Reviews of "Inferring Phylogenies"

There were six online and ten published reviews of the book. As reviews appeared, I put some reactions here. I have provided links to the actual reviews, but these may work only if your institution has the appropriate subscription.

• A review by "wiredweird" on the Amazon.com web site for my book. This review is impressed by the book and calls it a "goldmine". The reviewer gives it a full 5 stars, and mentions "this book's thud factor -- over 600 pages". It is a useful review which does not have much in the way of criticisms of the book, except to mention that is "reads like an encyclopedia" and that people who just want to run phylogeny programs may not want this much detail. This is fair enough, and worth warning them about. The review may be read here.

• Another review by "lynn" on the Amazon.com web site. This review gives it four stars and otherwise is just a single statement, saying only that "The book I bought is first printing version. Lots of typo inside..... I should correct them myself.-:( " Having just finished correcting over 200 typos for the second printing, I can only sympathize and hope that the buyers of the second printing will not have reason to be concerned about the number of typos.

• A review in Science by Fredrik Ronquist:
  
  • Ronquist, F. 2004. PHYLOGENETICS: A Broad Look at Tree-Building. Science 303 (5659): 767-768 (Feb 6 2004).
  Ronquist's review is mostly fair-minded and says many nice things about the book, though the large middle section of the review is mostly devoted to complaining about things he doesn't like, notably my failure to consider Willi Hennig and classification central to the field. He also finds my coverage of Bayesian methods inadequate. Then he finishes up by saying more nice things, almost as if to compensate. A few comments in reply to these points:
  • While Willi Hennig played a very important role in getting working morphological systematists to appreciate the utility of well-defined logical approaches, by the time his work was published in English, work on numerical phylogenies was ongoing, and he had no substantial influence on numerical methods, except through his influence on Farris and Kluge.
  • Systematists get so worked up declaiming the centrality of classification in systematics that I have argued the opposite. Well, we'll see what things look like a few years from now.
  • As to what should be called "Wagner parsimony" and what should be called "Fitch parsimony", my terminology may be mistaken (or maybe not). "Fitch parsimony" (in Ronquist's sense) was actually first used by Eck and Dayhoff (1966), though they did not explain their algorithms. Wagner didn't invent Wagner Parsimony (as I note in the book), but Farris and Kluge called it that, and gave algorithms that worked for ordinal scales and thus included the unordered two-state case. Fitch gave the algorithms for the unordered 4 state case (it works for n states too), and Hartigan (1973) for n states allowing multifurcations, and he provided the first proof as well. It might be best to use "Wagner parsimony" to denote the case of an ordinal scale, but it is not obvious whose name should be used for the unordered n-state case. Certainly Fitch's name is the leading contender.

• A review in Trends in Ecology and Evolution by Mike Steel:
  • Steel, M. 2004. Inferring phylogenies: an epic worth the wait. Trends in Ecology and Evolution 19 (issue 4): 173-174 (April 2004).
  Mike Steel has been a major contributor to phylogenetic methodology, including many contributions to quartets methods and Hadamard conjugation methods. His gracious and strongly positive review emphasizes the detail and breadth of my book. The only criticism he has is that is that it will be useful "less as a textbook than as a repository of details, references and background". I developed the book for use in a course, and it seems to work for that, but perhaps most courses would not have the same level and concentration on theory that my course does.

• A review by Stephen Wooding in the American Journal of Human Genetics:
  • Wooding, S. 2004. Book review: Inferring Phylogenies, by Joseph Felsenstein. American Journal of Human Genetics 74: 4074 (May 2004).
  A brief review, quite positive. Written in a rather clinical style, which is perhaps appropriate for the journal. Wooding's only complaint is a justified one: that I did not provide discussion of examples in which real data sets were analyzed.

