86
J. M. LAMBERT AND M. B. DALE
metho& will have been mainly concerned with differences: mapping
of agglomerated groups will therefore tend to reveal foci of habitat
resemblance, while mapping of subdivided groups will pinpoint attention
on boundaries between different environmental conditions. Nevertheless,
although such mapping may serve to define areas of overall similarity
or difference, the operative habitat factors have still to be identified; and
this requires separate examination of such other properties of the sites
as could be relevant. So far, all correlations on this basis have been
subjective only, but there’is in principle no reason why more objective
assessments should not be attempted.
An example of the opposite approach, i.e. the examination of site/
habitat relationships as the primary system, is seen in Whittaker’s
“gradient analysis” (Whittaker, 1956). Although his sites were ordinated instead of classified, the comparison of sites primarily by means of
habitat factors instead of species composition still remains the central
principle; “important” habitat factors were selected to provide axes
on which to ordinate the sites, and the latter were then inspected
subjectively for species relationships. Other examples can be found in
Greig-Smith (1964, pp. 191-7); in fact, this type of approach, through
habitat factors first, now seems to be becoming increasingly common
in some American and Continental work (cf. Gounot, 1961), but so far
much of it is essentially empirical in nature.
An interesting preliminary study in the independent assessment of
species/site and sitelhabitat relationships was made by Hughes and
Lindley (1955), who analysed separately the species composition of two
subjectively selected types of plant community and the chemical
properties of six subjectively selected soil series bjr’means of a distance
function ; but, although-they suggest the possibility of collation between
environmental factors and related vegetational characteristics as a
practicable exercise, no collation was actually attempted. The most
appropriate method of direct collation would appear to be by some
form of canonical analysis (vide e.g. Kendall, 1957); such methods deal
with the interrelationships of interdependent sets of variables but,
with the exception of a few preliminary studies in the social sciences,
canonical techniques have so far been little used in practice.
Unfortunately, there is still a general lack of empirical knowledge as
to the relative appropriateness and efficacy of different methods, a
lack which is due partly to the scale of computation and partly to the
extreme frailty of some of the available techniques. However, on
statistical as well as general grounds, at present it seems that the most
efficient means of objective collation are likely to reside in further
attempts to correlate habitat information directly with the resdts of
an initial species/site analysis. The question then arises as to whether
c
J. M. LAMBERT AND M. B. DALE
metho& will have been mainly concerned with differences: mapping
of agglomerated groups will therefore tend to reveal foci of habitat
resemblance, while mapping of subdivided groups will pinpoint attention
on boundaries between different environmental conditions. Nevertheless,
although such mapping may serve to define areas of overall similarity
or difference, the operative habitat factors have still to be identified; and
this requires separate examination of such other properties of the sites
as could be relevant. So far, all correlations on this basis have been
subjective only, but there’is in principle no reason why more objective
assessments should not be attempted.
An example of the opposite approach, i.e. the examination of site/
habitat relationships as the primary system, is seen in Whittaker’s
“gradient analysis” (Whittaker, 1956). Although his sites were ordinated instead of classified, the comparison of sites primarily by means of
habitat factors instead of species composition still remains the central
principle; “important” habitat factors were selected to provide axes
on which to ordinate the sites, and the latter were then inspected
subjectively for species relationships. Other examples can be found in
Greig-Smith (1964, pp. 191-7); in fact, this type of approach, through
habitat factors first, now seems to be becoming increasingly common
in some American and Continental work (cf. Gounot, 1961), but so far
much of it is essentially empirical in nature.
An interesting preliminary study in the independent assessment of
species/site and sitelhabitat relationships was made by Hughes and
Lindley (1955), who analysed separately the species composition of two
subjectively selected types of plant community and the chemical
properties of six subjectively selected soil series bjr’means of a distance
function ; but, although-they suggest the possibility of collation between
environmental factors and related vegetational characteristics as a
practicable exercise, no collation was actually attempted. The most
appropriate method of direct collation would appear to be by some
form of canonical analysis (vide e.g. Kendall, 1957); such methods deal
with the interrelationships of interdependent sets of variables but,
with the exception of a few preliminary studies in the social sciences,
canonical techniques have so far been little used in practice.
Unfortunately, there is still a general lack of empirical knowledge as
to the relative appropriateness and efficacy of different methods, a
lack which is due partly to the scale of computation and partly to the
extreme frailty of some of the available techniques. However, on
statistical as well as general grounds, at present it seems that the most
efficient means of objective collation are likely to reside in further
attempts to correlate habitat information directly with the resdts of
an initial species/site analysis. The question then arises as to whether
c