METHODS OF SAMPLING THE BENTHOS
243
densities in terms of sediment volume rather than surface area. Some
calculations have been made to show how sample volume affects the
catch both in terms of numbers of species and of individual. These
relate to the volumc,of different samples taken with the same gear.
Thus, Lee (1944) and Ursin (1960) found that the number of species
taken was more or less independent of sample volume. For number of
specimens, however, there is some degree of positive correlation with
sample volume, as shown by Lee (1944), Ursin (1960), and Ellis (1960).
As sample volume increases the number of specimens would be expected
to rise steeply at first, reaching an asymptote for the larger samples
which were sufficiently deep for all the animals to be taken. Ellis
also determined the total weight of animals with increasing sample
volume. This showed a fairly linear rise with depth, because the
deeper animals, although few in numbers, were large.
Such empirical calculations are necessary because there is uncertainty both as to the action of the grab and as to the depths at which
the animals live. If there is little increase in numbers of species or
individuals in the larger samples there can be some confidence that
the grab is digging deeply enough. If there is a great difference between
large and small samples it may be aa well to reject those below a
certain volume. Wherever possible the results should be finally
expressed in terms of surface area, not volume.
Apart from variation due to depth of sampling, there is an " edgeeffect '' which may prove significant, especially in samples of small
size. This is due to the inclusion or exclusion, both by the mechanics
of the sampling instrument and later by the observer's decision, of
animals living on the boundary of the sampled area. So far as animals
living mainly below the depth sampled are concerned, the capture of
their uppermost parts, such as the siphons of lamellibranchs, and their
inclusion in the list for the sample merely increases the effective
sampling depth ; but with animals occurring a t the side of the sampled
area, a statistical problem is involved. Frequently an animal is only
represented by a fragment, because it was at the edge of the sample,
and was broken off by the grab. If it can be identified, this fragment,
however small, will be listed aa one individual, and this represents a
bias which may be apparent when a number of samples are being
andysed statistically (cp. Ursin, 1900, p. 149). Where larger numbere
of fragments are present it is customary to count the numbers of one
part, for example the head, so that any odd fragments unmatched
with a head will not be counted. This is a statisti~lly-acceptable
method for any size of sample, and it should no doubt be strictly applied
to all samples when statistical treatment is contemplated, even to the
243
densities in terms of sediment volume rather than surface area. Some
calculations have been made to show how sample volume affects the
catch both in terms of numbers of species and of individual. These
relate to the volumc,of different samples taken with the same gear.
Thus, Lee (1944) and Ursin (1960) found that the number of species
taken was more or less independent of sample volume. For number of
specimens, however, there is some degree of positive correlation with
sample volume, as shown by Lee (1944), Ursin (1960), and Ellis (1960).
As sample volume increases the number of specimens would be expected
to rise steeply at first, reaching an asymptote for the larger samples
which were sufficiently deep for all the animals to be taken. Ellis
also determined the total weight of animals with increasing sample
volume. This showed a fairly linear rise with depth, because the
deeper animals, although few in numbers, were large.
Such empirical calculations are necessary because there is uncertainty both as to the action of the grab and as to the depths at which
the animals live. If there is little increase in numbers of species or
individuals in the larger samples there can be some confidence that
the grab is digging deeply enough. If there is a great difference between
large and small samples it may be aa well to reject those below a
certain volume. Wherever possible the results should be finally
expressed in terms of surface area, not volume.
Apart from variation due to depth of sampling, there is an " edgeeffect '' which may prove significant, especially in samples of small
size. This is due to the inclusion or exclusion, both by the mechanics
of the sampling instrument and later by the observer's decision, of
animals living on the boundary of the sampled area. So far as animals
living mainly below the depth sampled are concerned, the capture of
their uppermost parts, such as the siphons of lamellibranchs, and their
inclusion in the list for the sample merely increases the effective
sampling depth ; but with animals occurring a t the side of the sampled
area, a statistical problem is involved. Frequently an animal is only
represented by a fragment, because it was at the edge of the sample,
and was broken off by the grab. If it can be identified, this fragment,
however small, will be listed aa one individual, and this represents a
bias which may be apparent when a number of samples are being
andysed statistically (cp. Ursin, 1900, p. 149). Where larger numbere
of fragments are present it is customary to count the numbers of one
part, for example the head, so that any odd fragments unmatched
with a head will not be counted. This is a statisti~lly-acceptable
method for any size of sample, and it should no doubt be strictly applied
to all samples when statistical treatment is contemplated, even to the