• A review by Eddie Holmes in the Quarterly Review of Biology:
  • Holmes, E. C. 2004. Book review of "Inferring Phylogenies", by Joseph Felsenstein. Quarterly Review of Biology 79: 204 (June, 2004).
  A brief but glowingly positive review. I am particularly pleased with the analogy to Crow and Kimura's classic text on theoretical population genetics. I watched them write that book from the early 1960's on, critiqued some chapters for Jim Crow, and was pleased to be well-acknowledged in their Preface. I modelled my level of presentation and writing style in great measure on theirs, and hoped to have an effect similar to theirs. It is gratifying to see that connection made.

• A review by Jonathan Badger on the Amazon.com web site for my book. The review gives it four stars. It contains a brief description, but then is primarily concerned with complaining about my treatment of Bayesian methods. He is surprised at my "rather acrimonious treatment", and that I complained about them "on philosophical grounds" when I had so often been the target of philosophical arguments favoring parsimony methods. Well, my perception of Chapter 18 is different. Badger argues that Bayesian methods "will win or lose based on how well they work in practice". I don't think my arguments are philosophical ones, except in the sense that everything is philosophical. When I argue that your prior and mine might not be the same, and when I argue that the only valid issue distinguishing Bayesian and likelihood methods is the presence of a prior, those don't look like philosophical arguments to me. And in making arguments in favor of statistical methods (such as arguing about consistency) we are not simply withholding judgement to see which method works well in practice.

• A review by "folderol50" on the Amazon.com web site for my book. The review gives it only two stars, and consists, in its entirety, of the statement that
  

  "This book, although apparently containing everything, is written in a very opaque style which makes it impossible to simply read through. It probably is a good reference to look in for particular topics, but it is not at all usable as an introduction."

  I can't judge the matter objectively myself, of course, but I have gotten many reactions from readers who find the book readable by people at elementary and at advanced levels. My writing style is usually described as "clear", even when I might rather aspire to having it be described as "elegant". I suspect that this reaction is particular to this reader and not general.

• A review by David Penny in Systematic Biology:
  • Penny, D. 2004. Phylogeny in the comfort zone. Systematic Biology 53: 669-670 (August, 2004).
  This is a strongly positive review which says many nice things about the book. I really couldn't ask for a more positive review. Inspired perhaps by Laphroig, it meanders a bit in a charming way (I confess I had never heard of Laphroig before). I agree with Penny that I should have cited John Hartigan's important 1973 paper on the algorithm for counting steps on trees. Penny's argument for the increasing importance of biases as opposed to statistical noise as a source of incorrect phylogenies is a good and important point. I disagree with him that the book probably cannot be used for teaching -- as I said above, in response to a similar statement by Michael Steel, I use the book for teaching myself, so it actually can be used for teaching of a graduate course devoted entirely to phylogenies.

• A review by Alexei Drummond in Heredity:
  • Drummond, A. J. 2004. Inferring phylogenies: an instant classic. Heredity 93: 519. (November, 2004)
  As the title should make clear, this is an extremely favorable review. Drummond says a great many very nice things. He does think that the book can be used either as a teaching text or as a reference book. He does have some criticisms, which he says he can list "if pressed". He finds it "might be a tad sterile for the average biologist", as it lacks discussion of analysis of actual data sets, and he is critical of the chapter on Bayesian inference of phylogenies as too negative, saying that I do "a mild disservice to the continued development of statistical phylogenetics by downplaying one of the primary vehicles of recent progress." He points out the omission of MrBayes from the list of widely-used computer programs at the end of the book.

• A review in Northeastern Naturalist by Barbara Fifield:
  
  • Fifield, B. 2004. Inferring Phylogenies. Northeastern Naturalist 11: 361. (September, 2004)
  This review is simply a description of the topics the book covers. It is only 6 sentences long, and attempts no critique of the book.

• A review by Christine Petersen on the Amazon.com web site for the book

  The review gives the book a full 5 stars. It says that it might not be useable as the text for a graduate course, but rather "functions well as a reference book". It was used by a course she took, using the chapters as supplementary readings. Because of its completeness, she says she would recommend this book as "one title that most people working in phylogenetics would require for their bookshelf."

