Eight Weeks of Scratch Coding Failed to Boost Eighth Graders’ Problem-Solving Scores

An eight-week experiment in a rural Turkish middle school has found that block-based coding with Scratch did not measurably improve students’ mathematical problem-solving skills, offering a sobering counterpoint to the widespread enthusiasm for coding as a shortcut to sharper thinking. The study, published open access in Discover Education, tracked eight female eighth-grade students at a vocational religious middle school in Central Anatolia through a pretest, a sustained program of Scratch activities built around mathematical scenarios, and a posttest. When the researchers ran a paired-sample t-test on the scores, the difference between the two tests fell short of statistical significance, with t(7) = −1.063 and p = 0.323.

The research team, Sevim Sevgi and Elif Çelik, set out to test a popular assumption: that programming, when woven into mathematics lessons, cultivates the ability to break down complex problems and solve them systematically. Scratch, the block-based environment developed for children aged 8 to 16, is among the most widely used coding tools at the K-12 level, and prior work has suggested it can help students grasp mathematical concepts more concretely. Yet the literature the authors reviewed also revealed gaps. Very few studies using Scratch involve actual programming, fewer still take place outside computer science and technology classes, and there is lingering uncertainty about how computational thinking should be integrated into other subjects and how its development should be evaluated.

To frame the intervention, the researchers leaned on Polya’s classic four-step model of problem-solving: understanding the problem, developing a plan, carrying out the plan, and reviewing the result. This framework, published nearly eighty years ago, remains a global touchstone for thinking about how problems are solved, and earlier researchers had argued it maps reasonably well onto middle school coding instruction. In the adapted version, students first clarify what a problem asks, then devise an algorithm without writing code, translate that algorithm step by step into Scratch blocks, and finally test and debug their programs against the given scenarios. The authors caution, however, that not every coding problem demands the full sequence; some merely require applying a known algorithm or recognizing a familiar code pattern.

The study’s design was deliberately modest. The researchers employed a one-group pretest–posttest model, which they explicitly describe as a weak experimental design because it involves neither random assignment nor a matched control group. The participants were all eight female eighth-graders at the school, reached through stratified purposive sampling that covered roughly twenty percent of the eighth-grade population in the neighborhood. The authors acknowledge that the rural setting left them little choice: no other nearby schools had computers suitable for programming, only one class was available, and arranging a control group proved impossible. The study was approved by the Institutional Review Board of Erciyes University and conducted in accordance with the Declaration of Helsinki, with informed consent obtained from the students’ parents.

The measurement instrument was a twenty-problem test adapted from the Turkish Ministry of National Education’s standardized eighth-grade mathematics materials, chosen for their high validity and reliability and their uniform administration across the country. Students took the test before the coding program began and again after eight weeks of Scratch activities. The activities were sequenced by the complexity of the code blocks involved rather than by curriculum order, and students were given ready-made blocks for variable assignment and conditional statements while designing the remaining parts themselves. After the coding sessions, the researchers interviewed the students to gather feedback on the experience.

The raw numbers told a story of small, uneven movement. The group’s mean score rose from 8.5 on the pretest to 9.75 on the posttest, on a scale where the minimum and maximum possible scores were 0 and 20. No student hit either extreme on either test. Pretest scores ranged from 3 to 16, while posttest scores ranged from 4 to 18. On the pretest, three students scored between 0 and 5, two between 6 and 10, and three between 11 and 15; on the posttest, one student climbed into the 16-to-20 band. Two students, however, actually scored lower on the posttest, and the researchers note that both had attended the Scratch activities irregularly and had weaker overall achievement.

A closer look at individual problem types revealed a telling pattern. Students performed well when questions asked directly for the greatest common divisor or least common multiple, but struggled when those concepts appeared as intermediate steps in multi-step solutions. On the single prime number problem, the share of correct answers actually fell from 87.5 percent to 75 percent, though one student who had missed it initially solved it after the intervention, correctly rejecting 87 as prime. Prime factorization problems improved modestly, with more than half of students answering correctly on three of five such problems after the program, up from two of five before. The students who failed to improve on these items were largely the ones who had skipped the corresponding coding activities, hinting that engagement, not the tool alone, may shape outcomes.

The qualitative observations were arguably more revealing than the statistics. When the researchers analyzed the students’ coding behavior through the lens of Polya’s model, they found that most participants fixated on visuals during the problem-definition phase and largely neglected planning. Most favored top-down linear designs over branched or independent algorithms, and many successful-looking projects relied on copying and remixing existing code rather than genuine decomposition or abstraction. The authors warn that assuming an impressive or visually appealing Scratch project reflects systematic computational thinking is erroneous; collating and remixing code is a coding activity, but it does not by itself cultivate critical thinking, abstraction, or problem decomposition. Notably, the one student who developed independent algorithms went on to gain admission to a selective Science High School.

The authors attribute the null result to several converging factors: the students’ complete lack of prior coding experience, the short eight-week window, the cognitive complexity of programming itself, and the small sample size. Students handled simple Scratch operations, such as selecting sprites and adding sound effects, with ease, but found operator, control, and event blocks genuinely difficult, echoing earlier findings that activities demanding precise mathematical calculations can become an unintended barrier for younger learners. The results align with prior six-week studies of sixth graders that likewise found no significant effect of Scratch on problem-solving, and with research showing no relationship between computational thinking and academic performance outside collaborative learning contexts.

The study’s limitations are considerable and the authors state them plainly. A single school, eight participants, a weak experimental design, and a cross-sectional structure rule out causal inference and limit generalizability. The researchers also note that computational thinking lacks a consensus definition and that current assessment methods contain unreliable indicators, raising the possibility that other components of the skill went unmeasured. Their recommendations are correspondingly measured: longer interventions, experimental and control designs, larger and more diverse samples, expert-reviewed test questions, and a strategic effort to integrate algorithmic and computational thinking into primary and middle school curricula while rigorously evaluating the consequences. For now, the message for educators is that a few weeks of Scratch is unlikely to transform problem-solving on its own, and that the deeper work of curriculum integration remains unfinished.

Subject of Research: The effect of Scratch-based coding activities on the mathematical problem-solving skills of eighth-grade students

Article Title: The effect of scratch use on problem-solving skills

Article References: Sevgi, S., & Çelik, E. (2026). The effect of scratch use on problem-solving skills. Discover Education, 5(1), Article 1214. https://doi.org/10.1007/s44217-026-01532-8

Image Credits: AI Generated

DOI: 10.1007/s44217-026-01532-8

Keywords: Scratch, computational thinking, problem-solving, block-based coding, mathematics education, middle school students, pretest-posttest, Turkey, coding education, Polya, null result, educational research