What a BFI-2 Score Can Tell You Without Official Norms

What does ‘no official norms’ actually mean?

“No official norms” has a narrow meaning here: the Berkeley Personality Lab says the BFI-2 does not have an official manual with published norms. It does not mean that no BFI-2 scores have been studied, or that a score has no descriptive content. The Berkeley page points to a separate paper reporting means for an American sample aged 20 to 60. Those scores were converted to percentage of maximum possible (POMP), a way to express a score relative to the scale’s possible range. The page describes age-related means by gender for each broad domain; it does not present a universal table turning any reader’s score into a percentile. A norm is a reference distribution built from scores collected under a specified procedure in a defined group. It allows comparison with that group, if the group and scoring suit the question. A mean summarizes a group; alone, it does not show where an individual falls or what share scored below them. “No official manual” leaves open an exception: a provider may have built and documented its own reference sample. The absence of an instrument-author norm does not prove that work sound or unsound. The reader must inspect the provider’s evidence. Until the report identifies its comparison source and conversion method, treat a clear scale score as a bounded description, not a population rank.

Sources: Big Five Inventory - Berkeley Personality Lab

Which comparison is the report actually making?

A BFI-2 report number can answer different questions depending on how it was produced. A scale score locates responses on its scale; a within-person comparison orders one reader’s domains; a sample mean describes a group average; a percentile compares a score with a reference distribution. These are not interchangeable. Before accepting “high,” “low,” or “average,” identify which comparison the report makes. A scale score tells you where responses fall under the stated scoring rule; it does not tell you how common that value is among other people. The Behavioral Scientist’s “How to Read Personality Test Results” uses an expressly invented example: a score of 3.6 on a 1-to-5 scale is above the midpoint, but that does not establish the population average. The midpoint belongs to the response scale; a sample mean is calculated from respondents. This is not BFI-2 evidence; it clarifies the distinction. A report may also compare the same person’s domains. If one domain score is numerically larger, it is higher under those scoring rules. That does not make it unusually high compared with other people. Read “my highest domain” as an ordering within one profile, not a population rank, unless the report separately provides a suitable reference comparison. A sample mean answers another question: what was the average score in the particular sample studied? The Berkeley Personality Lab’s BFI-2 page says there is no official manual with published norms, while pointing to an American sample paper with age-related means converted to POMP, or percentage of maximum possible. These offer context for that sample, but a mean alone does not show an individual’s position within it or provide a percentile conversion. A percentile makes a relative claim by locating a score within a specified set of comparison scores. Thus the same score can receive different percentile ranks when the reference group changes. The Behavioral Scientist demonstrates this with invented groups: the same example score gets different ranks. These are not BFI-2 norms. The actual report should identify its reference group and calculation rule, including how ties are handled. Translate the report label into a precise statement: “my score on this scale,” “my highest domain within this profile,” “a mean in this sample,” or “my percentile in this reference group.” If it says “above average” without identifying the average or group, ask what comparison is intended. A scale score describes responses; a profile ordering summarizes that profile; a sample mean gives context. Only a percentile based on a defined distribution supports a population-standing claim, and its usefulness depends on whether that comparison group fits the question.

Sources: Big Five Inventory - Berkeley Personality Lab; How to Read Personality Test Results

What form and scoring rule produced the number?

