Confidence Intervals and Standard Error of Measurement on Achievement Tests

What a Confidence Interval Actually Tells You

Your child's achievement test report shows a standard score — say, 87 on Basic Reading — and next to it, in parentheses, something like (82-92) or (83-91). That range is the confidence interval, and it is telling you something important: the score your child received is not a perfect measurement. If your child took the same test again tomorrow, under identical conditions, their score would likely fall somewhere within that range.

No educational test measures ability with perfect precision. Every score includes some amount of measurement error — random variation caused by the child's alertness that day, the specific items that happened to appear, whether they guessed correctly on borderline items, and dozens of other small factors. The confidence interval quantifies that uncertainty.

A 95% confidence interval means there is a 95% probability that your child's true score falls within the stated range. A 90% confidence interval (which some reports use) means a 90% probability. The wider the interval, the less precise the measurement. The narrower, the more precise.

Why It Matters for Eligibility

Here is where confidence intervals become a practical advocacy tool rather than a statistics lesson. Suppose your child scores 87 on Basic Reading, with a 95% confidence interval of 82-92. A team may emphasize a score such as 85, but IDEA does not set a single standard-score cutoff for eligibility.

If a team points to the score of 87 as above a numerical line, the confidence interval shows the measurement uncertainty around that estimate. Ask how the interval and the rest of the evaluation data were considered; the interval itself does not determine eligibility.

The same measurement uncertainty applies when a point estimate falls below a number the team is emphasizing. A score and its interval should be interpreted with the rest of the evaluation data, not treated as an eligibility decision by themselves.

Standard Error of Measurement: Where the Interval Comes From

The Standard Error of Measurement (SEM) is the technical value used to calculate the confidence interval. It varies by subtest and by age or grade. Some subtests have smaller SEMs (meaning more precise measurement), while others have larger SEMs (meaning more uncertainty).

The formula is straightforward: the confidence interval is the standard score plus or minus a multiple of the SEM. For a 95% confidence interval, it is roughly ±2 SEMs. For a 90% confidence interval, it is roughly ±1.65 SEMs.

You do not need to calculate this yourself — the report should show the confidence interval directly. But understanding that different subtests have different SEMs explains why some intervals are narrow (±3 points) and others are wide (±7 points). Subtests with more items and more consistent measurement produce narrower intervals. Shorter subtests or those measuring more variable skills produce wider intervals.

How to Use Confidence Intervals at a Meeting

When reviewing your child's scores at an eligibility or IEP meeting, apply these principles:

Do not treat a single score as a fixed truth. A score of 87 is not necessarily meaningfully different from a score of 84 or 90 — their confidence intervals may overlap. If the team draws a hard line at a specific number ("Your child scored 87, the cutoff is 85"), ask how it considered measurement error and the rest of the evaluation; IDEA does not establish a single score cutoff.

Compare confidence intervals, not point estimates. Non-overlapping confidence intervals can support a difference between two scores, but overlapping intervals do not prove that the scores are the same. If Letter-Word Identification is 97 (92-102) and Word Attack is 82 (77-87), those intervals do not overlap. If Letter-Word Identification is 95 (90-100) and Passage Comprehension is 89 (84-94), the intervals overlap; ask the evaluator whether the publisher's score-comparison analysis finds a statistically significant or unusual difference.

Ask whether the evaluator used confidence intervals in their analysis. The evaluator who administered the test should have considered measurement error when interpreting scores. If the report presents only point estimates without confidence intervals, ask for them — they are available in the scoring software.

When a Score Sits Right on a Boundary

Confidence intervals are especially useful when a score falls near a classification boundary or a number the team is emphasizing. A standard score of 90 sits at the boundary between "Average" (90-109) and "Low Average" (80-89). With a confidence interval of 85-95, the child's true score could fall in either classification range.

When this happens, the classification label — "Average" or "Low Average" — is less informative than the confidence interval itself. Ask the team to discuss the range of possible true scores rather than anchoring to the label that the point estimate happens to fall into.

The same logic applies to composite versus subtest comparisons. If a composite score is 92 (87-97) and a subtest is 80 (75-85), the question is not just "is 92 higher than 80?" — it is "do these confidence intervals overlap?" In this case they barely do, suggesting the gap is meaningful. A composite of 92 (87-97) and a subtest of 85 (80-90) have overlapping intervals — the apparent gap may be measurement noise.

The Bottom Line

Every score on your child's report is an estimate, not a fact. The confidence interval tells you how good that estimate is. When eligibility decisions hinge on whether a score is above or below a cutoff, the confidence interval is the tool that prevents a few points of measurement error from determining your child's access to services.

The Achievement Test Decoder explains confidence intervals and SEM for each major test battery, with worked examples showing how to use them in eligibility arguments.