I have two books I like to leaf through from time to time: Michurin's "Conclusions from 60 years of work" (in Ukrainian; original text from 1935) and Lichtenshtein's "Editing a book on science" (in Russian; from 1957). Also, a Russian translation of Kepler's 'On Six-cornered snowflake' and other essays.
They aren't very old, but they are dated. There's a lot of ideology in Lichtenshtein - he quotes Lenin and Co about once a page, and I'm glad he does, because it both annoys me and works. They are comparatively narrow in scope; I likely won't have to use them in practice. I haven't ever read them from cover to cover. But I keep them, because...
They teach you to slow down. The new books tell you what to do and how to do it, yes. However, when you have bought labware, haunted your statistician into celibacy, grown your flies and are ready to sit down and count them - you have to slow down. There are small observations that never make it into articles, cryptic little notes in your journal like 'slide 740 is too dark, but better contrasted than 741 - overheated?' - the kind that makes up a large chunk of the 'intangible knowledge' in your field. Old books might have such asides (sometimes written in the margins by a reader before you), and emphasize the importance of not relying on memory.
They teach you that science is actually possible - in an inexpensive, tried, and criticable way. 'Suppose all your knowledge about a certain domain was deleted, how could you re-invent it...' is all good and well, but there's no way you can just re-invent methabolic pathways from scratch within a currently available lifetime - and no way at all if your brain hasn't been trained to notice things. I know a guy in our Ecology Department who is a great naturalist. One of them who will just as carefully weigh baby seals, measure stamens and collect rocks. That's what you want in people whom you send to Mars (or to Antarctica, in his case) - not merely skill or dedication or high IQ, because it all isn't worth shit if they bring you a damaged sample. He observes things... but for most people, he wouldn't be the best instructor. He rambles, and if you aren't in love with zoology, you won't stay long enough for him to finish. (Also, he's getting on in years and has a heart problem, and so doesn't get out into the field often anyways.) Books are more bearable - and there's more of them. (I think this point has been raised elsewhere on LW.)
They teach you about good illustration. Very often I see, in presentations, mags and even books, colourful photoes of plants - frequently, the flower; but as a botanist, I know I've also got to recognize the species out of bloom! I need to see the whole! Another problem I've run into when interested in someone's experimental setup is too dark images in scanned articles (and useless angles of view, for example if the author had wished to show a rack of test tubes instead of one single fitting.) The old pictures are representative.
They teach you about different, perhaps no less efficient ways of presenting data, too. Not just the sad tombstones with error bars hung upon their tops. Lichtenstein's section on tables is beautiful. He discusses the lengths to which a decent editor can and cannot go when preparing the manuscript for print. He showed me more about writing up the data than anybody in the grad school ever could.
They teach you to be picky. You can't, on your lonesome, much improve the quality of contemporary research in any given area - you're stuck not only with the 'current ways of being wrong' that the Lewis quote has addressed, but also with 'current standards of being right enough' - EY talked about it in the beisutsukai sequence. When I read Michurin, though, I have this chill between my shoulderblades, the feeling that he'd held himself to a much higher standard than I do, even if he happened to be wrong.
They teach you to look outside of your own field and to speak simply. Kepler engages the reader, lightheartedly and gently. Reading him actually made the world seem less cruel, more playful.
People have always had a religious or quasi-religious reverence for nature. In modern times, some people have started to see nature more as an enemy to be conquered than as a god to be worshiped. Such people point out that uncontrolled nature causes a tremendous amount of human suffering (to say nothing of all the misery that it causes other creatures), and that vast improvements to human welfare have largely been the result of us ceasing to love and fear nature and starting to control it.
There are several common responses to this. One response is that it is solipsistic for humans to measure the value of nature in terms of what is and is not good for us. This strikes me as right only insofar as it ignores the welfare of non-human creatures who have enough going on in terms of consciousness and/or sentience to matter; I think the objection would be without merit if one were to broaden the scope of concern to something like all creatures, present and future, capable of having experiences (who else is there to care about?). A second response is that seeing ourselves as highly effective lords over nature leads to dangerous overconfidence, which leads to costly mistakes in how we deal with nature. This is a very fair point, but what it really amounts to is a claim that we shouldn't underestimate the enemy, not that the enemy is really a friend. Anyway, the solution to that problem is to become better rationalists and get better at being skeptical regarding our powers, not to retreat into quasi-mystical Gaia worship. A third response is that getting into a "conquer nature" frame of mind puts people into a "conquer everything" frame of mind and leads to aggression against other people. This might have merit historically, but that problem is also best confronted directly, in this case by more effectively promulgating liberal humanistic values.
So what, if anything, is left to the idea that there is something special about nature worthy of particular regard? And by special I mean something beyond the fact that many people just plain enjoy it the way they enjoy lots of other things that nevertheless have no claim to any special status. I would say that the main thing that makes nature special in this sense is that when you are in nature or contemplating nature, you can be confident that the resulting thoughts and feelings are uncontaminated by all of the (visible and invisible) ideas and biases and assumptions that are present in your particular time and place. When you look at a waterfall and you like it, you can be pretty sure that: (i) it wasn't put there by anyone with an agenda; (ii) you weren't manipulated into liking it by contemporary ideology or social pressure or persuasive advertising or whatever; and (iii) the thoughts that you think while contemplating it aren't the thoughts anyone is trying to lead you into. In other words, nature is a way of guaranteeing that there is a little corner of experience that we are instinctively drawn to and that we can be confident doesn't represent anyone else's attempt to control us. And since other people are trying to control us all the time, even in relatively free societies (all the more so in oppressive ones), this is of real value.
I think the same basic point applies to some other things besides nature. Why do people still read old books* even when the knowledge in them has been refined and improved-upon in the meantime? In many subjects, we don't. Nobody learns geometry by reading Euclid, because there would be no point. But people do still read ancient works of philosophy. It seems to me that one good reason to do so is that for all the ways that these works have been analyzed and surpassed in the intervening years, the reader can be sure that what is written there is not the product of manipulation by the forces that are at work in the reader's own time and place. So it represents another way to gain valuable freedom and distance.
*Here I'm talking about non-fiction books. The merits of old creative works even when the innovations in them have become widespread in newer works is a different story. Often a point like the one in this post still applies, and sometimes the old stuff really is still just the best.