Over the past 38 years, American kids have been getting better at reading and math at age 9, slightly better at age 13, and not any better at age 17.

In other words, there has been considerable work at the lower grades, and by the time the kid's ready to hit the ejection buttion, all that work is wiped out. At least, that's what the data seem to say.

The data is from the federally funded National Assessment of Educational Progress (NAEP). The NAEP people visited the Seattle Times last week, talked about an hour and left us with a copy of their report, "NAEP 2008 Trends in Academic Progress." Of all the things in it, this discrepancy between the younger grades and high school stood out.

Darvin Winick, chairman of the National Assessment Governing Board, said that nationwide, “high school students are taking more math—and it is rather a marked increase.” At age 13, 30 percent of students nationwide are taking algebra in 2008. In 1990 it was 15 percent. At age 17, 19 percent were taking calculus or pre-calculus in 2008; in 1990 it was 8 percent. And yet, said Winick, for 17 year olds “math performance on the assessments has not gone up.”

Explain that, keeping in mind that it has happened over 38 years. What is it? Video games? Hormones? The reform math?

My last week’s column was about the “reform math,” which I didn’t think was very good. But if reform math is lowering scores (and I can’t prove it is) why would it hurt the 17-year-olds more than the nine-year-olds? And the patern is not just in math, but also in reading.

Some of NAEP's other conclusions:

--The “achievement gap” between America’s white, black and Hispanic students in math and reading has narrowed since 1971, but still is fairly wide and shows no significant change since 2004.

--White and Asian Americans tend to do about as well as kids the same age in European and East Asian countries, respectively. Among large ethnic groups (as Americans define them), Asian Americans are the highest-scoring in math.

--Girls and boys do the same, on average, in math at age 9, but boys pull a little bit ahead by age 13 and a bit more by age 17. This gap is not as wide as the racial and ethnic gaps, but it is persistent. With reading, the gap is the other way: Girls do better on average, than boys and have decade after decade. The gap at age 17 was virtually the same in 2008 as it was in 1971.

--The more education the parents had, the higher, on average, their kids score in math. This has remained true for 30 years, though the gap in average math scores between kids of college graduates and kids of high-school dropouts has narrowed somwhat.

--The charge that high-stakes tests are causing teachers to neglect the kids at the bottom and the top to concentrate on kids right near the cut-off line is not shown in the data. Over the years the kids at the bottom have been doing better in math and the kids at the top about the same as they always did.

--Average math scores are higher in Catholic schools than in public schools at all ages and over time. At age 17 in 2008, the Catholic schools’ scores were an average 4 percent higher. Thirty years ago they were 3 percent higher.

--There is not enough data to evaluate charter schools (which are not permitted in Washington), and Winick said charter schools are so different from each other that a group evaluation doesn’t mean much anyway. In his home state, Texas, he says, achievement scores for charter schools are “trimodal,” meaning schools tend to fall into three clusters depending on the type of school they are. Academic charters score higher than the average of public schools. Save-the-dropouts charters score lower than the average of public schools. And what Winick calls “social charters,” meaning charters founded on a principle other than academics, “do about as well as the public schools.”