Before interpreting a BFI-2 number, establish what was administered and how its responses became that number. The full BFI-2 contains 60 items and represents five broad domains through 15 narrower facets. The shorter BFI-2-S and BFI-2-XS use fewer items, so a report based on one of them is not simply the full inventory displayed more compactly. The Berkeley Personality Lab’s BFI page distinguishes these forms and links to their documentation. Ask the provider to name the exact form, language version, and item set; “BFI-2” alone may not identify what the report measured. The scoring rule matters just as much as the form. In a self-report inventory, some items are phrased in the opposite direction from the trait they contribute to. Reverse-keying means recoding such responses so that the direction of the item aligns with the scale before combining responses. A domain or facet score may then be calculated from its assigned items, but a provider could also transform or rescale that result for display. Without the provider’s method, do not assume that a number is a sum, an average, a percentage, or a norm-referenced score. The original paper, “The Next Big Five Inventory,” describes the BFI-2’s hierarchical domains and facets; that description does not establish how an unspecified website implemented or displayed them. Shorter forms also change how much information is available for interpretation. The study “The Next Big Five Inventory: Developing and Assessing a 15-Facet Model to Enhance Bandwidth, Fidelity, and Predictive Power” examined the full measure’s structure and validation in particular samples. Separate research by the same instrument developers on the BFI-2 short forms reports differences in measurement properties across the 60-, 30-, and 15-item versions, including lower average domain reliability and self-peer agreement for shorter forms, and cautions against using the extra-short form for facet scores. These are findings about those forms in the studied contexts, not a verdict on every report. They do make form identification necessary: a detailed facet label should not be treated as equally informative regardless of how many relevant items were answered. Keep two questions separate. First, did the provider score the identified items in a way consistent with the named form, including its handling of reverse-keyed and missing responses? Second, did it compare the resulting score with a reference group and convert it into a percentile or category? A clear answer to the first question can make the scale result interpretable without answering the second. Conversely, a polished percentile cannot repair uncertainty about what questionnaire or scoring rule produced its starting value. Ask the provider for the form name, language, item set, missing-response rule, scoring method, and whether the displayed number is raw, averaged, or transformed. Those details let you decide what the score represents before deciding how far its interpretation can go.

Sources: Big Five Inventory - Berkeley Personality Lab; The Next Big Five Inventory (BFI-2): Developing and Assessing a Hierarchical Model With 15 Facets to Enhance Bandwidth, Fidelity, and Predictive Power; Short and Extra-Short Forms of the Big Five Inventory–2: The BFI-2-S and BFI-2-XS

What can the BFI-2 measurement evidence support?

The BFI-2 research supports a bounded claim: its domain and facet scales were designed to measure related but distinct personality tendencies,. In The Next Big Five Inventory (BFI-2): Developing and Assessing a Hierarchical Model With 15 Facets to Enhance Bandwidth, Fidelity, and Predictive Power, Study 3 tested measurement properties in an internet validation sample of 1,000 people and a student validation sample of 470. The researchers examined scale reliability, the proposed domain-and-facet structure, relations with other personality measures, and associations with self-reported and peer-reported criteria. This supports claims about those scales in those contexts. It is not evidence that an unidentified website used the same items, language, scoring, or administration, and it does not locate an individual within a population.

These evidence types answer different questions. Internal consistency describes how closely items in a scale vary together; it does not show that a score is accurate for every person. Retest reliability asks whether scores remain similar over time, while structural analyses ask whether item patterns fit the proposed domains and facets. Associations with established measures or reported behaviors show whether scores relate to other variables in expected ways. Self-peer agreement asks whether a participant's account corresponds with another person's ratings. None of these is a percentile conversion. A high reliability estimate cannot substitute for a reference distribution, and a correlation with an outside measure cannot establish that a particular report's labels are warranted. Validity means evidence supporting a specified interpretation or use, not a certificate that every conclusion is correct.

Form matters too. Short and Extra-Short Forms of the Big Five Inventory–2: The BFI-2-S and BFI-2-XS compared the 30-item short form and 15-item extra-short form with the full inventory in an internet sample of 2,000 and a university sample of 423; its self-peer analyses used 230 university participants. The authors reported that, on average, the short form retained about 90% of the full form's domain-scale reliability and self-peer agreement, while the extra-short form retained about 80%. They also reported lower facet-level reliability and self-peer agreement for the extra-short form, cautioning against using its single items to assess facets. These are comparisons between forms in those samples, not guarantees for a provider implementation. For a reader, the practical inference is modest: a clearly identified BFI-2 score can serve as a tentative description of a domain or facet tendency, but its evidential support depends on the form and scoring actually used. The validation studies support measured tendencies in studied contexts; they do not supply the missing norm or justify a population rank by themselves.

Sources: The Next Big Five Inventory (BFI-2): Developing and Assessing a Hierarchical Model With 15 Facets to Enhance Bandwidth, Fidelity, and Predictive Power; The Next Big Five Inventory (BFI-2): Developing and Assessing a Hierarchical Model With 15 Facets to Enhance Bandwidth, Fidelity, and Predictive Power; Short and Extra-Short Forms of the Big Five Inventory–2: The BFI-2-S and BFI-2-XS

What makes a comparison group suitable?

