations clearly preclude the subjective assessment of
species-group area and distribution implicit in any use of “stand”
samples as “representative”, and which will inevitably result in the
introduction of an unknown subjective bias. This should not, however,
be taken to imply that all sites which lie on the grid should necessarily
be recorded, provided some external criterion is used to reject unsuitable sites. For instance, in an investigation specifically concerned with
variations in forest vegetation, newly felled sites lying within the region
under study might well be excluded as irrelevant to the immediate
problem in hand; or, in a mountainous area, some sites might have to be
omitted because they were completely inaccessible.
D. THE NATURE O F T H E MEASURES
The final decision to be made concerns the nature of the information
about each species to be collected at each site. A wide variety of
measures have been proposed, ranging from a simple presence-orabsence system to the complex (‘indices of importance” used by Curtis
and McIntosh (1951). For instance, measures of density, dry weight,
leaf area, percentage cover, vitality and so on have all been adopted
at one time or another in various phytosociological studies. Given that
the prime requirement is to obtain unbiased information as efficiently as
possible, the different types of measure must now be looked at in this light.
In the first place, there is little to be said on theoretical grounds in
favour of composite, mixed quantitative measures. Examples of these
are Curtis’s “Importance Value” (a sum of non-additive numbers, i.e.
relative density, relative frequency and relative cover) and the various
“cover-abundance” scales much used in continental work. The latter,
in fact, are little more than mere subjective estimations, given
respectability by being displayed in a pseudo-quantitative form.