Following up on a discussion, Arnold Kling asks:
Should we approach famous thinkers by digesting distilled versions, or should we study them in the original?
I'm for distilling, for reasons Arnold offers, but I'm also for reading the originals. Here are a few reasons why, drawn from a number of longer sources I have read and digested:
1. Secondary sources are unreliable and they do not capture or understand many of the original insights. To remove it from the distant past, what I get from John Rawls or Robert Nozick is quite distinct from what I get from their distillers.
2. Truly great thinkers require numerous distillers. Can you read just one book on Keynes? No. So you have to read a few. Shouldn't one of these then be Keynes himself? Yes.
3. The errors of top thinkers are often more interesting and instructive than their successes. Distillers have a hard time capturing these errors and their fruitfulness.
4. We often read great thinkers not to learn what they understood but also to set our minds racing and to find interesting new questions. Great thinkers are usually better at supplying this service than are their distillers.
5. Sometimes the value is in having read common sources and benefiting from the commonality per se. Great thinkers are usually more focal than any of their distillers and thus reading them is a good input for discussions with others.
6. Original sources often help you challenge or reexamine your world view or intellectual ethos. Distillers very often pander to that world view, while pretending to challenge you.
7. Consider a simple comparison. You can read either Adam Smith's two major books or any ten or even twenty books on him, toss in articles if you wish. It's a no-brainer which you should choose.
8. The best distillers often are original sources in their own right (and in part unreliable expositors), such as in Charles Taylor's excellent book on Hegel.
9. Distillation works best in very exact sciences, such as physics and mathematics. If you rely on distillation for an inexact science, you will do best at capturing its exact parts. You will be left with a systematic bias, and knowledge gap, regarding its inexact parts.
I could say more, but I fear this post is already too long.