Beyond the Big Five: HEXACO, Hierarchies, and Alternative Structural Models

By /

Last Updated May 22, 2026

The history of personality structure does not end with the Big Five. Once a five-domain architecture became dominant, the next question was unavoidable: is this the best large-scale map of personality, or only the most influential one? That question opened the door to alternative structural models—especially the HEXACO framework, hierarchical trait models, facet-rich architectures, circumplex approaches, network models, and proposals for higher-order meta-traits. These alternatives do not all reject the Big Five. Some revise it, some expand it, some reorganize its content, and some argue that personality must be modeled at several levels at once.

The resulting debate is one of the most important in contemporary personality psychology because it concerns not merely which labels researchers prefer, but how personality structure itself should be represented, compared, validated, and interpreted. A structural model is not a neutral filing cabinet. It determines which forms of individual difference become visible, which are compressed, which are separated, which are treated as central, and which remain difficult to see.

This article argues that going beyond the Big Five should not be understood as a rejection of broad trait science. It is better understood as a maturation of the field. The Big Five gave personality psychology a durable common language. HEXACO, trait hierarchies, facets, aspects, circumplex models, and network approaches ask whether that language is sufficiently precise for questions about morality, exploitation, emotionality, prediction, culture, development, pathology, and the structure of personality itself.


Main Library
Publications


Article Map
Personality Psychology


Related Topic
Cognitive Psychology


Related Topic
Developmental Psychology


Related Topic
Social Psychology
Series context: This article is part of the Personality Psychology knowledge series, which examines traits, identity, temperament, self-regulation, culture, development, relationships, institutions, moral character, mental health, creativity, physical health, and the enduring psychological patterns through which persons engage the world.
Restrained institutional illustration of a human profile surrounded by trait wheels, hierarchy diagrams, network structures, and dimensional models representing alternatives to the Big Five personality framework.
Alternative structural models such as HEXACO, trait hierarchies, and network approaches expand personality theory beyond the Big Five by examining additional domains, facets, and patterns of organization.

The central issue is not whether personality has five, six, or some other number of dimensions in a simplistic sense. The deeper issue is how to build maps that are empirically stable, theoretically meaningful, culturally responsible, predictively useful, and proportionate to the questions being asked. The Big Five remains one of the strongest maps available. But personality is too complex for any single map to exhaust the territory.

Why go beyond the Big Five?

The Big Five became dominant because it provided a durable broad structure for organizing personality description. Extraversion, agreeableness, conscientiousness, neuroticism, and openness offered a shared language for comparing individuals, building research programs, summarizing self-report and observer-report data, and relating personality to major life outcomes. The model’s success was not accidental. It brought order to a historically fragmented field.

But success generated its own pressure. Once researchers accepted the usefulness of five large domains, they could begin asking whether those domains were optimally specified, whether they were too coarse for certain explanatory tasks, and whether important personality content had been omitted, compressed, or redistributed in ways that obscured substantive differences. The question became not whether the Big Five is useful, but whether it is sufficient.

There are at least four reasons to go beyond the Big Five. First, some researchers argue that a five-domain solution compresses morally and socially important content into overly broad factors. Honesty, humility, sincerity, greed avoidance, fairness, manipulation, entitlement, and exploitative behavior may not be cleanly represented by standard Big Five Agreeableness. HEXACO responds directly to this problem.

Second, broad domains often need to be decomposed into facets or aspects if personality is to predict behavior with adequate precision. A person may be high in openness because of aesthetic sensitivity, intellectual curiosity, unconventionality, imagination, or emotional receptivity. Those are not identical psychological tendencies. A broad score may be useful, but the narrower structure may explain more.

Third, broad domains themselves may sit within higher-order hierarchies. Some models propose aspects between facets and domains; others propose meta-traits above domains; still others ask whether there is a general factor of personality. These proposals remain debated, but they show that personality structure can be examined upward as well as downward.

Fourth, alternative structures ask whether traits should always be represented as independent factors. Circumplex models, interpersonal models, and network approaches suggest that structure may also be geometric, relational, or organized through patterns of empirical association rather than through factor extraction alone.

