Tablet Drawing Kinematics and Developmental Differences

ISEF Category: Behavioral and Social Sciences

Subcategory: Development  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A simple drawing can hold more data than you think. Every pause, curve, and sudden change in speed leaves a trail on a tablet. That trail can act like a motion fingerprint. You can study whether those fingerprints differ in kids with developmental coordination differences.

What Is It?

This project looks at how a child moves while drawing on a tablet. Instead of judging only the final picture, you study the path the stylus takes across the screen. Think of it like comparing two songs, one by the final melody, and one by the timing between notes.

Kinematic features are just movement measurements. Jerk means how quickly movement changes, and pause duration means how long the stylus stays still between strokes. An interpretable classifier is a model that tries to sort patterns in a way you can explain. You can point to the features that mattered, instead of getting a mystery score.

Why This Is a Good Topic

This is a strong science fair topic because you can test a real question with measurable data. The same tablet task can produce many signals, so you can compare features, models, and tasks without needing expensive gear. It connects to child development, motor control, and digital assessment. A student can learn how to collect cleaner data, build features, and think about fairness, age matching, and model interpretation.

Research Questions

• How does stroke jerk differ between children who show more and less coordination difficulty?
• What is the effect of drawing task complexity on pause duration and stroke smoothness?
• Does using a stylus instead of a finger change the separation between groups?
• To what extent do simple kinematic features predict group membership better than chance?
• Which feature set, jerk, pause duration, or stroke count, gives the clearest interpretable model?
• How does age matching change the classifier's accuracy and false positive rate?

Basic Materials

• Tablet or iPad with a stylus and a drawing app that exports stroke data.
• Laptop for data export, cleaning, and analysis.
• Quiet room with a stable chair and desk.
• Simple drawing prompts, such as copied shapes or line tracing sheets.
• Parent consent form and child assent form.
• Spreadsheet software for organizing participant data.

Advanced Materials

• Research-grade tablet digitizer or graphics tablet with timestamped stroke export.
• Stylus that records pressure, tilt, or contact data.
• Python environment with scikit-learn, pandas, and numpy.
• R or JASP for statistical testing and effect size estimates.
• Video camera for coding posture, hesitation, or hand-off events.
• Secure data storage with de-identification workflow.

Software & Tools

• Python: Cleans stroke data, extracts features, and trains interpretable models.
• Jupyter Notebook: Lets you document analysis steps and show results clearly.
• scikit-learn: Builds classifiers and checks feature importance.
• RStudio: Runs statistical tests and plots group comparisons.
• JASP: Gives a point-and-click way to test group differences and effect sizes.

Experiment Steps

1. Define one drawing task that every participant can complete in the same way.
2. Choose the kinematic features you will extract, such as jerk, pause duration, and stroke count.
3. Plan a comparison group strategy that keeps age, handedness, and task difficulty as close as possible.
4. Build a clean feature table before you split data into training and testing sets.
5. Select an interpretable model and decide how you will explain its decisions.
6. Predefine how you will handle missing strokes, outliers, and repeated attempts.

Common Pitfalls

• Recording on different tablet sizes or at different angles, which changes stroke timing and jerk even when the child behaves the same.
• Mixing free drawing and copy tasks in one model, which can teach the classifier task style instead of developmental differences.
• Using too few children in each age band, which makes the results noisy and the feature rankings unstable.
• Letting the app smooth strokes automatically, which can erase pause patterns and flatten sharp changes in motion.
• Training and testing on strokes from the same child, which makes accuracy look better than it really is.

What Makes This Competitive

A stronger version of this project does more than report accuracy. It tests several task types, checks whether the same features still work after age matching, and shows which variables really drive the result. If you validate on a new group and report false positives, calibration, and feature importance, the work looks much stronger. Clear interpretation matters more than a flashy model.

Project Variations

• Use copy-tracing instead of free drawing to see whether task structure changes the signal.
• Compare finger input with stylus input to test whether device choice changes feature stability.
• Swap group prediction for age prediction to see whether the same features track normal development across grades.

Learn More

• PubMed: Search for review articles on developmental coordination disorder, handwriting kinematics, and child motor control.
• PubMed Central: Read open-access studies on tablet-based drawing analysis and developmental motor patterns.
• NIH NINDS: Find plain-language background on developmental coordination disorder and related motor signs.
• CDC Developmental Milestones: Check age-expected fine motor milestones for context when you frame your sample groups.
• Frontiers in Psychology: Search open-access papers on handwriting, motor development, and digital assessment.