Darwin Catholic has a post about his growing family library. Over the past 5 years the number of books in my "library" has been growing a steady but constant rate. But there are definite patterns in terms of how my collection has been piling up. I tend to keep technical books, but trade in non-technical books after one or two readings. Initially this meant I had a lot of programming related books (still have 'em), but over the past few years I've been depending on online supplements like Safari Library. Now my technical books tend to be more oriented toward science, the collected papers of Motoo Kimura or W.D. Hamilton, or D.S. Falconer's Introduction of Quantitative Genetics. My non-technical books I often take to the used book store and try to trade them in for more technical books. For example, recently I exchanged Future of Freedom & Lust in Translation for Scientific Reasoning: The Bayesian Approach. The result is over time my small library has a larger and larger proportion of books on genetics & mathematics and a constantly changing cast of works of history, sociology, psychology, etc. The rule of thumb that works for me is that I tend to retain books with a high load of contingent theoretical information, and am casual about trading in those works which are dense on loosely related fact. Across the sample space of facts I have rather good recall and I don't usually refer back to the original works from which I derived those facts because it is usually quicker to go to the internet and look up an isolated datum for confirmation then to rummage through the index of a book where it might be found. In contrast, theoretical constructs are harder to retrieve on the web in a fully fleshed out form without a lot of effort,¹ and quite often without regular usage I forget how to implement the detail of technical methods. I still remember the general outline of my course on the Peasants' War of 1525 as well as my linear algebra classes, but in the former case individual errors in recall don't result in the collapse of the whole structure of knowledge. In the latter case I've had to revisit linear algebra texts for details because small errors compound so as to make my recall close in worthless in many cases in implementation.

1 - A lot of the introductions to models on Wikipedia & elsewhere seem a bit simple. Then, when you go to papers there is the implicit assumption of a lot of background detail. So it seems to me that there is a space on the web for mid-level formalism and theory which is often missing, and can be most easily found in textbooks which you wouldn't normally find in your local public library.

Labels: Personal