American 15-year-olds performed above the international average in science, reading, and computational problem solving in the latest global comparison of student performance, but reading scores dropped 14 points from the previous cycle to their lowest level since the United States began participating in the assessment. Mathematics scores were statistically indistinguishable from the international average.
The results come from the 2025 cycle of the Program for International Student Assessment, a comparative study conducted every three years since 2000 and coordinated internationally by the Organization for Economic Co-operation and Development. Ninety-one education systems participated this cycle, including 38 member countries of the international organization and 53 additional countries and subnational systems. About 760,000 students took the test worldwide, making the cohort the largest in the study’s history. Data collection in the United States was completed in the spring of 2025, and the results for the country were released September 8 by the U.S. National Center for Education Statistics.
American students scored 502 points in science — two points higher than in 2022 and 20 points above the international average of 482. In reading, the U.S. average was 490 against an international average of 461, but 14 points below the American score three years earlier. That decline mirrored the international average drop of 14 points in reading. In mathematics, the U.S. average was 463, identical to the international average, and only two points below the country’s 2022 score — a smaller decline than the nine-point international drop. On the new computational problem solving measure the U.S. scored 513, against an international average of 500.
Science was the primary focus of this cycle, with a larger share of assessment items devoted to it and to related questions on students’ background and their schools. Reading and mathematics were minor domains, meaning fewer new items and a heavier reliance on trend questions carried over from earlier cycles so scores could be compared over time. The United States also took part in the first administration of a new innovative domain called Learning in the Digital World, which included computational problem solving using block-based programming tools and concept-map builders.
Two U.S.-specific patterns stood out across all four measured domains. First, the gap between the country’s highest-performing students at the 90th percentile and its lowest-performing students at the 10th percentile was wider than the comparable gap in many other participating systems; the acting commissioner of the National Center for Education Statistics, Matthew Soldner, noted the United States recorded the widest reading-performance gap between its highest- and lowest-scoring students of any participating country. Second, boys outperformed girls in mathematics and in computational problem solving, while girls outperformed boys in reading by about 24 points.
The broader international picture was grim. The international organization reported the lowest average performance ever observed in science, reading, and mathematics across its member countries. Reading scores across those countries fell about 28 points between 2015 and 2025, equivalent to roughly a year and a half of learning, and mathematics scores fell about 22 points over the same period. The share of 15-year-olds classified as low performers across science, math, and reading rose to one in five, up from 16 percent in 2022. Top performers this cycle included education systems in Beijing, Shanghai, Jiangsu and Zhejiang, along with Singapore, Estonia, Japan, Korea, Macao, Chinese Taipei, and the United Kingdom.
The U.S. result therefore reads as two facts held together: American teenagers now sit above the international average in three of the four measured areas because performance elsewhere has dropped faster, and the country’s reading score has still dropped meaningfully on its own terms. The Program for International Student Assessment does not produce state-level results by default, and results for individual schools or students are never reported. The next cycle of the assessment will be held in 2029, with results expected the following year, and the international coordinating body has moved the study to a four-year schedule going forward.
