The book says "The one thing that stands out above all else is the relationship between return and risk. Assets with higher returns invariably carry with them stomach-churning risk, while safe assets almost always have lower returns."

Yet when people do study the market there are ample ways to show that risk and returns are mispriced.

Your statement regarding mispricing would indicate that the correlation is not always perfect, but it doesn't have to be perfect to be statistically significant. If there is no correlation between historical returns and risk, then we should all just buy T-Bills and bonds. IIRC Bernstein considers volatility the major component of risk, as do most folks. Hence the questions regarding willingness to assume various degrees of portfolio percentage losses are included when attempting to determine an individual's risk aversion.

If you like Math, you might find Bernstein's other book, "The Intelligent Asset Allocator" more to your liking. This was actually his first book on the subject, but either bored most folks to tears or was too complex for them to understand, so he wrote "Pillars" to make the subject palatable to the average joe (me ;-).

2old