This book has many virtues, or at least its first few chapters do. There is probably no other book which presents such a clear and detailed history of the attempts at classifying languages into families, both the successful attempts and the unsuccessful ones (though Ruhlen seldom acknowledges any unsuccessful ones). The reader will find a great deal of useful information here on this enterprise, information which is hard to find elsewhere.
However, as is well known, Ruhlen has a number of large and dangerous bees in his bonnet, and those bees gradually take over the book. By chapter 7, real linguistics has been left behind, and we read only about Ruhlen's bees.
The huge shortcoming is Ruhlen's willingness to accept, in a wholly uncritical manner, just about every speculative mega-grouping of languages which has ever been proposed. This weakness appears in the earlier chapters, as Ruhlen unhesitatingly accepts a series of increasingly dubious "families', like Uralic-Yukaghir, North Caucasian, Khoisan, Austric, and Indo-Pacific. It comes to a head in chapter 6, where Ruhlen wholeheartedly endorses the vast but shriekingly speculative "Amerind" family proposed by Joseph Greenberg. Chapter 6 is by far the poorest of the first six chapters, and the reader will find much better information on the classification of American languages in Lyle Campbell's American Indian Languages (Oxford, 1997).
Then, from chapter 7 onward, Ruhlen's bees take over entirely, and the book falls apart. Ruhlen's familiar lack of understanding of linguistic methodology comes to the fore, and he descends into increasingly ridiculous claims about method and about results. True, he notes the hostility of professional linguists to the speculations he defends, but he fails to tell the reader anything much about the powerful reasons for that hostility, and he attempts to present the objections as resulting from little more than bad temper and supposed incompetence.
Ruhlen's profound lack of understanding of the formidable difficulties involved in relating any languages at all reaches a nadir on page 383, where he draws a preposterous parallel with biological classification, suggesting that identifying a language family is a task on a par with recognizing a class of butterflies. He should have pursued this analogy: he might have found out just how difficult and controversial biological classification really is.
As always, Ruhlen wants the reader to believe that languages can be successfully classified by the mere collection of miscellaneous resemblances -- which they cannot, as every professional linguist knows all too well. Waving away the laws of probability, he assures us breezily, on pages 255-256, that chance resemblances among languages are unlikely, and that they can be dismissed from consideration. But anybody who who has looked carefully at a few languages knows that chance resemblances are enormously frequent and statistically unavoidable: we have only a few speech sounds with which to construct thousands and thousands of words in every language, and chance resemblances are always with us. Consider English 'much' and Spanish 'mucho' ('much'), which are unrelated, or Italian 'due' ('two') and Malay 'dua' ('two'), which are unrelated, or Basque 'elkar' ('each other') and Dutch 'elkaar' ('each other'), which are unrelated.
Impervious to criticism, Ruhlen ventures to classify all the world's languages into just a few "families": 17 on page 258, and then only 12 on page 390. Readers should be aware that these "families" are, in most cases, no more than Ruhlen's pipe-dreams. Real linguists recognize well over 300 established families, and reducing that number by even one is an almost Herculean enterprise, requiring vast amounts of painstaking work. But Ruhlen doesn't believe in hard work; he believes only in collecting miscellaneous lookalikes from the pages of bilingual dictionaries. For Ruhlen, comparative linguistics is a trivial task, requiring no training, no experience, and no knowledge of the languages being classified, and he advocates ignorance over knowledge.
There are a few irritations even in the sensible sections, such as Ruhlen's (acknowledged) eccentric use of 'Indo-Hittite' for what the rest of the world calls 'Indo-European' -- a use which may bewilder innocent readers.
In sum, this book is a largely reliable source of information on the history of attempts at classification. But Ruhlen's grandiose conclusions, and indeed everything after chapter 6, is best ignored: it's fantasy, no more.