The point is not to abandon the Big Five. The point is to understand what it compresses, where it works well, where it works less well, and when another structural map provides better explanation. Going beyond the Big Five is therefore a sign of cumulative science. A field that has built a successful map can begin asking where the map still needs greater resolution.

Back to top ↑

The HEXACO model

The most influential structural alternative to the Big Five is the HEXACO model, developed principally by Michael Ashton and Kibeom Lee. HEXACO proposes six major dimensions: Honesty-Humility, Emotionality, Extraversion, Agreeableness, Conscientiousness, and Openness to Experience. It is measured through instruments such as the HEXACO Personality Inventory-Revised and is presented by its authors as a six-dimensional personality framework grounded in lexical and cross-language research.

HEXACO is important because it does not simply add a sixth label to the Big Five. It reorganizes personality space. Extraversion, Conscientiousness, and Openness remain relatively familiar compared with their Big Five counterparts, but the content of Agreeableness and Neuroticism is redistributed. HEXACO Emotionality includes fearfulness, anxiety, dependence, and sentimentality. HEXACO Agreeableness emphasizes patience, forgiveness, gentleness, and flexibility rather than anger-proneness. Honesty-Humility captures sincerity, fairness, greed avoidance, and modesty.

This means HEXACO is both an expansion and a repartitioning. It expands the space by naming Honesty-Humility as a broad domain, and it repartitions familiar content by reorganizing emotional and interpersonal tendencies. The structural claim is therefore more ambitious than “five factors plus one.” It says that personality becomes more intelligible when certain moral, emotional, and interpersonal tendencies are separated differently.

The model’s lexical foundation matters because lexical studies begin from the premise that important personality differences become encoded in natural language. If multiple languages support a six-factor structure, that strengthens the claim that the structure reflects recurring human concerns rather than an idiosyncratic scale construction. At the same time, lexical evidence is not automatically final. Language reflects culture, history, power, value, translation, and social salience. HEXACO is therefore strong because it has lexical and criterion support, but it still belongs inside ongoing structural debate.

The enduring value of HEXACO lies in its ability to make visible forms of personality content that standard Big Five language can blur. It provides a more explicit framework for discussing sincerity, fairness, greed, entitlement, exploitation, and humility as broad individual differences rather than as scattered subcomponents of agreeableness or moral judgment.

That is why HEXACO has become the leading alternative to the Big Five. It does not merely disagree about factor count. It asks whether a morally important region of personality space deserves its own domain.

Back to top ↑

Honesty-Humility and the six-factor claim

The most distinctive feature of HEXACO is Honesty-Humility. According to the model’s scale descriptions, high scorers tend to avoid manipulating others for personal gain, feel little temptation to break rules, show low entitlement, and are not strongly attracted to status and luxury. Low scorers are described as more willing to flatter, exploit, break rules for profit, and pursue self-importance. The theoretical importance of this dimension is that it isolates a cluster of dispositional tendencies related to sincerity, fairness, greed avoidance, and modesty that many researchers believe is underrepresented in the Big Five.

This matters because Honesty-Humility is especially relevant to morally charged interpersonal behavior. Many consequential social behaviors concern not merely warmth or sociability, but exploitation, deception, status-seeking, entitlement, reciprocity, fairness, and willingness to use others for gain. A person can be socially skilled, emotionally stable, ambitious, and even outwardly agreeable while still being manipulative or exploitative. A structural model that does not separate this content may miss a crucial dimension of interpersonal risk.

Honesty-Humility also helps clarify the difference between niceness and integrity. Agreeableness often concerns compassion, patience, cooperation, and interpersonal tolerance. Honesty-Humility concerns whether a person is disposed toward sincerity, fairness, modesty, and low exploitation. These domains can overlap in ordinary judgment, but they are not the same. Someone may be gentle and conflict-averse but still status-seeking. Another person may be blunt or low in warmth but deeply fair. HEXACO gives the field a vocabulary for this distinction.

