34 
BULLETIN OF THE BUREAU OF FISHERIES 
In Tables 2 to 29 are given in detail the results obtained from each tagging 
experiment listed above. The actual date of recapture (or, more properly, the 
date on which the recapture was reported) is given for each locality, and at the 
bottom of each column is given the total number reported from each locality, the 
percentage of the total number tagged taken in that locality, and the average time 
that elapsed between the time of tagging and the reported recapture. Careful con- 
sideration has been given to the relative advantages of calculating the percentage 
returned from each locality on the basis of the number tagged in each experiment 
and on the basis of the total number returned from each experiment. If these 
percentages are based on the number returned, it is implied that the fish tagged 
and, in fact, all the fish found with them at the time of tagging were distributed 
only to the localities from which tags were reported and in the percentages given. 
This does not give consideration to the possibility that fish may have gone to 
localities were no fishing is conducted and from which no returns could be expected. 
Furthermore, the percentage of fish taken in a given locality may be profoundly 
affected by extreme variations in the percentage taken in another locality if these 
percentages are based on the number of fish returned. By basing these calcula- 
tions on the number tagged these difficulties will be avoided and the results of the 
separate experiments are strictly comparable. Accordingly, this has been done. 
In determining the average time en route the median has been selected as the 
measure of central tendency best describing the average time required by the fish 
in passing from point to point. The median seems best for the purpose, since it 
gives less weight to extreme measures than does the mean, yet is not subject to the 
instability of either the empirical or the theoretical modes. The median seems 
especially appropriate for use in handling such data as we have to work with in 
this paper, in which a few of the frequency distributions are extremely skewed 
(those provided by the returns from points close to the place where the tags were 
attached) , while others are approximately normal. The returns from some localities 
are numerous and from others we have only a few. 
The interpretation of the median as used for this purpose is very simple. It 
may be given thus: At the expiration of the median time as given in the tables 
one-half of the total number of returns reported from that particular point had been 
taken. 
Table 2. — Returns from 120 red salmon tagged June 2, 3, and 5, 1923, at Kelly Rock trap, Unga 
Island. Recaptures, 13 ( 10.8 per cent ) 
Date 
Unga 
Island 
Loca 
Ikatan- 
False 
Pass 
ity of reca 
Egegik 
pture 
Naknek 
Kvichak 
•Tune 8 
1 
1 
June 11* 
2 
June 14*-- _ 
1 
June 16* - 
1 
June 17- 
1 
June 22* 
1 
June 26* 
3 
June 29*-- _ 
1 
July 25*-- 
1 
Total 
1 
6 
2 
1 
3 
Percentage returned 
0.8 
5.0 
1.7 
0.8 
2.5 
Median time in days 
3 
8.5 
21.5 
52 
23 
Note. — I n this and subsequent tables the asterisk is employed to indicate a break in the continuity of the table. It indicates 
that one or more dates have been omitted from the table just preceding the date thus marked. No returns were recorded on these 
omitted dates. 
