RESEARCHES ON RACES OF HERRINGS. 85 
whole cm. division line, for one or two cases of this kind were actually 
observed in time to prevent this error. It is probably very difficult to 
exclude completely occasional lapses of this nature in examining large 
numbers of fish at the high rate of speed required. 
It is unfortunate in some respects that the experimental sample men- 
tioned above was not a fresh sample exactly comparable with a research 
sample, and it would probably be better in future work to re-examine a 
batch of the research sample in order to determine the error of measure- 
ment. 
An experiment was carried out to determine how accurately the instru- 
ment would measure. A number of slips of paper (30) were ruled with lines 
parallel to one end, which was placed against and was parallel with the 
‘“end-board ” of the instrument. The points measured on the ruled lines 
were similar to those measured on herrings, so that in measuring them it 
was necessary to move the instrument about in approximately the same 
way as when measuring the research herrings. The slips of paper repre- 
sented in fact paper herrings. These 30 slips were measured twice at a 
rate greater than the maximum rate at which the research samples of 
herrings were examined. Characters 1 and 2 were read to -2 mm., and 
Characters 3 to 9 to the nearest -5 mm., just as in the research sample. 
The average difference between the first and second measurements of 
Characters 1 and 2 was less than -1 mm., and in only 4 cases was the 
difference as much as -3 mm. The average difference between all the 
measurements of Characters 1 and 2 and the actual distance—as measured 
by a 15-cm. ivory rule divided to fifths of a millimetre—was also less than 
‘1 mm., and in only 5 cases were there differences of -3 mm. 
The average difference between first and second measurements of 
Characters 3 to 9 was less than -1 mm., and in only one case was the 
difference more than -5 mm. The difference was exactly -5 mm. in 30 
cases, and in 159 pairs of measurements the results were exactly the same. 
In all the measurements of Characters 3 to 9 the average difference from 
the actual distance measured was about -13 mm. 
In a large number of measurements, however, it is considered that the 
instrument may be taken as reading accurately on the average, since the 
plus and minus variations would tend to balance each other, although 
ranging between plus and minus the maximum error mentioned above. 
The average algebraical error of all the measurements taken in the 
paper-herring sample mentioned above was less than +-03 mm. in 
Characters 1 and 2, and +-04 mm. in Characters 3 to 9. 
