This is an excellent and approachable introduction to the Lambda Calculus and its relevance to computer science. It introduces the calculus as a means of describing computer programs, and not merely as a mathematical construct. In the space of just over a eighty pages it develops the mathematical foundation for the calculus without the all too common cycle of theorem statements and proofs that can exhaust many readers. It also develops enough syntactic sugar for the calculus that the result should look familiar to any functional programmer. The remainder of the book defines the semantics of the calculus, and then shows how to implement these semantics on traditional computer systems (including shared memory multiprocessors) using graph reduction. Oh, and while at it, it also includes a proof of the Church Rosser theorem and an overview of the typed lambda calculus in appendices. That so much material is covered so succinctly, while remaining comprehensible, is what recommends this book.
Readers should remember that this book was published nearly a quarter century ago. Functional programming is now more widely accepted (consider Scala, Erlang, Haskell, F#, ML, and so on) and so this book often feels dated. Most noticeably the syntax chosen for lambda application requires far too many parenthesis. Apparently this was chosen to make predictive parsing easy, but I sorely missed the curried, left associative style for function application that is favored today. I also have to say, that with all due respect to John Backus and his Turing Award lecture, that the references to FP do feel like an anachronism. The extensive discussion of incremental garbage collection near the end of the book also seems out of place. Even the typeface and code formatting conventions hark back to an earlier age of computing (I'm not suggesting this is for the worse, rather that it is just different: imagine reading ALGOL-60 or MIX code today).
In the end though the dated feel of the text is a relatively minor annoyance. Readers just need to remember that all really good mathematics is timeless, even when it is expressed in programming terms that are not!