What role should student evaluations play in assessing professor performance?

NYT blogger Stanley Fish (former Duke English professor and very well known academic) has some not-so-kind words for student evaluations. Fish's main point is that the end of the semester is rarely the best vantage point for determining what you have gained from a particular course and its professor.  Money quote:

Students tend to like everything neatly laid out; they want to know exactly where they are; they don’t welcome the introduction of multiple perspectives, especially when no master perspective reconciles them; they want the answers.

But sometimes (although not always) effective teaching involves the deliberate inducing of confusion, the withholding of clarity, the refusal to provide answers; sometimes a class or an entire semester is spent being taken down various garden paths leading to dead ends that require inquiry to begin all over again, with the same discombobulating result; sometimes your expectations have been systematically disappointed. And sometimes that disappointment, while extremely annoying at the moment, is the sign that you’ve just been the beneficiary of a great course, although you may not realize it for decades.

Fish is alarmed to see some states using student evals as a component of performance-based pay for faculty.  His fear is that faculty who provide high entertainment value will eventually crowd out those who provide solid learning and development.

My own take as associate dean who regularly sees student evaluations: It provides useful information about classroom performance (especially at the tails of the distribution) but it should not be the only metric.  Faculty also need to be involved, especially for evaluating whether a course is being taught at the proper level of rigor.  Also alumni can provide a perspective that current students cannot bring.