Excellent for two parts, then it becomes a filing cabinet.

Strengths first, because that is the correct order.

The opening two parts do the thing a book like this is supposed to do and mostly doesn't. He describes an experiment, gives the actual numbers, says what he thinks it shows, and then says where he has since been argued with. I counted eleven places in the first 150 pages where he flags his own uncertainty. That bought him an enormous amount of patience from me later on, and he needed it.

Yellow tabs, ranked by how often I've gone back to them:

1. Anchoring. The wheel-of-fortune study, where a plainly random number drags people's estimates around, is the cleanest four pages in the book.
2. The planning fallacy. I read the chapter, then opened my kitchen renovation folder, then read it a second time.
3. Prospect theory, laid out well enough that I could explain it to my father, who is 74 and did not want it explained.
4. The remembering self against the experiencing self. That one got a blue tab too, because it's also the hinge where the book changes shape.

Now the faults. There are three and they are all the same fault wearing different labels.

1. It runs roughly a third too long. Parts Three and Four re-present earlier material under new headings. I timed it, because I time things: the first half took nine hours across eleven evenings, the back half took eleven hours and I retained less of it.
2. The replication trouble is not entirely his fault but it is entirely my problem. Several of the priming studies arrive as settled and haven't held up since. I looked four of them up, marked them green, and now I genuinely don't know what to do with the chapters they're sitting inside.
3. He assumes the reader wants to be told what to do at the end. I did not. The material on institutions and checklists reads as though it was written for a different audience, one with a budget.

Why is Part Four in here at all? I'd put that question to an editor rather than to him. Somebody clearly decided a book this size needed an applications section, and the applications section is the thinnest writing in it.

I read the whole thing over about six weeks of commuting, twenty-two minutes each way, which is a bad way to do it; the chapters are longer than the journey and I kept losing the thread between the platform and my desk.

If you're only going to read part of it, take Parts One and Two and stop after the endowment effect. I've given that advice to three people. None of them took it, and two stalled around page 300 exactly where I said they would.