170
J. HESLOP-HARRISON
Harberd (1957, 1958), namely the danger of attributing false significance to differences between populations in genecological trials in
consequence of unsatisfactory sampling procedures. Harberd has been
concerned with the particular case of a single genotype being incorporated several times in a sample because of the clonal spread of a
parent plant to an unknown degree. By decreasing the within- population variance in a trial this can lead to the attribution of “spurious
significance” t o differences between populations. Both Harberd (1 957)
and Wilkins (1959) are driven to the conclusion that the within- population variance is not a particularly useful statistic in genecological
studies, and that the value of a trial is most likely to be increased by
increasing the number of populations sampled rather than the sample
size. For a trial of given scale, the ultimate is a sample of one plant from
each population; then, as Wilkins (1959) says, the variance would
automatically. be that within a purely statistical assemblage of unrelated individuals, and there would be no reason to expect any two
such assemblages from the same geographical area to show non-adaptive
differences.
A further source of confusion in genecological studies lies in the form
of material collected for experimental garden or other investigation.
The investigator in general has two choices: he can transfer mature
living plants, or he can grow his material from seed. In the former case,
he is sampling from a selected population; in the latter, he is sampling
from a population which has descended from selected ancestors but has
not itself suffered selection. The implications of this are sometimes overlooked. When populations occupy unlike habitats but are in sufficiently
close proximity to permit gene exchange, differentiation will only proceed if selection pressure is high enough to out-balance gene flow (p.
204). I n the extreme case, adaptation will be a generation-to-generation
matter, with an essentially random dispersal of genotypes over the
entire area each year and a subsequent stringent selection for adapted
genotypes in each habitat “sub-population”. In this case, to rely upon
seed samples for the estimation of genotypic differences is to guarantee
that they will not be found. Even when remote populations are being
compared, a seed sampling method may not provide an adequate picture
of what the actual surviving population in a given site is like. In the
absence of disruptive gene-flow from other populations, recombination
is unlikely to turn up radical deviants in any quantity, but where an
adaptive response depends upon a nice balance of polygenes in an outbreeding population amodal types are certain to occur in each generation, normally destined to succumb in the wild unless they happen to be
included in a seed packet. The aberrant individuals recorded in various
studies on photoperiodic and other responses of ecotypes may have had
J. HESLOP-HARRISON
Harberd (1957, 1958), namely the danger of attributing false significance to differences between populations in genecological trials in
consequence of unsatisfactory sampling procedures. Harberd has been
concerned with the particular case of a single genotype being incorporated several times in a sample because of the clonal spread of a
parent plant to an unknown degree. By decreasing the within- population variance in a trial this can lead to the attribution of “spurious
significance” t o differences between populations. Both Harberd (1 957)
and Wilkins (1959) are driven to the conclusion that the within- population variance is not a particularly useful statistic in genecological
studies, and that the value of a trial is most likely to be increased by
increasing the number of populations sampled rather than the sample
size. For a trial of given scale, the ultimate is a sample of one plant from
each population; then, as Wilkins (1959) says, the variance would
automatically. be that within a purely statistical assemblage of unrelated individuals, and there would be no reason to expect any two
such assemblages from the same geographical area to show non-adaptive
differences.
A further source of confusion in genecological studies lies in the form
of material collected for experimental garden or other investigation.
The investigator in general has two choices: he can transfer mature
living plants, or he can grow his material from seed. In the former case,
he is sampling from a selected population; in the latter, he is sampling
from a population which has descended from selected ancestors but has
not itself suffered selection. The implications of this are sometimes overlooked. When populations occupy unlike habitats but are in sufficiently
close proximity to permit gene exchange, differentiation will only proceed if selection pressure is high enough to out-balance gene flow (p.
204). I n the extreme case, adaptation will be a generation-to-generation
matter, with an essentially random dispersal of genotypes over the
entire area each year and a subsequent stringent selection for adapted
genotypes in each habitat “sub-population”. In this case, to rely upon
seed samples for the estimation of genotypic differences is to guarantee
that they will not be found. Even when remote populations are being
compared, a seed sampling method may not provide an adequate picture
of what the actual surviving population in a given site is like. In the
absence of disruptive gene-flow from other populations, recombination
is unlikely to turn up radical deviants in any quantity, but where an
adaptive response depends upon a nice balance of polygenes in an outbreeding population amodal types are certain to occur in each generation, normally destined to succumb in the wild unless they happen to be
included in a seed packet. The aberrant individuals recorded in various
studies on photoperiodic and other responses of ecotypes may have had