• A review by Pascal Tassy in the Bulletin de la Société Français de Systématique, Février, 2005, No. 33, pages19-20.

  This is a review in the internal society bulletin. The Bulletin is received by the members of this society. It is a very gracious and entertaining review which praises the book while noting that persons opposed to statistical approaches to phylogenies may not find it satisfactory. The review is in French. I am not quite sure whether to count this as an on-line review or an in-print review since I think the Bulletin may primarily be distributed on-line.

  Some points:

  • Tassy wonders why I chose not to cover stratigraphic indices that measure the fit between a phylogeny and stratigraphy of fossils. It was partly laziness on my part and partly the feeling that it would not be easy to know how to compromise fit between the phylogeny and the stratigraphy and fit of the phylogeny and the change of characters, when these were not all bookkept in a unified way. I mentioned this, saying (on page 548) that

    I will not describe these measures or the tests of which are best, preferring to concentrate on methods that have some explicit means of compromising the phylogeny and the stratigraphy.

  • He also wonders with amusement whether my coverage of long branch attraction is not excessive compared to coverage of charges that there are countervailing phenomena called "short branch attraction" that favors parsimony. In short: no, my coverage is the proper balance, at least, in my view.
  • In an effort to understand his review (as I only read French in a crude way) I plugged the sentences into Google's French-to-English translator. One of the results was extraordinary, and helps explain the incendiary nature of debate in systematics. Tassy's sentence

    Si vous voulez un manuel pour vous convaincre que la phylogénétique est d’essence statistique, c’est bien celui-là qu’il faut acheter.

    was translated as

    If you want a handbook to convince you that the phylogenetic one is of statistical gasoline, it is well that one which it is necessary to buy.

• A review in Evolutionary Anthropology by Charles Lockwood:
  
  • Lockwood, C. 2005. Seeing the forest and the trees. Evolutionary Anthropology 14: 122-123.
  A strongly positive review that agrees that the book has many strengths while expressing some surprise that I did not spend time on character analysis and polarity, issue of "cladistics" that Lockwood is surprised I did not include, but at the same time he sees that this has its advantages. He ends up by saying that "Felsenstein provides an essential and authoritative reference that already has me thinking about phylogenetic questions in a different way."

• A review in Journal of Classification by F. James Rohlf:
  
  • Rohlf, F. J. 2005. Book review: J. Felsenstein, Inferring Phylogenies, Sinauer Assoc., 2004, pp. xx + 664. Journal of Classification 22: 139-142.
  A positive review which describes the book as "important and useful", one which "should be of interest to many readers of this journal" and which "will be a rich source of ideas for future research". Rohlf has been a major figure in the literature of numerical classification, and on occasion he has contributed as well to the literature of phylogenetic inference. He describes the chapters in some detail. One embarassing inaccuracy he points out is my misattribution of the UPGMA method to Sokal and Michener's paper of 1957. It was apparently first published by Rohlf in 1962 and by Peter Sneath in the same year.

• A review in Evolution by Mike Sanderson:
  
