It's an interesting book, and a useful one because everything is backed up by studies and empirical research. It has three different categories: laws, hypotheses and conjectures.

The laws are solid and unobjectionable, if a bit stodgy. A few examples from the appendix:

A system that is used will be changed. (Lehman)
Testing can show the presence but not the absence of errors. (Dijkstra)
Good designs require deep application domain knowledge. (Curtis)

The hypotheses are a little looser:

Object oriented programs are difficult to maintain (Wilde)
Object oriented designs reduce errors and encourage reuse (Booch)
Group behavior depends on the level of attention given (Hawthorne)

And the conjectures are much the same:

Distribution ends where the customer wants it to end.
Process improvements require action based feedback.
* Measurements are always based on actually used models rather than on desired ones.

Which is fine, as far as it goes. You can't really argue with the conclusions. However, it is slightly out of date, even having been published in 2003 -- no-one really objects to object oriented programs these days, and discussion about CASE tools is more along the lines of "where are they now" rather than an actual discussion. However, there are solid sections on static and dynamic verification that work very well as an introduction to methods that many programmers may not be familiar with.

This isn't a book you'll pick up and read every weekend. It's far too theoretical and abstract. However, it's an excellent way to formalize thought and use as a touchstone when you're concerned about the theory.