Science Fair Project Ideas for High School Students (2026): 60+ Award-Winning Topics

What Makes a Science Fair Project Actually Win

Most science fair projects fail for the same reason: the topic is too broad, the experiment has no control group, and the conclusion is "we found X is interesting." Judges reward the opposite - a narrow question, a clean experimental design, and a result that the student can defend under pointed questioning.

This guide is built from what actually wins at ISEF, Regeneron STS, and state-level fairs in 2024 and 2025, plus what we have seen work for Sir Luther Center students whose science fair projects have grown into peer-reviewed publications.

The format is simple: 60+ ideas across six subject areas, each tagged with difficulty (Beginner, Intermediate, Advanced) and a one-line note on what the strong version of that project looks like.

Biology and Life Sciences

Beginner

1. Effect of music genre on plant growth. Strong version: measure stem height, leaf count, and chlorophyll content (use a $20 SPAD-style meter) across 4+ genres, with silence as a control. n >= 6 per group. 2. Antibacterial properties of household substances. Strong version: zone-of-inhibition assay on E. coli K12 (safe strain) for honey, garlic extract, turmeric, and silver-colloid water. Photograph zones daily. 3. Caffeine and heart rate in Daphnia. Strong version: dose-response curve across 5 caffeine concentrations, n >= 10 per group. Use a microscope and a metronome. 4. Memory and color: which colors are easiest to recall? Strong version: 30+ participants, controlled exposure time, counterbalanced color order.

Intermediate

5. Microplastic content in local water sources. Strong version: filter 5L samples from 4 sources, dye with Nile Red, count under fluorescence microscopy. Report particles/L by polymer type if you can stain-differentiate. 6. Probiotic effects on gut bacteria diversity. Strong version: 16S rRNA sequencing kits from Genohub or similar are now under $100. Pre/post sampling on yourself (with parent consent) or a willing family member. 7. Tardigrade tolerance to UV radiation. Strong version: dose-response across 5 UV exposure durations, survival check at 24h and 72h. Photograph specimens. 8. Plant response to mycorrhizal fungi inoculation. Strong version: same species, same soil, half inoculated with commercial mycorrhizae. Measure root mass, leaf area, and time to flowering.

Advanced

9. CRISPR-Cas9 knockout effects in C. elegans. Strong version: requires lab access (community college or university partnership). Knock out a single gene, document phenotype with imaging across 3 generations. 10. Bioinformatic analysis of public genomic datasets. Strong version: use NCBI or UK Biobank open data to look at one variant across populations. No lab needed - this is one of the most publishable categories for online students. 11. Effect of microbiome composition on anxiety-like behavior in mice. Requires IACUC oversight. Most students partner with a university lab. 12. Detection of plant disease using machine learning on leaf images. Strong version: train a CNN on PlantVillage dataset, then test on photos you collect from local farms. Report accuracy, precision, recall.

Chemistry

Beginner

13. Vitamin C content of fruits over time. Strong version: iodine titration, samples at day 0, 3, 7, 14. Test fresh vs frozen vs canned. 14. Effect of pH on enzyme activity (catalase from potato). Strong version: 5 pH levels (3, 5, 7, 9, 11), measure oxygen produced via water displacement, n >= 5 per condition. 15. Optimal sunscreen SPF: UV light penetration through different brands. Strong version: UV-sensitive beads or photodiode, controlled film thickness, blind labeling. 16. Crystallization rate vs solute concentration. Strong version: time to first crystal at 4+ saturation levels, photograph crystal morphology.

Intermediate

17. Heavy metal contamination in local soil. Strong version: atomic absorption spectroscopy at a partner lab, or use a Hanna multi-meter for accessible metals. Report lead, cadmium, arsenic in mg/kg. 18. Synthesis and characterization of biodegradable plastic from cornstarch. Strong version: tensile strength testing, biodegradation rate in soil over 8 weeks. 19. Electrochemistry: building and optimizing a saltwater battery. Strong version: vary electrode metals (Cu, Zn, Mg, Al), measure voltage and current over 24h, model internal resistance. 20. Photocatalytic degradation of dyes using TiO2. Strong version: methylene blue under UV light, measure absorbance at 664nm every 10 minutes.