The six-factor claim therefore has both empirical and ethical significance. Empirically, Honesty-Humility can improve prediction for outcomes involving deception, exploitation, greed, cheating, corruption, workplace deviance, and low-integrity conduct. Ethically, it reminds personality psychology that socially consequential character does not always fit neatly into warmth or emotional stability. Some forms of harm are not about emotional volatility or low sociability. They involve self-serving use of others.

At the same time, Honesty-Humility must be interpreted carefully. It is a personality construct, not a moral verdict on the whole person. A low score should not be treated as proof of wrongdoing, criminality, corruption, or inevitable exploitation. A high score should not be treated as proof of virtue in all contexts. Like all traits, it describes probabilistic tendencies measured under specific conditions.

The strength of Honesty-Humility is that it improves structural visibility. It makes a morally important region of personality space harder to ignore. But responsible use still requires validation, context, humility, and proportional interpretation.

Back to top ↑

How HEXACO reorganizes Big Five content

The relation between HEXACO and the Big Five is partly parallel and partly non-isomorphic. Extraversion, Conscientiousness, and Openness to Experience remain relatively close across the two frameworks. The more complex differences involve Big Five Agreeableness and Neuroticism, whose content is repartitioned into HEXACO Agreeableness, Emotionality, and Honesty-Humility. This is where the structural debate becomes substantive rather than merely terminological.

Big Five Agreeableness often includes warmth, trust, altruism, compliance, modesty, and tender-mindedness. In HEXACO, some of this content shifts toward Honesty-Humility, especially sincerity, fairness, modesty, and greed avoidance. HEXACO Agreeableness becomes more centered on forgiveness, patience, flexibility, and gentleness. This changes how interpersonal traits are interpreted. It separates low exploitation from low anger or high patience.

Big Five Neuroticism also differs from HEXACO Emotionality. Standard neuroticism often includes anxiety, depression, vulnerability, anger, self-consciousness, and emotional instability. HEXACO Emotionality includes fearfulness, anxiety, dependence, and sentimentality, while anger-related content is more strongly related to low Agreeableness. This reorganization matters because it changes the theoretical meaning of emotional distress, fearfulness, dependence, anger, and interpersonal flexibility.

The structural implication is important: models are not interchangeable translations. If one model places anger-proneness within Neuroticism and another places it closer to low Agreeableness, the same behavior may be interpreted differently. If one model treats sincerity and modesty as part of Agreeableness while another treats them as part of Honesty-Humility, then correlations with outcomes may change. Prediction, interpretation, and theory all depend on how the map is drawn.

This is why debates about personality structure cannot be reduced to preference for five or six labels. When models sort dispositional content differently, they change the empirical meaning of trait scores. A trait architecture determines which psychological tendencies are treated as belonging together and which are treated as distinct. It shapes the kinds of questions that are easy to ask.

HEXACO’s contribution is therefore structural, not cosmetic. It argues that personality space is better represented when morally relevant exploitation/fairness content is separated from broader agreeableness and when emotionality is reorganized around fear, anxiety, dependence, and sentimentality rather than the full Big Five neuroticism package.

The model’s value lies in that reorganizing move. Whether one ultimately prefers Big Five or HEXACO for a given purpose, the comparison forces personality psychology to become more explicit about what each domain actually contains.

Back to top ↑

Hierarchical models of personality

Another route beyond the Big Five does not multiply domains so much as embed them in hierarchy. Hierarchical models hold that broad traits sit above narrower facets, and may themselves sometimes load onto even broader higher-order dimensions. The core idea is that personality is not structured at a single level. It is layered.

At the broadest familiar level, one may speak of extraversion, agreeableness, conscientiousness, neuroticism, and openness. At the next level, one may speak of aspects such as enthusiasm and assertiveness, compassion and politeness, industriousness and orderliness, withdrawal and volatility, openness and intellect. At still narrower levels, one may speak of facets such as sociability, activity, positive emotions, anxiety, depression, self-discipline, deliberation, trust, compliance, aesthetic sensitivity, or intellectual curiosity.

Hierarchical models matter because broad domains are both powerful and incomplete. They summarize large regions of personality space, but they also compress important internal variation. Two people may be similarly high in conscientiousness while one is high in orderliness and moderate in industriousness and the other is highly industrious but not especially orderly. Two people may have similar openness scores while one is aesthetically sensitive and the other is intellectually curious. The hierarchy keeps those differences visible.