  • Sanderson, M. 2005. Where have all the clades gone? A systematist's take on Inferring Phylogenies. Evolution 59: 2056-2058.
  A review that praises the book for its clear explanations of the mathematical and computational aspects of inference of phylogenies, and for making these accessible to biologists. It contains a number of thoughtful and serious criticisms. Mike is a gentleman and a scholar, with firm views that he is not afraid to state, and this makes his review a good starting point for a wider discussion of the state of systematics. That aside, let me comment on the criticisms themselves:
  • Sanderson suggests that the book is not an adequate review of the "large literature on parsimony-based phylogenetics". I guess I have so little understanding that I don't get as much as he seems to out of the book he suggests as an alternative there, Kitching et al.'s Cladistics: the theory and practice of parsimony analysis. [On reading that book more after writing this sentence I can see that I was being unfair to it -- it is wonderfully clear and will be very useful to those who want an explanation of use of the most widely-used kinds of parsimony methods, as long as they don't want to use a statistical approach or inquire about other methods of inferring phylogenies].
  • He argues (as he and Junhyong Kim have in a paper) that my confidence in methods with fully-specified models is not adequately supported in the literature, which he feels gives as least as much support to rougher and quicker approaches. I think that the last word has not been said on this -- it depends on how well those rougher methods will do in coping with tricky biases that may be corrected by model-based methods. We shall see.
  • He chides me for dismissing the importance of classification (saying that my statements on that at the end of Chapter 11 are "a bizarre thumb in the eye to systematists" which is "badly out of touch with reality"). He points as evidence to the heat generated by controversies among systematists about the Phylocode: I would say that the effort put into this controversy is further evidence that systematists do not have their priorities straight. In their day-to-day work they really do not make much use of higher taxa in classifications, but they show a strange obsession with fighting about them for reasons that seem to me to be an historical curiosity.
  • Sanderson laments my "curious omission of one of the central agendas of modern systematics: the discovery of clades". Clades are very important, but are hard to infer with any degree of certainty. He seems averse to incorporating our uncertainty about them into conclusions we draw from them. I wonder whether that is not excessive confidence in the clades.
  • He says that my "one comment on the most widely used method of rooting trees in real phylogenetic data analysis, outgroup analysis, borders on dismissive" as I pointed out that it amounts to knowing the answer in advance. Well, it does amount to this. If saying this upsets Sanderson, he should ask whether he is not giving the inference of the rooting a status that it does not deserve. (On further consideration, this is too simple. If we have the ingroup and our method gets only unrooted tree topology, then adding the outgroup does let us find out where it connects. But if we are already have both the ingroup and the outgroup in an unrooted tree topology, then using the outgroup method does amount to knowing the answer in advance.)
  • In his concern that I omit the importance of inferring clades, he noted that "in one respect, Felsenstein implicitly acknowledges the importance of clades, because his own bootstrap test delivers confidence levels on clades, not trees." This is an interesting point, but a bit off the mark. For most of our phylogeny methods, the tree specifies partitions (in effect branches of the tree), not clades. To the extent that the rooting is uncertain, bootstrap methods do not deliver confidence levels on clades.
  Sanderson's is the most thoughtful and serious critique of the book yet published, and will be very useful to the fields of systematics and molecular evolution.

• A review by Roger Burks on the Amazon.com web site

  Burks credits the book with being good for computer scientists and for biologists versed in biostatistics, but says that it "is very difficult reading for other scientists who do not fully understand the complex math presented in the text. It also does not give a concinct summary of the assumptions and failings of each method." He feels that it is not good for "students who don't have a good handle on such things." He gives it four stars.

Other reviews that were said to have been coming:

These reviews were at one point thought to be coming. However we have passed the fourth anniversary of the publication of the book. It is unusual to have reviews appear that late. One doubts whether these reviews will ever appear.

• One was said to be coming in Plant Systematics and Evolution.
• Two reviewers were supposed to be writing reviews for Molecular Phylogenetics and Evolution, though it is not clear yet whether this was going to result in one, two, or zero reviews. Zero, apparently.
• A review is supposed to be be coming in Theoretical Population Biology.

Improbable phrases

Incidentally, Amazon compiles a list of SIP's ("statistically improbable phrases" that appear in my book). Their list turns out to be:

polymorphism parsimony, quartets methods, parsimony score, quartets distance, consensus supertree, multifurcating trees, least squares branch lengths, expected pattern frequencies, distance matrix methods, coalescent trees, short quartets, ancestral selection graph, unit branch length, unrooted tree topology, unrooted bifurcating trees, coalescent genealogy, given tree topology, tree rearrangement, character state tree, partial bootstrap, different tree topologies, least squares tree, phylogenetic invariants, same tree topology, postorder tree traversal

I think this is a great list -- it really gives you a sense of some things my book covers that some others might not.

Those interested in a list of typos and their corrections for the book should look at the web page here.

Those interested in the example data sets used in the book will find most of them available for download here.