Advanced

21. Computational drug docking against a SARS-CoV-2 protein. Strong version: AutoDock Vina, ChEMBL library, free cloud compute. Highly publishable. 22. Synthesis of metal-organic frameworks for CO2 capture. Requires lab access. ZIF-8 is the standard starter. 23. Spectroscopic identification of unknown organic compounds. IR + NMR if you have lab access through a university partnership. 24. Green chemistry: solvent-free synthesis of aspirin. Strong version: compare yield, purity (melting point), and waste vs traditional synthesis.

Physics and Engineering

Beginner

25. Optimal blade angle for a small wind turbine. Strong version: 3D-print blades at 5 angles, fixed wind speed, measure voltage output. n >= 3 trials per angle. 26. Effect of bridge truss design on load capacity. Strong version: build 4 designs from popsicle sticks at fixed weight, load to failure with a bucket of sand and a kitchen scale. 27. Magnus effect on projectile trajectory. Strong version: video-track a spinning ball with Tracker software, compare to theoretical model. 28. Solar cell efficiency vs angle of incidence. Strong version: measure power output (V x I) every 10 degrees from 0 to 90, repeat 3 times.

Intermediate

29. Building and tuning a PID controller for a balance robot. Strong version: Arduino, IMU sensor, document tuning process and step response. 30. Acoustic levitation of small objects using ultrasonic transducers. Strong version: open-source build (TinyLev), document standing wave nodes and maximum mass supported. 31. Aerodynamic drag of car body shapes in a homemade wind tunnel. Strong version: smoke streamlines, force measurement via load cell, n >= 5 trials per shape. 32. Quantum random number generation using a smartphone camera. Strong version: use sensor shot noise, document statistical tests (NIST suite) for randomness.

Advanced

33. Cosmic ray detection using a homemade cloud chamber. Strong version: count tracks/hour over 7 days, correlate with solar activity from public data. 34. Building a low-cost MRI phantom and pulse sequence simulator. Strong version: pair with a university radiology department. 35. Optimizing a maglev train design for energy efficiency. Strong version: 3D-print track, measure power draw at 5 speeds, compare to wheeled control. 36. Machine learning for fault detection in 3D-printed parts. Strong version: train CNN on stress-test images, deploy in real time during prints.

Environmental Science

Beginner

37. Air quality monitoring in different parts of your city. Strong version: PMS5003 sensor (~$25), 10 locations, 3 readings per location at the same time of day. 38. Effect of road salt on freshwater plant growth. Strong version: 5 salinity levels, Elodea or duckweed, measure biomass and chlorophyll over 14 days. 39. Composting efficiency: hot vs cold methods. Strong version: temperature probes, weekly C:N ratio, time to finished compost. 40. Plant species diversity in disturbed vs undisturbed lots. Strong version: 1m x 1m quadrat sampling, Shannon diversity index, n >= 10 quadrats per site.

Intermediate

41. Tracking microplastic contamination through a watershed. Strong version: 6 sites upstream to downstream, monthly sampling, polymer ID by FTIR if available. 42. Carbon sequestration rates of different cover crops. Strong version: above- and below-ground biomass, soil C measurement via loss-on-ignition. 43. Pollinator preference for native vs ornamental plants. Strong version: timed observation counts, 4+ plant species, 3 sites, multiple weeks. 44. Effect of light pollution on insect populations. Strong version: light trap counts at 5 distances from a streetlight, multiple nights, identify to family.

Advanced

45. Remote sensing analysis of deforestation using Sentinel-2 data. Strong version: free Google Earth Engine, NDVI time series for one region over 5 years. Highly publishable. 46. Bioremediation of heavy metals using engineered bacteria. Requires lab partnership. 47. Climate model downscaling for local precipitation prediction. Strong version: use CMIP6 public data, statistical downscaling to your county. Python skills required. 48. eDNA monitoring of fish populations in local streams. Strong version: kit-based sampling, partner with a university for sequencing.

