As a Web site designer, I've long been an advocate of JakobNielsen's ideas -- to an extent. Usability is arguably the most important aspect of any design project, and an aspect too often ignored by many software and Web site designers.
Mr. Nielsen, in his book, very aptly points out typical errors and common stumbling blocks of interface design, and presents very convincing arguments and methods for solving these problems. However, strict adherence to Mr. Nielsen's interface design techniques, at the expense of less easily measured human factors, will often result in a sterile and boring product. Both are eminently efficient and usable, but are also wonderful examples of visual blandness -- nearly devoid of the human and aesthetic factors that contributes to a depth of personality and a richness of sensory stimulation.
Although Mr. Nielsen never specifically advocates this, the logical conclusion of his approach is an interface design whose personality and soul have been stripped away in a slavish preference for pure, unencumbered efficiency and usability. Contrary to Mr. Nielsen's examples, the quest for usability should not abrogate the need to avoid ugliness.
For the sake of efficient usability, I wonder if Mr. Nielsen has replaced his impractical, hard-to-maintain backyard lawn with efficient asphalt paving. Or maybe pulled out his expensive, hard-to-clean, dirt collecting, living room carpet and replaced it with an efficient concrete floor. I'm joking of course, but even if Mr. Nielson thinks this way, most do not. Yet, this is the result achieved by many of his user interface examples.
Perhaps on the planet Vulcan where everyone thinks like Mr. Spock, Mr. Nielsen's conclusions and methods might be the eminently rational final word on good interface design. But on Earth the value of his conclusions and usability tests must be weighed against the somewhat hard-to-measure and difficult-to-quantify factors of illogical human personality and perception.
Although Mr. Nielsen's observations, conclusions and suggestions continue to be very valuable in helping to pull interface design towards much needed greater usability and functionality, his mistake seems to be that this is all he sees as being important.