Some hierarchical approaches emphasize facets. Others emphasize aspects as an intermediate level between domains and facets. Still others propose higher-order meta-traits or even a general factor of personality. These higher-order proposals remain debated. They may reflect substantive personality structure, social desirability, evaluative bias, method variance, or some combination of these. The debate itself is useful because it makes researchers clarify what kind of hierarchy they are proposing.

The most practical lesson of hierarchical trait theory is that the right level depends on the question. Broad domains may be best when the criterion is broad, the sample is large, or the goal is communication. Facets may be best when the criterion is narrow, the mechanism is specific, or the person’s profile is uneven. Aspects may provide a middle level when domains are too broad and facets too numerous.

Hierarchical models therefore extend the Big Five by deepening it. They do not require abandoning broad factors. They require treating broad factors as one level of an organized structure rather than as the entire architecture of personality.

A mature personality science should be able to move up and down this hierarchy: from item to nuance, nuance to facet, facet to aspect, aspect to domain, domain to meta-structure, and back again.

Back to top ↑

Facets, aspects, and granularity

The question of granularity is central to alternative structural models. Broad domains are useful for large-scale description, but they can obscure substantively important distinctions. A person may be high on agreeableness because they are compassionate, because they are compliant, because they avoid conflict, because they trust others, or because they are patient. Those patterns are related, but not identical. A person may be high on openness because of imagination, aesthetic sensitivity, intellectual curiosity, emotional receptivity, unconventionality, or tolerance for ambiguity. Treating all of that content as one score can be useful, but it can also flatten interpretation.

This is one of the strongest arguments for hierarchical architecture. Models that include multiple levels can preserve the communicative convenience of broad factors while allowing narrower subcomponents to do explanatory work. The issue is not whether breadth or precision is better in the abstract. It is whether the structural level chosen is appropriate for the research question.

The bandwidth-fidelity tradeoff is the classic way to express this problem. Broad domains offer bandwidth: they cover wide regions of personality and often provide reliable summaries. Narrow facets offer fidelity: they capture more specific psychological content and may predict specific outcomes more accurately. A broad measure may be better for broad life outcomes; a narrow measure may be better for specific behaviors. Neither level is universally superior.

Granularity also matters for applied interpretation. A broad score can be easier to communicate but may overgeneralize. A facet score can be more precise but may invite false certainty if reliability and validity are weak. A long hierarchy of narrow scores may look impressive while becoming difficult to interpret responsibly. More detail is only valuable when it clarifies the construct, improves prediction, or supports better reasoning.

Aspects provide a useful middle path. They are broader than facets but narrower than domains. For example, separating extraversion into enthusiasm and assertiveness can clarify whether a person’s extraversion reflects warmth and positive emotional engagement or social dominance and agency. Separating conscientiousness into industriousness and orderliness can clarify whether self-regulation is expressed through effort, productivity, structure, or neatness.

Granularity is therefore not just a psychometric issue. It is a question of explanatory scale. A model must be detailed enough to capture meaningful differences, but not so fragmented that it loses theoretical coherence.

The best structural models help researchers and practitioners ask: how much detail is needed here, and what is lost or gained by moving to a broader or narrower level?

Back to top ↑

Circumplex and network approaches

Not all alternatives to the Big Five are factor-expansion models. Some researchers use circumplex frameworks, especially in interpersonal personality research, where traits are organized in geometric relation rather than simply as independent domains. The interpersonal circumplex, for example, often represents interpersonal behavior along dimensions such as agency/dominance and communion/warmth. Traits and behaviors are positioned around a circular space, showing how they relate through proximity and opposition.

Circumplex models are useful because many interpersonal qualities are relational. Dominance, submissiveness, warmth, coldness, friendliness, hostility, assertiveness, and deference are not merely independent traits. They have patterned relations to one another. A circumplex can show that some traits are adjacent, some are opposite, and some are blends of broader interpersonal dimensions. This geometric representation can be more informative than a simple list of factors when the domain is inherently relational.

