In view of the heated discussion of the California High School Exit Examination, it is important to know that, generally, this much we can say: High school exit examinations don't work, and in some cases, they backfire. States don't gather information on the effects of the test because the political risk is too great. Imagine voters' outrage if a study found that a state had spent hundreds of millions on a test that did no good.

Yet indications from various sources are that that is the case – the tests do no good. Reports from Massachusetts find that students who pass the exit exam there are as likely to need remedial work in college as students were before the test existed.

We know from several pieces of research, most particularly from an accidental experiment in California, that the test can backfire. As in many states, a concatenation of fears, pressures and agendas led to the adoption of the California High School Exit Examination, known as CAHSEE.

Students took the CAHSEE annually, starting in their sophomore year. The CAHSEE was first scheduled to hit the class of 2004. Thus, the class of 2005 took the test as sophomores in 2003. However, after the 2003 administration, the California State Board had second thoughts, commissioned a study and – oops – was advised that the courts would likely throw out the results if applied to the class of 2004.

In other states, the study noted, courts had overturned exit exam results when they judged that the test's sanctions went into effect too fast for the students to have been taught what they were being tested on. You can't punish kids for not knowing stuff they haven't been taught, the courts said.

The board delayed the effective date of the CAHSEE until 2006. Thus the class of 2005 first took the test as 10th-graders in 2003 thinking they had to pass it to get a diploma, but by the time they were 11thgraders in 2004, they knew they were home free. The class of 2006 and future classes were under the gun.

Researchers at Stanford looked at the achievement and graduation rates of the classes of 2005, 2006 and 2007. They found that, as 11th-graders, the classes of 2006 and 2007 scored no higher than the class of 2005. The CAHSEE requirement did not improve achievement.

They also discovered that students in the classes of 2006 and 2007 who scored in the bottom 25 percent of the test as 10th graders were substantially less likely to graduate than similar students in the class of 2005. Remember, when the class of 2005 took the test as sophomores in 2003 they thought they had to pass, but later learned they didn't. The mere imposition of the CAHSEE requirement on the classes of 2006 and 2007 caused more students to drop out. This is not, according to exit test advocates, how exit examinations are supposed to work.

The impact was disproportionately large for minorities and for girls. This is important because while it might be thought that minorities suffered from attending lower-quality schools, girls are distributed over all schools. The researchers were also able to rule out that minorities and girls might have been counseled into less-demanding courses that didn't prepare them as well for the test.

Another group of researchers at the University of Minnesota tried to determine if state high school exit examinations made the diploma more meaningful to employers. The answer was a resounding "No." It didn't matter if the exit exam was relatively easy or tough. The Minnesota team concluded, "These examinations must be seen as a colossal waste of education and human resources, harmful to those whose educational attainments are curtailed by failing them and of little use to those who pass them."

Instead of a do-or-die exit exam, the superintendent of public instruction and the State Board of Education might consider the "body of evidence" route in which many aspects of a student's performance are considered, but no one bit of evidence can alone keep a child from earning a diploma.

This actually would take us back to Alfred Binet's approach to constructing IQ tests. All tests in Binet's battery had the same weight. A child who scored high, say, on tests 1, 2, 3, 4 and 5, but low on 6 would get the same score as a child who scored high on tests 2, 3, 4, 5 and 6, but low on 1.

Surely a group of educators could provide California's education policymakers a list of items that should be considered when granting a diploma or not. Given today's fetish for math and science, they might suggest that an A in calculus be weighted more heavily than knowing when to use the subjunctive tense in French, although that would not be my choice. But low performance on one particular item in the list could not prevent a student from walking across the stage.

In any case, we can say without question that making any test an absolute requirement for graduation not only doesn't work as desired, it flops.

Gerald W. Bracey has had a 42-year career in education. His most recent book is "Education Hell: Rhetoric vs. Reality"