Whichever road it [evolution] used, it appears to have happened repeatedly, because swim bladder genes evolved and converged four times in teleost fish, providing many structures from which a lung could develop. Some estimate that fish overall evolved air breathing 68 independent times. These fish took evolutionary paths that differed in the details, but they reached the same destination dozens of times, predictably.
Graham concludes, “air breathing has independently evolved among the fishes at least 38 times and perhaps as many as 67 times,” a point that Karel Liem echoes in his analysis of ABOs [Air-Breathing Organs].
"Graham" is Jeffrey B. Graham, author of the aptly titled book Air-Breathing Fishes: Evolution, Diversity, and Adaptation. The book was published before high-throughput genomics techniques, such as the analysis that gives evidence for four convergences in teleost fish mentioned in the previous sentence. As such, it's based on classic biology: comparing lots of fish and noting just how many fish breathe air, using organs that have many, many different shapes with one oxygen-gathering function. You could rewrite Red Fish Blue Fish with all the different fish that breathe air in Graham's book (an idea I'm just going to leave out there for free considering the burgeoning children's convergent evolution market).
Notice also that I was content to use the hedge words, "Some estimate," and that's good because I made another slip of the keyboard, putting 68 instead of 67. (I try to type independently to avoid plagiarism and I must have fused the two numbers from the quote maybe? Sigh.) But the important word is in the final sentence of my paragraph above: whether 38, 67, 68, or something else, these are big numbers and are all in the range of "dozens" (as in greater than 24).
That in itself was a dramatic surprise to me, and should be a dramatic surprise to most people. The very fact that we are so surprised shows us that our mental image of evolution should be changed. Evolution is not always an inefficient and undirected process that depends on lucky chances. Or, if it is, there are so many lucky chances that it can be counted on to increase resource efficiency and even complexity. Conway Morris refers to evolution as a "search engine" working on a planetary level, returning complexity and intelligence. That's not what most people think of when they hear "evolution" -- and yet that's what Conway Morris and Graham see, and it's one of the points of my book. Here's one more bit of evidence to back it up.