A comparison group is suitable when it resembles the people and conditions relevant to the question the score is being used to answer. That means checking more than a label such as “adults” or a large respondent count. Ask who took the measure, how they were recruited, when they responded, which BFI-2 form and language they completed, and how the scores were collected and converted. The International Test Commission’s *Guidelines on Test Use* says technical documentation should make it possible to evaluate the appropriateness of norm groups and the relevance of evidence for the population and intended use. It also cautions against conclusions drawn from norms that are irrelevant to the people being assessed or outdated. Sample size and sample fit answer different questions. A larger sample can make estimates less sensitive to chance variation, but it cannot by itself show that participants represent the group named in a report. A volunteer website sample, for example, might contain many responses while still differing from the population a reader assumes the percentile describes. That is an illustration of the distinction, not a claim about any particular BFI-2 provider. Recruitment, eligibility, response patterns, and collection dates help explain what a sample can represent. Subgroup coverage matters too: if a provider reports comparisons by age or language, the reader needs to know whether those categories were actually represented and whether the scores were produced comparably. The comparison also has to match the use. A group assembled to support broad self-reflection does not automatically justify a threshold for employment selection, and evidence for one purpose cannot simply be carried into another. The ITC guidelines distinguish technical qualities, relevant populations, and required use; they also recommend identifying the norms, scale types, and equations behind report labels. Berkeley Personality Lab’s BFI-2 page notes that no official manual with published norms exists, while pointing readers to American age-related means from a separate study. Those means may provide context, but their existence does not establish that an online provider’s comparison sample matches a particular reader or supports that provider’s percentile conversion. A useful request is specific: “Who was included in the comparison group, how were they recruited, when and in what language did they complete which BFI-2 form, and how were their scores converted into this label?” If the answer identifies a relevant group and a transparent procedure, assess that provider evidence on its own. If it does not, the score may still describe a position on a scoring scale, but the population comparison remains unsubstantiated. For self-reflection, that is a reason to keep the interpretation bounded; it is not evidence that the person’s responses have no meaning.

Sources: ITC Guidelines on Test Use; Big Five Inventory - Berkeley Personality Lab

Can a third-party report supply its own norms?

Yes. A provider can create its own reference distribution even though the Berkeley Personality Lab says there is no official BFI-2 manual with published norms. These are separate questions: one concerns an official manual; the other concerns comparison data collected by a report provider. The absence of the first does not prove the second impossible or establish that a provider has done it well. A report must show its evidence before a percentile can be read as a comparison with other people.

A percentile is a position calculated against scores from a stated group using a scoring and conversion procedure. To audit one, trace the chain: which BFI-2 form and language generated the responses; how responses became a scale score; who makes up the comparison distribution; and how the score became the reported rank or band. If a link is missing, the reader cannot tell what population the label describes. The Berkeley page says there is no official norms manual but points separately to age-related means for an American sample expressed as percentage of maximum possible. Such a mean is descriptive context, not a percentile table or a provider's conversion rule.

A research sample, a provider's site-user database, and a sample designed to represent a defined population may describe people who answered questions, but they do not support identical inferences. A study sample may address measurement properties rather than where later users stand. Site users may differ from the population implied by a report. A purpose-built reference sample may be more relevant, but adequacy still depends on who was recruited, when, in what language and form, and for what use. Size can improve precision under suitable sampling; size alone does not establish relevance or representativeness. These are questions to ask, not grounds to assume an undisclosed provider used a flawed sample.

The International Test Commission's *Guidelines on Test Use* offers a general standard: interpretation should account for the people assessed and the purpose, and documentation should let users judge the evidence. It cautions against conclusions based on irrelevant or outdated norms. This broad guidance is not a BFI-2-specific endorsement or a single required norming recipe. Applied here, it suggests a practical test: can the provider explain why its reference group fits the comparison and how the rank follows from the score and data?

If a report says “above average” or gives a percentile without identifying its reference group and conversion, the careful reading is that the comparison is unclear. That does not prove the number was fabricated or that the responses have no reflective value. It means the reader should not treat the label as an established population rank until the provider supplies enough information to evaluate it. Ask for the form and language, the sample's composition and collection date, its intended population, and the score-to-percentile method. The fair exception is a documented provider norm; the deciding issue is whether its evidence fits the stated comparison, not whether the provider is the instrument author.

