Point Biserial Correlation measures the relationship between getting a single test question right or wrong and a student’s overall performance on the entire test. It helps you figure out: “Is this question working well to identify my stronger vs. weaker students?”
Think of it this way: You have a 20-question test. For question #15, you want to know: “Do students who get this question right tend to do well on the whole test? And do students who get it wrong tend to struggle overall?” Point biserial correlation gives you that answer.
What the Numbers Mean:
- Ranges from -1 to +1
- Positive numbers (0.2 to 0.8) = Good! Students who get this question right tend to do well overall
- Numbers near 0 = Problematic – Getting this question right or wrong doesn’t relate to overall performance
- Negative numbers = Big problem! Strong students are getting this question wrong, weak students are getting it right
Classroom Example: Question: “What is 2 + 2?”
- High positive correlation (0.6): Students who get this right generally do well on the whole test
- Near zero (0.1): Whether students get this right seems random compared to their overall performance
- Negative (-0.3): Your best students are getting this wrong—maybe it’s poorly written or has a trick
When You Might See This:
- Standardized test reports analyzing individual questions
- District assessment data
- When reviewing why certain test questions didn’t work well
- Education courses or professional development on test analysis
What It Tells You About Your Questions:
- Good questions have moderate positive correlations (0.3-0.7)
- Questions that are too easy (everyone gets right) show low correlation
- Questions that are too hard (everyone gets wrong) show low correlation
- Confusing or poorly written questions often show negative correlation
Quick Takeaway: If a question has a low or negative point biserial correlation, consider revising or removing it—it’s not helping you identify which students understand the material.