Guide
How to Read Personality Test Results
One personality report gives you a score of 65. Another gives you 65%. A third puts you in the 65th percentile. Have all three told you the same thing?
You cannot tell from the number alone. A score of 65 might be a total from your answers, while the 65th percentile describes how your score compares with other people's scores. Those are different kinds of information. Before deciding whether a result is high, low, or useful, you need to know which kind you have.
Let's read an example report and work out what one score can tell us. You can use the same questions on a real report. You do not need to take a test or share a result here. If you are still deciding which test to take, start with the guide to choosing a personality test.
Start with the report
Example report: invented scores
These numbers were invented for this lesson. No one answered a questionnaire to produce them, so they can help us practice reading a report without claiming to measure anyone's personality.
- Questionnaire
- None. This example uses the Big Five trait names.
- Scale
- 1 to 5. Higher scores mean more of each trait in this example.
- Reference group
- Two invented groups, shown below, for comparing the conscientiousness score.
- Measurement precision
- Unknown. This example includes no precision estimates.
| Trait | Example score |
|---|---|
| Openness (O) | 3.2 |
| Conscientiousness (C) | 3.6 |
| Extraversion (E) | 2.7 |
| Agreeableness (A) | 3.8 |
| Neuroticism (N) | 2.5 |
We will follow the conscientiousness score, 3.6, through each step. Conscientiousness concerns tendencies such as being organized, dependable, and careful about completing tasks. Knowing the trait's name still doesn't tell us what 3.6 means.
1. Find out what produced the score
Start with the name and version of the questionnaire. “Big Five” identifies a broad way of organizing personality traits, not one particular test. Two websites can use the same five headings while asking different questions, scoring them differently, or comparing your results with different groups. The shared headings do not make the scores interchangeable.
Look for the questionnaire's name, its form or version, the language you took it in, and an explanation of the scoring. These details help you find the instructions and research that apply to your result. For help with that search, see how to identify the test. The guide to the Big Five traits explains the traits themselves.
Our report has no questionnaire behind it. We can still use its invented numbers to learn how to read a score, but they cannot tell us about an actual person. With a real report, identifying the questionnaire lets you go further and examine how well its scores measure the traits they name.
2. Check what the score means
In a real report, 3.6 could be a raw mean: the average of the responses after they have been scored according to the questionnaire's instructions. Other reports show a total, convert the score to a different scale, or give its rank in a group. Check the score's definition before comparing it with another number.
On our report's 1–5 scale, 3.6 lies above the midpoint of 3. But the midpoint of a scale is not necessarily the average of people's scores. To know whether 3.6 is above average, we would need scores from a comparison group. So far, we have only the scale.
You could divide 3.6 by 5 and get 72%. The arithmetic works; “72% conscientious” does not. Five is the highest score on this scale. Dividing by it expresses 3.6 as a percentage of that maximum, not as a percentage of someone's personality.
The same problem appears when you compare the rows. Agreeableness is 3.8, which is larger than 3.6. Does that make agreeableness the more unusual result? We cannot tell until we compare each score with other scores on its own trait. A larger score on one trait need not be more unusual than a smaller score on another.
3. Ask: compared with whom?
A percentile rank tells you where a score stands among a particular set of comparison scores, called the reference group. This answers the question the raw value could not: how high or low is this result compared with these other results? Keep 3.6 fixed and see what happens when we change the group.
The same score in two groups
Each group below contains ten invented scores, not norms from a real personality test.
| Group | Scores, lowest to highest | Below 3.6 | Equal to 3.6 |
|---|---|---|---|
| A | 1.1, 1.8, 2.2, 2.8, 3.0, 3.2, 3.6, 4.0, 4.2, 4.8 | 6 | 1 |
| B | 2.9, 3.6, 3.8, 4.0, 4.1, 4.3, 4.4, 4.6, 4.8, 4.9 | 1 | 1 |
Both groups include a value exactly equal to 3.6, so we need a rule for ties. Here we use the midrank rule: count every lower value and half of each equal value, then divide by the group size and multiply by 100. Written as a formula, that is 100 × (number below + 0.5 × number equal) ÷ group size. Other reports may handle ties differently, so use their rules when checking the calculations.
- Group A: 100 × (6 + 0.5 × 1) ÷ 10 = 65th percentile.
- Group B: 100 × (1 + 0.5 × 1) ÷ 10 = 15th percentile.
The score is still 3.6. In Group A, six values are below it; in Group B, only one is. That is why the percentile falls when we switch groups. We changed the comparison, not the score or the person it might describe.
A real report needs the same context. Website volunteers, students at one institution, and a sample designed to represent a broader population can have different score patterns.
Here is a real case. The BFI-2 is a Big Five test scored on the same 1–5 scale as our example. In Christopher Soto and Oliver John’s research on the BFI-2, 1,000 adults who took it on a personality website averaged 3.92 on open-mindedness, while university students averaged 3.66. So the same open-mindedness score would tend to rank higher among the students. A comparison with one group does not automatically describe your position in another.
Find out who was in the reference group and how they were chosen. The guide to test norms explains these choices and shows how the same two percentiles change when a report counts ties differently.
4. Keep the number when the report adds a label
A report may add a label such as “upper range” to help you read a result quickly. That label tells you which category contains the score, but it leaves out how far the score is from the category boundary or from another score.
