Michael Petrelli, at Flypaper, writes approvingly of the Ohio State Superintendent’s recommendation to limit testing time to two percent of instructional time, and test preparation to one percent of instructional time.* He seems to suggest that this is reasonable and therefore, by definition, not over-testing.
This strikes me as an argument similar to arguments that spending for testing–or any other program at issue–only represents X% of the budget, as a reason for doing it. It is good to know how much time and money is being spent, and may even be useful to know how it fits into the big picture, but that only gives us a sense of the cost piece of the cost-benefit analysis, not the benefit piece.
So, off the top of my head this afternoon, I would suggest that the use of tests that serve no educational purpose for the students might be, by definition, over-testing. The use of tests with poor reliability might be, by definition, over-testing. The use of tests that are longer than necessary to acquire results adequate for the stated educational purpose might be, by definition, over-testing. And this could be true even if the tests themselves fit within the “reasonable” two percent of instructional time.
*In Iowa (based on a 1080 hour school year), this would amount to 21.6 hours for testing and another 10.8 hours for test preparation, for a total of 32.4 hours per year. [Note: students on a fifteen minute lunch schedule have 45 hours per year for lunch.] The graphic below depicts assessments, many of them standardized, that Iowa students may be expected to take throughout their K-12 careers.