Sources: Big Five Inventory - Berkeley Personality Lab; ITC Guidelines on Test Use

Why does form, language, and setting change the comparison?

A BFI-2 result is not automatically comparable across forms or languages. The same trait label can hide differences in which items were asked, how they were translated, and who answered them. Measurement invariance means evidence that a measure operates comparably across groups at a specified level. This matters for a report without official norms because an outside comparison is useful only if the reader's score and the reference scores can reasonably be treated as measurements of the same thing. The OECD’s *Survey of Adult Skills and the measurement of human capital* illustrates the issue in the 2023 Survey of Adult Skills, which used the 30-item BFI-2-S self-report. The OECD explains that cultural and linguistic differences can lead people in different countries to interpret similar questions differently. It therefore standardised scores to have equal means and variances within each country. Those scores can support analyses of relationships between reported skills and other characteristics within a country, the OECD says, but are not suited to comparing trait levels between countries. That is a survey-specific analytic choice, not a norm table for an individual BFI-2 report. Translation is one part of the problem; local evidence is another. In *The Bangla Big Five Inventory-2: a comprehensive psychometric validation*, researchers compared Bangla responses from 646 Bangladeshi participants who passed four attention checks with an earlier English-language internet sample of 1,000 people. The Bangladeshi sample was young and predominantly university students: 84% were students, and its mean age was 24.25. The study examined whether measurement properties held across language, gender, and education. Its discussion reports that Bangla and English versions did not achieve measurement invariance, and points to item interpretation and response styles as possible contributors. The authors also report that all items departed from a normal distribution in the Bangla sample, with responses tending toward agreement with more desirable items and disagreement with less desirable ones. These findings describe those samples and analyses; they do not establish that every Bangla speaker answers alike, or that the English form is a universal standard. The practical distinction is between evidence that a translation is understandable and evidence that its scores support a particular comparison. A careful translation can preserve the intended wording while still producing different item behavior or response patterns across settings. Conversely, a difference in scores does not by itself show whether the cause is trait levels, item meaning, response style, or sample composition. A report that compares someone with a group using another language, form, or administration should identify evidence for that exact comparison. If it cannot, read the result as a bounded description of responses under that version and setting, and withhold the population-rank claim.

Sources: Survey of Adult Skills and the measurement of human capital; The Bangla Big Five Inventory-2: a comprehensive psychometric validation

What does the newer cross-country evidence change?

The OECD’s 2026 technical report makes cross-country ranking of BFI-2-S scores in the 2023 Survey of Adult Skills harder to defend, but it does not show that an individual BFI-2 result is useless. Nor does it provide a percentile table for a personal online report. The study examined short forms used in a particular international survey. Its findings do not substitute for the reference group and scoring method a provider should disclose. In conventional item-level analyses, only Agreeableness met the initial configural model criteria; even there, the result for equal factor loadings was mixed across fit criteria. The other domains did not meet or clearly satisfy the tests for straightforward comparison. The report therefore cautions that country rankings could be biased. Measurement invariance means a measure operates comparably across groups at a specified level. Metric invariance concerns whether items relate to the underlying domain similarly, and supports comparing relationships or variances. Scalar invariance is a stronger condition needed to compare latent means. Metric invariance alone cannot establish that one country’s average score is higher, much less where an individual stands in a percentile ranking. (OECD, *Psychometric properties of the BFI-2 inventory in the 2023 Survey of Adult Skills*.) There is also a material qualification. When the OECD modeled the BFI-2-S using its 15 facet scores rather than six individual items per domain, metric invariance held for Extraversion, Agreeableness and Emotional Stability across the 21-country analysis; it did not hold for Conscientiousness or Open-Mindedness. This is a better result for three domains, but it still is not scalar invariance and does not license comparing their average levels as if the groups shared the same score origin. Using alignment optimization, which allows some item parameters to vary, the report found noninvariant items below its cited 25% threshold. It suggests comparability may be less harmed than strict tests indicate. That counterpoint tempers, rather than erases, the caution: it does not turn every domain, form or score into an interchangeable measure. The OECD’s *Survey of Adult Skills – Reader’s Companion* explains that the survey’s within-country standardized scores have mean zero and standard deviation one in each country, so their country means cannot be compared. The technical report itself says its results offer a first picture and that further work is needed to establish when these scales provide valid and reliable estimates. Do not assume a BFI-2-S label has identical meaning across countries because the domain names match. Ask what form, language, comparison group and conversion generated the report’s claim. This OECD study cannot answer those provider-specific questions or supply an individual norm.

