THE USE O F STATISTICS I N PHYTOSOCIOLOGY
97
the efficiency of the programme itself, and here the ecologist is entirelydependent on advances in machine design and in programming technique. What is frequently forgotten in estimating’ the time and cost of
employing computer-analysis is that these must be balanced against
the time and cost in man-hours of the actual collection of the data:
to contemplate an analysis extending over several days-or
even
weeks-would not be exorbitant if the data themselves had taken
several years to collect.
The question of the extent to which computer analyses can eventually
replace a trained ecologist in the survey of vegetation is naturally
controversial. The well-established phytosociological schools apparently
place great emphasis on the need for experience not only in the collection
of the data but also in the subsequent analytical work. However, here
we sympathize fully with Webb (1954, p. 365) when he says : “It seems
curious that plant-sociologists who collect such large masses of data in
quantitative form should then rely merely on their faculty of recognition for sorting out the communities, instead of using a quantitative
yard-stick”; and later: “For a scientific taxonomy it is not enough to
assure one’s critics that with sufficient experience one can learn to
recognize the units; there must be some means of defining them.
Without this plant-sociology can only be a craft to which one is apprenticed, and not a science which one can learn. . . . Not only is such a
procedure uncommunicable ; it is also unreliable.”
Although we ourselves are fully convinced of the value and power
of appropriate statistical techniques in vegetational analysis, this does
not of course imply that we consider that the professional ecologist will
himself ultimately become outmoded. Statistics is only a servant, and
must not become the master. Its function is to enable a worker to
examine a new situation free from preconceived ideas, to aid him in
the mechanical task of sorting the data he acquires, and to guide him
to the points of greatest significance in a complex situation; but the
particular problem to be tackled, the nature of the data to be collected,
and the method of assessing the results, must always be determined
by the investigator himself.
ACKNOWLEDGMENTS
We are extremely grateful to Professor W. T. Williams for muchuseful
discussion, particularly on the more statistical points, during the preparation of this paper; and we should also like to thank Mr P. GreigSmith for making the typescript of the 2nd edition of his book Quantitative Methods in Plant Ecology available to us before it was published.
97
the efficiency of the programme itself, and here the ecologist is entirelydependent on advances in machine design and in programming technique. What is frequently forgotten in estimating’ the time and cost of
employing computer-analysis is that these must be balanced against
the time and cost in man-hours of the actual collection of the data:
to contemplate an analysis extending over several days-or
even
weeks-would not be exorbitant if the data themselves had taken
several years to collect.
The question of the extent to which computer analyses can eventually
replace a trained ecologist in the survey of vegetation is naturally
controversial. The well-established phytosociological schools apparently
place great emphasis on the need for experience not only in the collection
of the data but also in the subsequent analytical work. However, here
we sympathize fully with Webb (1954, p. 365) when he says : “It seems
curious that plant-sociologists who collect such large masses of data in
quantitative form should then rely merely on their faculty of recognition for sorting out the communities, instead of using a quantitative
yard-stick”; and later: “For a scientific taxonomy it is not enough to
assure one’s critics that with sufficient experience one can learn to
recognize the units; there must be some means of defining them.
Without this plant-sociology can only be a craft to which one is apprenticed, and not a science which one can learn. . . . Not only is such a
procedure uncommunicable ; it is also unreliable.”
Although we ourselves are fully convinced of the value and power
of appropriate statistical techniques in vegetational analysis, this does
not of course imply that we consider that the professional ecologist will
himself ultimately become outmoded. Statistics is only a servant, and
must not become the master. Its function is to enable a worker to
examine a new situation free from preconceived ideas, to aid him in
the mechanical task of sorting the data he acquires, and to guide him
to the points of greatest significance in a complex situation; but the
particular problem to be tackled, the nature of the data to be collected,
and the method of assessing the results, must always be determined
by the investigator himself.
ACKNOWLEDGMENTS
We are extremely grateful to Professor W. T. Williams for muchuseful
discussion, particularly on the more statistical points, during the preparation of this paper; and we should also like to thank Mr P. GreigSmith for making the typescript of the 2nd edition of his book Quantitative Methods in Plant Ecology available to us before it was published.