Page 198 - J. C. Turner - History and Science of Knots

Studies on the Behaviour of Knots 187

Reliability of Data
Few investigators give any indication of the reliability of their data. The
first people to measure knot strength, the British Alpine Club [2], simply state
that their values were the mean of, usually, two tests; Wright and Magowan
[36] stated that they made 7-13 tests of each knot, and give values of 1.2-
13.2% for the `percentage variation'; Microys [27] gives a range of values, but
does not say how these were calculated (they have obviously been rounded off);
and Wildsports [35] state that their values have a `precision' of ±3%. Neither
`percentage variation' nor `precision' are standard statistical terms, and their
meanings in this context are uncertain. In some of his tests, Barnes [7, p. 136]
used six replicates each time; he claimed `the extreme differences in averages
was less than 2%', but also said that he had rejected any `erratic figures', on
the not necessarily correct assumption that the knot had been wrongly tied.
Others of his tests had many more replicates [7, p. 123]. But none of these
data give any certain guide on how big a difference between values is needed
to reach any set confidence level. All other investigators report a single value
only, and do not state whether these are results of single tests or means of
several. A few anglers' knots have been stated to have 100% efficiency, e.g.,
[11, p. 86], but in the absence of confidence limits these figures are difficult to
interpret. Few investigators give any technical details of their tests; however,
I would expect the errors due to machine operation to be much less than those
due to the variability of the knot and rope samples and the factors mentioned
in the next section. Rope is expensive and the tests time-consuming, so that
there are obvious economic limits on the number of replicates of each test
that can be made. But there seems little worth in publishing values with
no indication of their possible variation or of the differences between values
for different knots needed for confidence in the results. Otherwise, all we are
saying is that knots weaken rope. In what follows, I take the published figures
at face value.
Cogency of Results
It is not certain that all investigators were testing exactly the same things,
since the details of the tests were not published. The variables most likely to
be important are the way the knot is tied and the load is applied, and the
nature of the rope used; sometimes the environmental conditions might have
been significantly different.
Knots: One investigator said: [26] `it was found impossible to tie a Bow-
line in exactly the same manner, no matter how proficient we were'. Barnes
[7, p. 124] said that hours of practice were needed before he could tie his
anglers' knots with a uniform breaking strength. No others have commented
on any such difficulties. It is important that all knots should be properly
dressed, arranged to give all parts smooth curves without twists or kinks,