Ay yi yi, so many questions to consider. I'll probably be answering various of these piecemeal over the course of the day, so apologies in advance for filling up the thread with multiple replies.
I never really understood the "teach to the test" criticism. What is the alternative to "teaching to the test"? Teaching a subject in such a way that students can't demonstrate proficiency?
A well-designed test (SAT Math is a good example) makes it so that teaching to the test is equivalent to teaching the subject. You cannot do well on the SAT Math without actually being good at math. In this respect NCLB gets it right -- you shouldn't be able to mask your terrible teaching with unstandardized, unrigorous, and arbitrarily graded tests of your own design. If your students can't demonstrate basic proficiency you are doing something wrong.
From a Bayesian perspective, if your students fail standardized tests, it's a lot more likely that you're a crap teacher, rather than you're some genius teacher who somehow still can't get his kids to demonstrate basic proficiency.
FTL partly covered this, but I'll give a slightly different wording at least. There two major assumptions in what you've said. The first is that any given standardized test is a good metric of what students know. This is demonstrably untrue in many, though not all, cases. The second is that any given standardized test is a good metric of what students
should know. While this delves into more philosophical problems about what we should be teaching students (and some of my opinions are outlined above), you'll find a lot of agreement among teachers, especially science teachers, that many state standardized tests don't do a great job with this either.
As I said, the failing is in testing knowledge rather than application or analysis. Mathematics tests usually get around this problem because they are, inherently, application tests; there is very little to "know" about math. (Once you know what positive integers are, and what the four basic operations are, everything further is an application of those to more complex situations.) Yet even at that, the students are not often required to show their work; instead, we leave it as a multiple-choice question. Standardized science tests are often little more than knowledge tests, however, and this is a huge failing. Consider which of the following two questions more usefully demonstrates mastery of physics:
(1) Determine the kinetic energy of a baseball with a mass of 95 g travelling at a velocity of 35 m/s.
(2) A baseball is pitched, hit by the batter, and becomes a pop-fly [insert graphic showing the path for someone not familiar with the sport]. Describe the energy changes that happen to the baseball as it moves along this path from pitcher to batter to fielder. Where is energy gained, and where does it come from? Where is energy lost, and where does it go?
Now, which of those is going to show up on a standardized test?
The "teaching to the test" criticism isn't so much that the tests exist, though countries with better education systems often don't have the same kinds of standardized tests (Finland, for instance, which was mentioned above). The problem is that the tests suck, and the nature of the testing system means they will continue to suck. And the result is that the curriculum... well, sucks.