Psychology and Behavioral Science

Beginner

49. Effect of music tempo on test-taking speed and accuracy. Strong version: 3 tempo conditions + silence, counterbalanced order, n >= 30, IRB-style consent forms. 50. Color and emotion: which colors elicit which feelings? Strong version: standardized emotion scale (PANAS), 4 color conditions, n >= 40. 51. Sleep duration and reaction time in teenagers. Strong version: self-report sleep log + a free reaction-time app, n >= 25, within-subject design. 52. Social media use and self-reported mood. Strong version: pre-registered hypothesis, validated mood scale (PHQ-2), 2-week diary study.

Intermediate

53. Effect of exercise type on working memory. Strong version: cardio vs strength vs control, n-back task, n >= 30 with random assignment. 54. Bilingualism and cognitive flexibility (Stroop test). Strong version: matched groups by age and education, standardized Stroop, n >= 40. 55. Mindfulness apps and stress reduction in students. Strong version: 4-week RCT, validated stress scale (PSS-10), n >= 50. 56. Decision-making under time pressure: framing effects. Strong version: classic Tversky framing problems, 4 conditions, n >= 60.

Advanced

57. EEG correlates of focused attention during reading. Requires partnership with a psych or neuroscience lab (Muse headband works for simpler builds). 58. Computational modeling of risk-taking behavior. Strong version: fit prospect-theory parameters to behavioral data using Python. 59. Cross-cultural meta-analysis of growth mindset interventions. Strong version: PRISMA-style review of published studies, effect-size calculations. 60. Implicit bias and hiring decisions: a behavioral experiment. Strong version: IRB-style review, n >= 100, deception protocol with debrief.

Computer Science and Math

All Levels

61. Machine learning for early detection of plant disease from photos (beginner with PyTorch tutorials, advanced if deployed to a phone). 62. Cryptographic analysis of a classic cipher and a modern weakness (intermediate; great for math fair). 63. Predicting election outcomes from social media sentiment (intermediate; uses public Twitter/X academic API alternatives). 64. Building an accessibility tool: real-time captioning for the deaf community (advanced; combines ML, UX, and impact judging).

What Judges Actually Care About

Across ISEF, Regeneron STS, and state fairs, the same five criteria show up:

1. Specific question. "Does X affect Y, in this population, under these conditions?" beats "studying X." 2. Control group. No control = no project. This is the #1 reason strong topics get scored down. 3. Sample size and statistics. n >= 5 is the bare minimum for any quantitative project; n >= 30 for behavioral work. Report means, standard deviations, and an appropriate test. 4. The student can defend the design. Judges will ask, "Why this method? What would you do differently?" If you cannot answer in your own words, the project is not yours. 5. Honest discussion of limitations. Strong projects say "this is what we did not test and why it matters." Weak projects oversell.

Turning a Science Fair Project Into a Peer-Reviewed Publication

This is the question we get most often from students who win at the regional level: "What is next?"

A strong science fair project is roughly 60% of a publishable paper. What is usually missing:

• A formal literature review with 15-30 cited sources, properly framed.
• A standardized methods section that a stranger could replicate.
• Statistical analysis at the level a reviewer expects (not just averages).
• A discussion section that connects results to the existing literature.
• Submission to a peer-reviewed journal that actually accepts high school work - this is a real list, not a free-for-all.

At Sir Luther Center, Harvard student mentors take students through exactly this transition over 8 weeks. The project you brought to a science fair becomes a published paper with a DOI and a citation - a credential that lasts beyond high school.

420+ Sir Luther students have published in peer-reviewed journals - many starting from a project that was originally built for a school science fair.

How to Pick the Right Idea for You

Three quick filters:

1. Subject you already enjoy. A 12-week project on a topic you find boring is misery. Pick something you would read about for fun. 2. Resources you actually have. Be honest about lab access, equipment, and time. A great kitchen-sink project beats a mediocre fantasy lab project. 3. A specific question, not a topic. "Microplastics" is a topic. "Does road runoff increase microplastic concentration in our local creek between site A and site B?" is a project.

Next Steps

If you have an idea you want to develop into a publishable paper, the fastest path is to start with a mentor who has done it before. Book a 15-minute call to talk through your idea, or browse the Research Publication Fellowship for the full 8-week program structure.

For younger students (grades 6-9), the Young Researchers Program is built specifically around taking a first science-fair-style project to a published outcome.