Shifting Standards: Experiments in Particle Physics in the Twentieth Century

Introduction

• 1 The history of the changing standards for scientific discovery in particle physics discussed in the prologue suggests that it might be interesting to investigate whether other aspects of experiment and the reporting of experimental results in that field have changed with time. If we look at the history of the use of statistics, we find considerable additional change. As Jed Buchwald (2006) remarked in his discussion of the treatment of discrepant experimental results, “well into the eighteenth century experimenters chose to publicize that single golden number which they deemed to be the very best one of all the values that their labor had produced” (566). This statement of results changed over the next several centuries into more standardized treatments of data. Initially, other techniques were sometimes used to deal with discrepant results. For example, some scientists , believing that, as they proceeded with a measurement, the later results were better and more reliable than the earlier ones, used a procedure in which they first took the mean of the first two measurements. They then proceeded to take the mean of this first mean with the third measurement and so on. This procedure had the effect of more heavily weighting the later, and presumably better, measurements. We can see an illustration of the increasing reliability of measurements if we examine Robert Millikan’s measurements of the charge of the electron as a function of time (figure I.1). Although Millikan did not use this procedure to calculate his value of the charge of the electron, we see that the spread of the values becomes smaller and that the values converge as he made more measurements. The percentage of events included in his published paper also increased with time. Isaac Newton was, if not unique, certainly rare in his use of the actual mean of a set of results as the best value, although one could not provide a mathematical justification for this until the development of probability Introduction 2 • Introduction theory and least-squares fitting in the early nineteenth century. This standardization in analyzing results was strengthened in the late nineteenth century with the use of the standard deviation and the invention of the χ2 test by Karl Pearson. The word “standard” is important here. By analyzing data with the same technique, scientists could compare experimental results or at least see whether there was intersubjective agreement. Experimenters throughout the period studied in this book are consistent in providing estimates or calculations of experimental uncertainty,1 although the mathematical techniques used vary considerably. Thus, Kennelly and Fessenden (see chapter 1), although not providing an uncertainty, presented the maximum and minimum values obtained, which gives an estimate of the experimental uncertainty. Hall and others provide an estimate based on the calculation of the probable error or the standard deviation. The prologue shows the use of standard deviations as both a measure of uncertainty and of significance. More recent experiments calculate the maximum likelihood of the distribution of events given various hypotheses and use Bayesian decision trees, neural nets, and multivariant analyses in presenting an experimental result, estimating the uncertainty, and providing an estimate of the significance of a result. The use of statistics is only one of the important issues involved in the presentation of experimental results. It seems worthwhile to examine these other issues to see whether they have also changed over time. In this book I examine several of these issues by looking primarily at papers published Figure I.1. Millikan’s values for e, the charge of the electron, as a function of time. Introduction • 3 in Physical Review, which began as the journal of record for the American Physical Society and is now one of the major archival journals.2 I begin with a paper published in 1894 (volume 1 of Physical Review was published in 1893) and continue up to the present, looking at papers at approximately ten-year intervals. Such a history can provide only snapshots of experimental practice at a given time, but, like snapshots of a vacation trip, they can be of value. These snapshots will also provide a feel for the practice of experimental science and for the style of scientific papers at various times. I will discuss, almost exclusively, experiments concerned with elementary particles and their properties. This is because some of the issues, particularly those of scale, are most apparent in that field, although the other issues discussed also apply...