On the dust jacket of Kent Staley's excellent book, The Evidence for the Top Quark, there is a quotation from my manuscript review stating that the book is “a significant contribution to the history and philosophy of both experiment and science.” Having identified a source of possible bias, an issue in science that Staley discusses in some detail, let me state that on my second reading, I found it to be an even better book. It is both informative and enjoyable to read.

Staley's book is not a complete history of the discovery of the top quark. It is, rather, a detailed account of how the Collider Detector at Fermilab (CDF) group came to their conclusion that they had evidence for the top quark. Staley's fascinating account is based not only on the published papers, but also on internal group documents and extensive personal interviews. The study of the early work by the CDF group notes the important difference between having evidence for the top quark and their later claim—based on more data and on work by another group (D-zero)—that the groups had made an ‘observation’ of the top quark.

Along the way, Staley gives us an interesting account of the theoretical background to the experiment, the story of how the detector was designed and built (along with the decisions that were made in the process), and how the paper was written. In discussing the latter, Staley makes the point that writing such a paper is not ‘writing up’ a result, but instead an important part of determining what that result is. The last two chapters discuss many of the issues raised in his historical account from the perspective of the error statistical view advocated by Deborah Mayo.

There are so many good things in this book that it is impossible to discuss all of them in a brief review. My own favorite is the discussion of possible bias in the analysis of the data. Staley shows us that in a large, complex experiment involving hundreds of physicists, there are many decisions concerning which analysis procedure to use on the acquired data, and that within such a procedure, physicists must decide which values of the selection criteria, or ‘cuts’, to use. In the CDF experiment, there were two different types of possible experiments: counting experiments, in which the number of candidate events were counted and compared to expected background, and kinematic experiments, which examine the detailed characteristics of each candidate event. Within the counting experiments there were three different methods of ‘tagging’ top quark candidates: secondary vertex tagging (SVX), soft lepton tagging (SLT), and dilepton events. Within each analysis procedure, one had to decide on the values of the selection criteria. This was often done when the effects of the criteria on the number of events was known, thus allowing for the possibility that the cuts were tuned to achieve the maximum number of events. In the SLT analysis, for example, one of the cuts was on the minimum value for the transverse momentum of the lepton. Should the cut be made at 2 GeV/c or 4 GeV/c? The 2 GeV/c cut allowed more candidate events, but also increased the background. Members of the group worried that the cut had been tuned. (Although Staley presents lots of technical detail, the reader who is not familiar with the arcane details of high-energy physics, or even physics in general, will get lots of value by reading the chapters as stories.) Staley notes that this was a source of possible bias and presents us with some of the internal group discussions on how the decisions were made. He also discusses this philosophically by considering the question of predesignation, which is setting the value of the cuts in advance. Unfortunately, in such a complex analysis, one may not even know that the procedure is working properly, or what values are reasonable for the cuts, without examining real data. Monte Carlo simulated events may not be adequate, precisely because one doesn't know the exact characteristics of the target events.

Staley also discusses the history of the theoretical background to the experiment, including the fact that a philosophical view, in this case, Marxism, can have an influence on the theoretical approach taken. He argues persuasively for such an influence in the early papers. However, by the time of the Kobayashi and Maskawa paper (which suggested the existence of the top quark), this influence seems to have disappeared. Staley also discusses the question of whether American scientists overlooked that paper because it was written by Japanese physicists and published in a Japanese journal. He shows, however, that although there were few early citations of this paper, after it was cited in several western papers, it became one of the most cited papers in the high-energy physics literature. Also, arguing against such a bias is the fact that the paper was not often cited in the Japanese literature.

There are also other interesting historical discussions, including discussions concerning the design and construction of the CDF detector, the contributions of the D-zero collaboration, and others, but space limitations preclude further discussion.

Staley has also made some important philosophical points. I found particularly interesting and valuable his distinctions for success:

1. 1. Evidential success, in which the experiment severely tests the hypothesis ( $P(e/ h) $ is large and $P(e/ \neg h) $ is small), and the hypothesis passes or fails the test;

2. 2. Epistemic success, in which the experiment gives reasons for believing that the hypothesis is true and the experimenters are justified in believing that statement; and

3. 3. Experimenters’ success, in which the experiment is an epistemic success and the experimenters can provide the appropriate scientific community with a justification that it is an epistemic success.

Staley places his philosophical views within the context of Deborah Mayo's error-statistics approach, and he offers a very nice argument for that approach, although it is not completely persuasive to this Bayesian. He also applies it quite well to the historical case. I particularly liked his argument that the statistical significance of an experiment should increase with increasing data. Interestingly, in the top quark experiment, this did not happen immediately, although by the end of the second experimental run, the significance had increased considerably.

This small reservation does not detract in any way from my view that this is an important book that should be read by anyone seriously interested in the history and philosophy of both experiment and science.