Network approaches raise a different possibility. Instead of assuming that traits are caused by latent factors, network models examine traits, symptoms, behaviors, or item responses as systems of mutual association. A network model might ask how anxiety, withdrawal, irritability, distrust, avoidance, low energy, and self-criticism relate directly to one another. The emphasis shifts from hidden dimensions to patterns of interconnection.

Nomological-web and clustering approaches also broaden the structural conversation. Traits can be grouped not only by lexical covariance, but by their empirical relations to outcomes, behaviors, values, motives, or institutional contexts. A cluster of traits may matter because it predicts certain forms of leadership, work behavior, health risk, creativity, exploitation, moral conduct, or social conflict. In this view, structure is partly defined by consequences.

These alternatives reveal different structural intuitions. A factor model asks what latent dimensions summarize covariance. A circumplex asks how traits are arranged in relation to one another in a continuous space. A network model asks how trait-relevant components connect and reinforce one another. A nomological approach asks which constructs belong together because of their empirical relations with other phenomena.

Each model changes the meaning of “structure.” None is automatically superior for every question. Factor models are powerful for broad summarization. Circumplex models are powerful for relational domains. Network models are useful when direct associations and feedback loops matter. Nomological models are useful when prediction and external validity are central.

The field gains when it recognizes these as complementary maps. Personality structure is not exhausted by counting factors.

Back to top ↑

What these models explain better

Alternative structural models matter when they improve explanation, prediction, interpretation, or conceptual clarity. HEXACO is often especially informative for behavior involving exploitation, manipulation, greed, entitlement, fairness, sincerity, rule-breaking, and low-integrity conduct because Honesty-Humility isolates a morally important domain not cleanly represented in the standard Big Five. That is not a trivial addition. It changes how researchers can study antisocial and self-serving behavior.

Hierarchical models matter when broad domains are too coarse to predict specific outcomes or when intermediate trait layers capture psychologically meaningful distinctions lost at the domain level. A broad trait may predict a broad outcome, but a facet may better predict a specific behavior. Orderliness and industriousness may differ in their relation to work habits. Assertiveness and sociability may differ in their relation to leadership. Compassion and politeness may differ in their relation to caregiving, conflict avoidance, or moral judgment.

Circumplex models explain better when the content is relational. Interpersonal dominance, warmth, submissiveness, hostility, and affiliation are difficult to understand as isolated traits because they derive meaning from social position and interpersonal orientation. Circumplex models can show relational structure more naturally than some factor models.

Network models may explain better when the goal is to understand how trait-relevant tendencies reinforce one another over time. For example, social anxiety may increase avoidance, avoidance may reduce social practice, reduced practice may intensify anxiety, and shame may amplify withdrawal. Such patterns may be better understood as feedback systems than as simple reflections of one latent factor.

More broadly, alternative models explain better whenever a five-score profile is too compressed for the question. They can clarify why two individuals with similar Big Five profiles differ in antagonism, sentimentality, honesty, volatility, or social dominance. They can show why traits cluster at more than one level. They can identify when morally important behavior is hidden inside a broad interpersonal domain. They can reveal that structure may be hierarchical, geometric, networked, or criterion-linked.

The best argument for alternative structural models is not novelty. It is explanatory gain. A model earns its place when it makes personality differences more intelligible, predicts meaningful outcomes more accurately, improves measurement, clarifies theory, or prevents important content from being hidden by an overly broad map.

The strongest personality science is therefore comparative. It asks not “Which model is always right?” but “Which model best serves this question, this evidence, this population, this criterion, and this level of interpretation?”

Back to top ↑

Culture, translation, and structural comparability

Alternative structural models raise difficult questions about culture and comparability. Personality taxonomies are often built from language, and language is never culturally neutral. Lexical studies assume that socially important differences become encoded in words, but which differences become salient, how they are named, and how they cluster can vary across languages, histories, institutions, and social worlds.

This matters for both Big Five and HEXACO research. Cross-language evidence can support broad structural patterns, but it can also reveal that factor structures are not equally stable everywhere. Some lexical spaces may make certain distinctions highly visible; others may distribute them differently. A trait such as humility, assertiveness, emotional restraint, social dominance, obedience, sincerity, or openness may carry different meanings across cultural settings.

