FEEDING TIMES 
175 
expected only about once in a hundred million times. The distribution however is clearly 
skew, and if the data are replotted on a logarithmic scale of time (Fig. 2) the histogram 
approximates to a normal distribution. 
2. Recoveries of T. deliensis on a logarithmic scale of days. The same data as Fig. 1, but time 
scale logarithmic so that each interval represents three times as many days as the 
preceding. 
The extreme value now falls into place as an unusual variant to be expected once 
in three or four hundred times. 
There is the possibility that such an extreme value may be caused by the chigger dying 
while attached and being finally dislodged by the scratching of the rat. While such a possibility 
is freely admitted (there is no record of whether the mite was alive when recovered) the value 
should, strictly, be taken into account in calculating the “ mean feeding time ” as defined 
above, since it would undoubtedly be counted in calculating the infestation rate. If the extreme 
value is ignored the statistics become: mean 3.0, Sd. 1.93, se. 0.30. 
The most useful statistics would seem to the arithmetic mean, which is needed 
for population calculations, and the mode and the range, to illustrate the feeding habits. Since 
each group is closely centered the mode is taken as the central value of the modal group. In 
none of the distributions obtained is it possible to estimate any lower limit of feeding time and 
therefore only the upper limit is estimated. The upper limit is liable to large random 
variations, as in the example above, and it seems better to take the value within which 95 per 
cent, of the observations lie, corresponding to a value of the mean plus twice the standard 
deviation in a normal distribution. In the example given above the observed 95 per cent, 
range (within which 41 of the observations lie) is up to eight days. Taking the adjusted value 
of the mean (3.0) and standard deviation (1.93) the expected value would be 6.86 days. 
Two-day exposures 
Rats recovered within 48 hours of release have been exposed to a similar risk of infestation 
for two successive nights. There is no reason to suppose that actual infestation on one night 
is likely to affect the probability of infestation on the following night (although the probabilities 
for the two nights will be correlated), so that on the average the two probabilities should be the 
same. That this is so is shown in T able 2, from which it is clear that the probability of remaining 
uninfested for two nights is very close to the square of the probability of remaining uninfested 
for one night. The smaller number of recaptures after two nights is due to the method of 
trapping at one site for about five nights at a time. 
MALAYA , No. 26. 1953 