Imagine a separate, invented 0–100 scale. Its reporting rule labels values below 50 “lower side” and values of 50 or more “upper side.” Scores of 49 and 51 receive different labels while remaining only two points apart. Scores of 51 and 90 share a label while sitting 39 points apart. If you saw only the labels, you could easily mistake the second pair for the more similar one.
Keeping the numbers lets you check that impression: the differently labeled pair is closer on this scale. The category exercise in the MBTI guide gives you another way to practice the distinction. This invented rule illustrates what a category can hide; it is not the official MBTI scoring system.
Researchers have documented both parts of the problem. Methodologists Robert MacCallum, Shaobo Zhang, Kristopher Preacher, and Derek Rucker showed what happens when you split a continuous score into two groups. People on opposite sides of the cut, like 49 and 51, are treated as different. People on the same side, like 51 and 90, are treated as equal, whatever the actual distances. The split also makes the measurement less reliable, not more precise.
Labels also change most easily near a boundary. The publisher of the official MBTI reports that about half of people get the same four-letter type when they take it again. When one preference changes, it is usually the one that was least clear, where the person’s answers leaned least strongly to either side.
5. Ask whether a small difference means anything
Suppose someone takes a real test twice and receives 3.6, then 3.7. The scores differ by 0.1. Before treating that as a change in personality, you need to know how much the test's scores vary under the relevant testing conditions. A test that gives slightly different scores on different occasions can produce a small difference even without a meaningful change in the trait.
Real tests do vary this way. When Soto and John developed the BFI-2, 110 students took it a second time about eight weeks later, and their two conscientiousness scores correlated .83. That is high agreement: people who scored higher the first time usually scored higher again. It does not mean each person received the same number twice. Timo Gnambs’s meta-analysis of 74 samples found the same pattern across many Big Five questionnaires. For retests up to two months apart, the typical correlation was about .82, and occasion-specific influences such as mood accounted for about a tenth of the variation in scores.
This is a question about measurement precision. One useful measure is the standard error of measurement, which estimates the typical size of random measurement error under specified conditions. Researchers can use it to calculate an interval (a range around a score), but only if the error estimate and the calculation suit the test and the testing conditions. Printing another decimal place does not supply this evidence. The 2014 Standards for Educational and Psychological Testing explain score precision and intervals in chapter 2; the section “Standard Errors of Measurement” begins on printed page 39.
Our fictional report has no such evidence, so we cannot draw an interval around 3.6 or decide how much a second score would need to differ before the change mattered. With a real report, look for the provider's precision estimates and the testing conditions they cover. The guide to reliability explains how researchers examine score consistency.
6. Separate a description from a decision
Now imagine a real report with good evidence for its description: you tend to prefer preparation and order. That could help you recognize a pattern across situations. Perhaps you enjoy knowing what comes next and dislike being rushed into an unfamiliar activity. The description gives you something useful to consider when choosing how to approach your day.
But “you prefer preparation” and “you should always use a detailed daily schedule” are different claims. The first describes a tendency. The second says that one particular activity will help you. A detailed schedule might help on a day with several appointments and become frustrating on a day filled with unpredictable interruptions.
To choose between those possibilities, you need information about the activity and the circumstances, as well as the person. Evidence that supports a description does not automatically support every recommendation attached to it. The testing Standards, chapter 1, printed page 11, make this distinction by tying validity to the particular interpretation and use of a score.
Make the next question specific: “Would writing down my first task the night before help me get started?” Try it, then notice whether it helps you begin and whether the task still makes sense when morning comes. You could also compare it with choosing a task over breakfast. The Behavior Menu helps you consider different actions toward a goal. Your result can suggest options worth trying; your experience helps you choose among them.
Try reading these results
Return to the problem at the start: similar-looking numbers can mean different things. Imagine that three reports contain the following statements. What can you learn from each one, and what is still missing?
- “Raw mean: 3.6 on the 1–5 response scale.”
- “65th percentile compared with the group described in the report, using the calculation explained there.”
- “Conscientiousness: 65%.” No further definition appears.
Read the answer
The first tells you the average of the scored responses on the 1–5 scale. You still do not know how common that result is. The second tells you where the score stands in that group, using the report's calculation; it does not give the raw score. The third does not define its percentage. It could be a score converted to a percentage, a percentile rank, or something else. You would need the scoring explanation before you could interpret it.
None tells you what percentage of a person is made up of conscientiousness. Each defined score answers a more specific question. Once you know that question, you can use the result without asking it to choose a routine, career, or relationship for you.
Say what the score means
When you return to a real report, try describing the result in one sentence. Include the questionnaire, what kind of score it gives, and any reference group. Then ask what the result helps you understand. If you cannot tell whether “65” is a total or a rank, you have found a specific question to resolve before making a decision.
For our example: “The fictional report gives a score of 3.6 on a 1–5 scale; using our midrank rule, it is at the 65th percentile of invented Group A and the 15th percentile of invented Group B.” Keeping the scale, rule, and groups with the number makes the result understandable. Removing them leaves you with 3.6 and a guess.
Use How to Evaluate Personality Tests to examine the evidence behind a real report. For practice in a group, use the printable classroom exercises and answer key.