Hair Breakage 276 Variability and statistical significance in Weibull results: As seen from the S-N plots, there is considerable variability in fatigue data, which introduces concerns about the statistical significance of results obtained by this approach. In addition, there may also be questions about the number of replicate samples needed to yield results from which we are comfortable making predictions. Theoretical studies suggest that especially high numbers of replicate samples are needed to confidently detect relatively slight changes in breakage tendency. This is not disputed, but approaches can be used to improve our confidence in these results. Firstly, as will become apparent, fatigue testing results tend to yield considerably larger differences between samples than conventional stress-strain testing that is, it is more common to be comparing results that differ by one or two orders of magnitude, as opposed to 10‑15%. Secondly, running systematic experiments allows for comparing trends, rather than placing too much emphasis on any particular data point. By means of illustration, Figure 17 shows a plot of characteristic lifetimes as a function of repeating stress for Afro hair at 60% relative humidity. The black data points represent analyses based on 50 fibers per stress range (i.e. 350 total data points), while the gray points are based on the presence of 75 fibers per stress range (i.e. 525 total data points). In each case, shape parameters are also shown. The regression coefficient (R2) for results based on 50 fibers per data point is 95.5%, thus indicating strong evidence for the trend line. It is apparent that there may be considerable uncertainty in any point itself (for example, results at 0.0115 and 0.0125 g/μm2 appear somewhat high, while those at 0.0135 and 0.0145 g/μm2 may be somewhat low), but nonetheless, there is high confidence in the trend. Upon increasing the number of fibers to 75 at each condition, a number of points converge to a slightly altered regression line, with the coefficient rising to 96.7%. By the existing methodology, this additional data collection represents approximately two months of work, which leaves us more confident in the absolute numbers, but does not alter the overall conclusion. It is also emphasized that this exercise has been performed on Afro hair, which is perhaps a