IQ yo-yo: test changes alter retardation diagnoses.

Since average scores on particular IQ tests rise a few points every 3 or 4 years, those tests become obsolete after a couple of decades. In order to reset the average score to 100, harder IQ tests are devised every 15 to 20 years.

Trickier tests have no practical impact on people who score within the normal IQ range of 90 to 110. But so-called renormed IQ tests create a yo-yo effect in the number of mental retardation placements in U.S. schools, a new study finds.

Rates of mental retardation among children appear to bottom out near the end of a particular test's run, followed by a sharp rebound with the introduction of a tougher test, say Tomoe Kanaya, a graduate student at Cornell University, and her colleagues. Scores on the new test then increase over time, pulling many children from just below to just above the store of 70, which stands as the rough cutoff for mental retardation. That trend continues until the next test revision comes along.

As already demonstrated for children with IQ scores in the normal range, kids scoring near 70 lose an average of nearly 6 points when administered a renormed test, Kanaya's team reports in the October American Psychologist.

Mild forms of mental retardation often prove difficult to diagnose. Psychologists look not only for an IQ of slightly less than 70 but also for impaired social and practical skills.

"Our findings show the importance of focusing on children's [real-life] functioning when assessing mental retardation," says psychologist Matthew H. Scullin of West Virginia University in Morgantown. For several years after the introduction of a revised IQ test, he adds, "two children in the same classroom with the same cognitive ability could be diagnosed differently, simply because different tests were used for each child."

In their study, the Cornell scientists analyzed IQ data from 8,944 special education assessments conducted by psychologists in nine school districts across the country. Testing was often repeated and ran from 1989 through 1995.

Among the kids who took the old test and then the version revised in 1991, the proportion scoring just below 70 and recommended for special education tripled. The effect was absent in children who took the same test on two occasions.

The impact of the yo-yo effect on IQ scores extends beyond the classroom, Scullin says. People convicted of murder now avoid the death penalty if they're deemed mentally retarded, but they stand a far better chance of scoring below 70 on a recently revised IQ test than on an older test. The same goes for adults trying to qualify for Social Security payments for a mental disability.

"Talk about high-stakes IQ testing," remarks educational psychologist Frank Gresham of the University of California, Riverside.

The new findings are consistent with those from much smaller studies conducted in various school districts over the past 20 years, Gresham says. During that same period, however, diagnoses of mental retardation have decreased, while designations of learning disabilities have soared. In Gresham's view, this trend largely reflects the stigma attached to labeling a child as mentally retarded.