274
CHARLES J. KREBS AND JUDITH H. MYERS
case has it been shown to be a valid assumption. Leslie et al. (1953)
showed that Microtus agrestis were not sampled randomly between the
marked and unmarked segments of the population. Some voles are
trap-prone and others are trap-shy. The same results were obtained for
M . californicus by Krebs (1966). Tanton (1965, 1969) showed a seasonal
change in probability of capture for Apodemus sylvaticw and Clethrionomys glareolus. Tanaka (1963, 1972) has shown that the probability
of capture is different for unmarked and for marked voles of Microtus
montebelli, Clethrionomys rufocanus, and C . smithi.
One way to provide randomness of capture might be to prebait
animals for several days or weeks before trapping begins. Tanaka
(1970) prebaited C . rufocanus for three days and showed that this
amount of prebaiting increased the probability of capture of unmarked
voles slightly. Andrzejewski et al. (1971) showed that C. glareolus
which were trap-shy were caught more readily in permanent trap sites
than in random trap sites which were not prebaited. Capture was not,
however, at random in either of these studies, and Krebs (1966) found
that continuous prebaiting at permanent trap sites was not sufficient
to provide random sampling in Microtus californicw. Even though
prebaiting does not equalize the probability of capture over all individuals, it may still improve census estimates. We have found that
M. townsendi cannot be live-trapped even at high densities without
prebaiting (Krebs, unpublished). The same problem is found in N.
pennsylvanicus during the summer (Krebs et al., 1969).
If the assumptions of standard capturerecapture analysis cannot be
met in rodent populations, two courses of action are available. First,
recent techniques for mark and recapture estimation with unequal
catchability can be utilized (Marten, 1970; Seber, 1970). The difficulty
is that again some uniformity assumptions must be made (e.g. that an
individual has a fixed probability of capture throughout its life).
No one to date has used these techniques on a long-term field study.
Second, one can attempt to enumerate the population by saturation
live-trapping at frequent intervals and hope that the errors involved are
relatively small. This approach was adopted by Chitty and Phipps
(1966) and has been used by Krebs (1964a, et seq.). If permanent
trapping stations are used, the enumeration approach seems to provide
the best technique for studying population processes in small rodents.
Unfortunately, many workers on small rodent populations do not
appreciate the problems of density estimation, and the literature is
filled with examples of population estimates derived from the Lincoln
Index with no attempt to satisfy the assumptions, examples of indices
of density such as snap-trap catches being interpreted quantitatively,
and sampling techniques applied with no appreciation of sampling
CHARLES J. KREBS AND JUDITH H. MYERS
case has it been shown to be a valid assumption. Leslie et al. (1953)
showed that Microtus agrestis were not sampled randomly between the
marked and unmarked segments of the population. Some voles are
trap-prone and others are trap-shy. The same results were obtained for
M . californicus by Krebs (1966). Tanton (1965, 1969) showed a seasonal
change in probability of capture for Apodemus sylvaticw and Clethrionomys glareolus. Tanaka (1963, 1972) has shown that the probability
of capture is different for unmarked and for marked voles of Microtus
montebelli, Clethrionomys rufocanus, and C . smithi.
One way to provide randomness of capture might be to prebait
animals for several days or weeks before trapping begins. Tanaka
(1970) prebaited C . rufocanus for three days and showed that this
amount of prebaiting increased the probability of capture of unmarked
voles slightly. Andrzejewski et al. (1971) showed that C. glareolus
which were trap-shy were caught more readily in permanent trap sites
than in random trap sites which were not prebaited. Capture was not,
however, at random in either of these studies, and Krebs (1966) found
that continuous prebaiting at permanent trap sites was not sufficient
to provide random sampling in Microtus californicw. Even though
prebaiting does not equalize the probability of capture over all individuals, it may still improve census estimates. We have found that
M. townsendi cannot be live-trapped even at high densities without
prebaiting (Krebs, unpublished). The same problem is found in N.
pennsylvanicus during the summer (Krebs et al., 1969).
If the assumptions of standard capturerecapture analysis cannot be
met in rodent populations, two courses of action are available. First,
recent techniques for mark and recapture estimation with unequal
catchability can be utilized (Marten, 1970; Seber, 1970). The difficulty
is that again some uniformity assumptions must be made (e.g. that an
individual has a fixed probability of capture throughout its life).
No one to date has used these techniques on a long-term field study.
Second, one can attempt to enumerate the population by saturation
live-trapping at frequent intervals and hope that the errors involved are
relatively small. This approach was adopted by Chitty and Phipps
(1966) and has been used by Krebs (1964a, et seq.). If permanent
trapping stations are used, the enumeration approach seems to provide
the best technique for studying population processes in small rodents.
Unfortunately, many workers on small rodent populations do not
appreciate the problems of density estimation, and the literature is
filled with examples of population estimates derived from the Lincoln
Index with no attempt to satisfy the assumptions, examples of indices
of density such as snap-trap catches being interpreted quantitatively,
and sampling techniques applied with no appreciation of sampling