PUBLISHED: June 17, 2006 | By DAVID M. HERSZENHORN | New York Times

More than 8,900 New York City fifth graders are in danger of being held back because of failing scores on annual reading and math tests, city education officials said yesterday. That contrasts sharply with last year when a big rise in scores led Mayor Michael R. Bloomberg to say that his effort to end social promotion had improved student achievement.

City third graders posted better results with the number in danger of repeating the grade declining for a third year in a row. And there were gains among seventh graders, who were subject to the mayor's promotion policies for the first time this year.

Students who fail to make the cut on the annual exams are urged to attend summer school. They must retake the exams in August, and if they fail again, they must repeat the grade.

While the results were mixed, it was the fifth grade numbers that drew attention because the most stellar increases on city exams last year were in that grade.

In interviews, city education officials offered many explanations for the poorer fifth grade results, which saw the number failing to earn promotion rise to 8,921, or 15.2 percent of students subject to the mayor's promotion rules, from 5,450 students, or 8.6 percent last year.

The officials said that new state tests were used this year in place of citywide exams and that the tests were given in the winter instead of the spring. The changes, they said, made it unfair to compare year-to-year results.

Officials also cautioned that the full test results would not be issued by New York State until August and that the promotion decisions were based on a preliminary analysis of scores. But some analysts said the latest data lent credence to skeptics' assertions that last year's steep jump in test scores of fourth and fifth graders was too good to be true.

"There is some evidence here to suggest that the grade four test results last year were inflated," said Robert Tobias, the former director of testing for the city schools and now a professor at New York University's Steinhardt School of Education.

Lori Mei, the city's current testing director, disputed any notion that the new results cast any doubt on last year's success. "I am saying categorically no," she said.

Unlike the last two years in which Mr. Bloomberg and Schools Chancellor Joel I. Klein held triumphant news conferences to tout test results, this year's promotion data was e-mailed to reporters by Mr. Klein's press office yesterday. Officials said a news conference would be held when the full state test results are released in August.

City Hall referred questions to the Education Department. Mr. Klein left for the weekend yesterday afternoon without issuing any comment. Later, through his press secretary, he issued a statement thanking the State Education Department for its help in providing preliminary results.

"We look forward to receiving the official scores from the state later this year," Mr. Klein said.

The city's preliminary analysis of scores was needed because the state this year assumed responsibility for annual reading and math tests of all students in grades three to eight to comply with the federal No Child Left Behind law. In previous years, the state handled the exams in grades four and eight while the city gave the tests in grades three, five, six and seven.

Last year, the proportion of city fourth graders scoring at or above grade level in reading increased 9.9 percentage points. But that gain — by students who are this year's fifth graders — was outpaced by all of the state's large districts except Buffalo, prompting critics to ask whether the test was too easy. State officials insisted it was not.

"Given the fact that we had this meteoric rise in fourth grade last year, given the fact that it looked suspicious," Mr. Tobias said, "I was looking forward to this year's fifth-grade results to get a sense if those fourth-grade increases were real."

Of yesterday's results, Mr. Tobias said, "This is exactly what I would expect to see if there were problems with the fourth-grade test last year."

Ms. Mei suggested other reasons for the decline in fifth-grade performance, particularly the earlier administration of this year's test. Later exams had allowed many struggling students to benefit from three months more of Saturday classes for remediation and test preparation.

"We have charts that show a real connection between Saturday sessions children attended and how they performed," Ms. Mei said. "Given the fact that the test was given in January those at-risk kids attended many fewer sessions."

In third grade, the number of students failing to earn promotion declined to 5,237 or 9.2 percent from 8,244 or 13.4 percent. In seventh grade, the number failing to meet promotion criteria dropped to 3,552 or 5.3 percent from 7,329 or 11.2 percent.

But the seventh-grade data did not reflect math scores, which historically account for the majority of failures in seventh grade. The city did not use the math scores this year because of recent curriculum changes.

The city faced many logistical headaches this year as the state took full control of testing.

Because the state developed all new exams it had warned that results would not be available until late summer. The city, however, was eager to have test results to make decisions about whether students should go to summer school under the mayor's promotion rules.

At first, city officials said students would have to take two sets of exams this year, one for the state and the other for the city. But after an outcry by parents, the city and state reached a deal allowing city officials to tabulate preliminary results for use in promotion decisions.

Depending on their scores, students place in one of four performance levels, with Level 1 being far below standards and Level 3 being at or above grade level. In previous years, the city's cutoff for promotion was Level 2 — still below grade but demonstrating some mastery of skills.

It is unclear if the city's decisions about who should go to summer school will precisely match the state's criteria for Level 1.

Critics of standardized testing seized on all of these issues to yet again condemn the city's use of the exams. "It goes to prove one more time that these tests are really a sham," said Jane Hirschmann, of the Time Out From High-Stakes Testing, a leading anti-test group, accusing the city of making decisions in a way that no test expert "would validate or find reliable."