Sources: Psychometric properties of the BFI-2 inventory in the 2023 Survey of Adult Skills; Survey of Adult Skills and the measurement of human capital

How should you use a BFI-2 score without norms?

Use a transparently scored BFI-2 result as a tentative description of tendencies, then test it against examples and counterexamples from your experience. Without a suitable reference distribution and disclosed conversion, the score does not establish where you rank among other people. That limit does not prevent reflection; it changes the question from “Am I high?” to “What does this label describe, and when does it fit?” Do not turn reflection into diagnosis, population rank, or a recommendation about which job to take.

Suppose a report labels one domain “high” but gives no comparison group. Ask what high means there: a point on the scoring scale, a comparison with your other domains, or a rank against other respondents. If the report does not say, do not assume the broadest claim. Instead, ask what the domain description might look like in a decision. If it suggests seeking structure, consider a recent task: did you make a plan before starting, and did that help? Then look for a counterexample. Perhaps you began an urgent task without a plan, or used someone else’s structure. Both observations matter. This illustration supports reflection; it does not validate the score for you or show that a tendency is fixed.

Keep the report’s language close to observable behavior. “I write down steps when a task has several dependencies” is easier to examine than “I am always organized.” The first statement leaves room for context and exceptions. You might discuss one situation with a colleague or coach instead of treating a trait label as a verdict. If the description feels unfamiliar, consider whether the items, language, or circumstances affected your answers. A mismatch is a reason to inspect the interpretation, not proof that you or the measure is defective.

Reliability concerns how consistently a measure produces scores. Even a consistent score cannot supply a missing comparison group: repeating a measurement does not show how you compare with others. The International Test Commission’s Guidelines on Test Use advises against conclusions based on irrelevant or outdated norms and calls for scale labels. For reflection, a description can be a prompt if you keep its reach modest. A consequential decision such as hiring or promotion should not rest on this score alone; it does not establish suitability or performance.

The Berkeley Personality Lab’s Big Five Inventory page says there is no official BFI-2 manual with published norms, while pointing to separate American sample averages. Those averages provide research context, but do not tell you your rank or validate a provider’s label. Turn the label into a question about a recurring choice or collaboration: when does this tendency appear, and what is one recent exception? Keep the answer provisional. If the number remains unclear, ask the provider what it represents before drawing a comparison.

Sources: Big Five Inventory - Berkeley Personality Lab; ITC Guidelines on Test Use

What should you ask the report provider next?

Ask the provider to trace one label from your answers to its interpretation: “Which BFI-2 form and language did I complete, what does this number represent, and what reference group and conversion support the percentile or high/low label?” This separates scoring from comparison. The Berkeley Personality Lab’s “Big Five Inventory” page says there is no official BFI-2 manual with published norms, while noting American age-related sample means. A provider could still document its own comparison sample, but a sample mean alone does not show where an individual falls in a distribution. The “ITC Guidelines on Test Use” treat appropriate interpretation as dependent on the reason for testing and its context. An answer should identify the population the comparison describes and explain how the label follows from it. If those details are unavailable, keep the result as a description under its scoring method; do not turn it into a population rank or consequential work judgment. Use it to ask: when does the described tendency appear, and what example would complicate that reading? If your remaining question concerns decisions or collaboration at work, the live Work Pattern Report at /assessment offers a low-stakes reflection across work-pattern continuums. It has no norms and does not recommend a job, so it cannot fill the BFI-2 report’s missing comparison. It may help organize observations; /topics offers assessment-literacy guidance. Ask the provider to show how a label moves from scoring rule to reference group to interpretation, then decide whether that chain supports the claim you received.

Sources: ITC Guidelines on Test Use; Big Five Inventory - Berkeley Personality Lab