Translation is not merely technical. A translated item may preserve dictionary meaning while shifting social implication. A question about “standing up for oneself” may evoke confidence in one context, disrespect in another, resistance in another, and inappropriate self-assertion in another. A question about modesty may reflect humility, gender expectation, class etiquette, religious virtue, strategic self-presentation, or fear of social sanction depending on context.

Measurement invariance becomes essential. Before comparing groups, researchers must ask whether the same items and scales function similarly across populations. A structural model that appears comparable at the domain level may conceal facet-level differences. Conversely, a domain may appear unstable because particular items fail to travel well across language or culture.

HEXACO’s cross-language foundation is valuable precisely because it engages these questions. But no model escapes the need for cultural humility. Personality structure may show broad regularities across human populations, yet still be shaped by local language, ecological conditions, family systems, economic arrangements, religion, gender norms, institutions, and power.

Structural comparability also has ethical significance. If a model developed in one context is applied uncritically in another, it can mislabel culturally meaningful behavior as personality difference. Emotional restraint, deference, communal obligation, modesty, confrontation, ambition, or skepticism may be interpreted differently depending on social world. A personality model must not treat one cultural style as the universal baseline.

Going beyond the Big Five therefore also means going beyond naïve universalism. A stronger structural science asks which personality patterns generalize, which require local interpretation, and how measurement can respect both common human variation and cultural specificity.

Back to top ↑

Limits of alternative structural models

Alternative models do not escape the classic limits of personality structure. More dimensions can increase nuance, but they can also increase complexity and measurement burden. A six-factor model may improve coverage of honesty and exploitation, but it also requires careful explanation and validation. Facet-rich models can increase precision, but they can overwhelm interpretation if too many scores are presented without theory. Network models can reveal connections, but they can become unstable across samples or difficult to interpret causally.

Hierarchies also raise the problem of where to stop. Domains, aspects, facets, nuances, states, items, behaviors, situations, and narratives all contain information. The existence of narrower levels does not mean every applied or theoretical problem requires maximum granularity. Detail without discipline can become fragmentation. A good model is not the one with the most labels. It is the one that organizes the right amount of information for the purpose at hand.

Cross-language and cross-cultural evidence can support structural alternatives, but it can also reveal instability in translation, varying lexical density, and differences in the local salience of personality descriptors. A model may be broadly useful while still requiring cultural adaptation. Universal claims should be made carefully.

There is also a deeper limitation. No structural model—Big Five, HEXACO, hierarchical, circumplex, or network-based—fully captures motive, attachment, identity, conflict, life narrative, moral formation, religious life, trauma, institutional constraint, or the meaning of a person’s choices. Structural models are maps of dispositional space. They are not complete theories of personhood.

This limitation is not a failure. It clarifies scope. Personality structure is one layer of personality science. It helps describe recurring individual differences and their organization. But to understand a person, one must also consider development, context, relationships, goals, values, identity, memory, biography, culture, and institutions.

Alternative structural models are therefore most powerful when they remain humble about what structure can and cannot explain. They can improve the map, but they do not become the territory. They can clarify the architecture of traits, but they do not replace the living person.

The best use of structural models is disciplined pluralism: use the map that fits the question, compare models where appropriate, validate claims carefully, and avoid treating any taxonomy as the final word on human individuality.

Back to top ↑

Professional use and applied boundaries

Alternative structural models can be professionally useful in research, psychometric education, assessment design, coaching reflection, organizational learning, leadership development, ethics training, selection-system critique, clinical formulation, and personality science communication. They help professionals understand why a five-domain profile may be too coarse for some questions, why HEXACO may be useful for integrity-related constructs, why facets can improve interpretation, and why structural level must match the intended use.

A professional scaffold based on HEXACO, hierarchy, and alternative structure can support legitimate work: comparing five- and six-factor models, demonstrating exploratory factor analysis, teaching item-to-factor logic, examining bandwidth-fidelity tradeoffs, exploring Honesty-Humility as a construct, and showing why broad traits should not be overinterpreted. These are appropriate uses in professional education, research prototyping, consulting support, organizational learning, and methodological demonstration.

