This is a very interesting book arguing that Bach mainly used only 1 singer per part in the performances of his own vocal works. The document most cited in this book is the "Entwurff" of 1730 written by Bach to the Leipzig town council. In it, Bach expressed his discontent with the state of the musical resources allowed him by the council; specifically there Bach requests for at least 3 singers per part for each of his four choirs. It is that request that some use to dismiss the use of only 1 singer per part as advocated by Joshua Rifkin in 1981.
The arguments in the first half of the book concentrated on Bach's extant scores (eg, nearly all vocal works had only 1 set of vocal parts), iconography (eg, illustrations of singers not sharing scores), and contemporaneous accounts of Baroque practices. At this point of the book, I was not completely convinced by this information although I was convinced that when Bach asked for 3 per part, he did not mean 3 per part in a performance but rather 3 singers per part in a pool from which he could select from. The arguments in the second half of the book focused on resource (eg, what was available to Bach), balance, and instrument-singer ratio (eg, how many singers and instrumentalists were employed). One interesting consideration presented in the second half is the fact that doubling the number of singers does not double the volume of sound but rather adds sonority and only modestly increase volume. The arguments from the second half are far stronger in my opinion and they strengthen and validate the arguments from the first half. The book is very well illustrated and includes the original text and a translation of the 1730 Entwurff.
Joshua Rifkin's 1981 recording of the Mass in B minor used 1 singer per part. Andrew Parrott's own recording in 1985 used a mix of 1 and 2 singers per part which I consider probably the most satisfying recording of the Mass in B minor that I've ever heard. Rifkin later went on to record a number of Bach cantatas with 1 singer per part. The argument from the 1700s that the use of one singer per part brings clarity and beauty is spectacularly evident in these recordings. For me, the musical result of using 1 or 2 singers per part trumps any historical argument.
Unfortunately for various reasons that I won't go into but are related to the size and nature of today's music venues, 1 singer per part does not work as well in live performances as it does in a recording.