Fact 1. Our children are tested to an extent that is unprecedented in our history and unparalleled anywhere else in the world. While previous generations of American students have had to sit through tests, never have the tests been given so frequently, and never have they played such a prominent role in schooling. The current situation is also unusual from an international perspective: Few countries use standardized tests for children below high school age—or multiple-choice tests for students of any age.
Fact 2. Noninstructional factors explain most of the variance among test scores when schools or districts are compared. A study of math results on the 1992 National Assessment of Educational Progress found that the combination of four such variables (number of parents living at home, parents' educational background, type of community, and poverty rate) accounted for a whopping 89 percent of the differences in state scores. To the best of my knowledge, all such analyses of state tests have found comparable results, with the numbers varying only slightly as a function of which socioeconomic variables were considered.
Fact 3. Norm-referenced tests were never intended to measure the quality of learning or teaching. The Stanford, Metropolitan, and California Achievement Tests (SAT, MAT, and CAT), as well as the Iowa and Comprehensive Tests of Basic Skills (ITBS and CTBS), are designed so that only about half the test-takers will respond correctly to most items. The main objective of these tests is to rank, not to rate; to spread out the scores, not to gauge the quality of a given student or school.
Fact 4. Standardized-test scores often measure superficial thinking. In a study published in the Journal of Educational Psychology, elementary school students were classified as "actively" engaged in learning if they asked questions of themselves while they read and tried to connect what they were doing to past learning; and as "superficially" engaged if they just copied down answers, guessed a lot, and skipped the hard parts. It turned out that high scores on both the CTBS and the MAT were more likely to be found among students who exhibited the superficial approach to learning. Similar findings have emerged from studies of middle school students (also using the CTBS) and high school students (using the other SAT, the college-admission exam). To be sure, there are plenty of students who think deeply and score well on tests—and plenty of students who do neither. But, as a rule, it appears that standardized-test results are positively correlated with a shallow approach to learning.
...
http://www.alfiekohn.org/teaching/edweek/staiv.htm