But professional use does not mean unrestricted assessment use. A synthetic dataset is not evidence about real people. A six-factor score is not a moral verdict. A low Honesty-Humility score is not proof of misconduct. A facet profile is not a diagnosis. A structural model is not a hiring system. A network graph is not a personality destiny map. The more consequential the decision, the stronger the validation burden.

Alternative trait tools are appropriate for education, research prototyping, reproducible workflow development, psychometric demonstration, consulting support, organizational learning, coaching reflection, and model-comparison work. They are not appropriate as standalone systems for hiring, promotion, termination, clinical diagnosis, educational placement, legal evaluation, insurance decisions, surveillance, relationship matching, or individual prediction.

Any consequential use involving real people would require validated instruments, qualified interpretation, documented intended use, informed consent where appropriate, privacy protections, fairness and measurement-invariance analysis, careful communication of uncertainty, and appropriate ethical and legal oversight. If genetic, clinical, workplace, student, or vulnerable-population data are involved, the governance burden is even higher.

The intended professional use is analytic, educational, methodological, and reflective. The purpose is to reason more carefully about personality structure—not to convert alternative models into unsupported classification, moral labeling, or gatekeeping systems.

Back to top ↑

Mathematical lens: nested, expanded, and repartitioned trait space

Alternative structural models can be clarified by writing personality architecture more formally. Suppose \(\mathbf{x}\) is a vector of observed indicators. A broad latent-factor model represents those indicators as:

\[
\mathbf{x} = \mathbf{\Lambda}\mathbf{f} + \boldsymbol{\delta}
\]

Interpretation: \(\mathbf{x}\) is the vector of observed item responses, \(\mathbf{\Lambda}\) is the loading matrix, \(\mathbf{f}\) is the vector of latent trait dimensions, and \(\boldsymbol{\delta}\) contains item-specific variance and error. Structural models differ partly in how many latent dimensions they posit and how those dimensions are interpreted.

In a five-factor model, the broad latent vector can be represented as:

\[
\mathbf{f}_{BF} = (E, A, C, N, O)
\]

Interpretation: The Big Five vector summarizes Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness. It is a high-bandwidth summary of personality space.

In HEXACO, the broad latent vector becomes:

\[
\mathbf{f}_{HX} = (H, E_m, X, A_h, C, O)
\]

Interpretation: The HEXACO vector includes Honesty-Humility \(H\), Emotionality \(E_m\), Extraversion \(X\), Agreeableness \(A_h\), Conscientiousness \(C\), and Openness \(O\). The six-factor model is not merely larger; it repartitions some emotional and interpersonal content.

HEXACO’s structural claim can be represented as a repartitioning of some Big Five content:

\[
(A_{BF}, N_{BF}) \rightarrow (A_{HX}, E_{HX}, H_{HX})
\]

Interpretation: Big Five Agreeableness and Neuroticism do not translate one-to-one into HEXACO Agreeableness and Emotionality. Some content is redistributed into Honesty-Humility. This is why comparing the models is not just a matter of counting factors.

Hierarchical models add nesting. If \(D_j\) is a broad domain and \(f_{jm}\) is one of its narrower facets, then:

\[
f_{jm} = \alpha_{jm} + \beta_{jm}D_j + \varepsilon_{jm}
\]

Interpretation: A facet \(f_{jm}\) inherits variance from broad domain \(D_j\), but also retains facet-specific variance \(\varepsilon_{jm}\). This explains why two people can have similar domain scores but different facet profiles.

Higher-order structure can be represented by allowing domains to load onto broader meta-traits \(M_k\):

\[
D_j = \sum_{k=1}^{K}\gamma_{jk}M_k + \zeta_j
\]

Interpretation: Domain \(D_j\) may reflect one or more higher-order meta-traits \(M_k\), while \(\zeta_j\) preserves domain-specific variance. Whether higher-order factors are substantively meaningful, evaluative, methodological, or partly artifactual remains debated.

A circumplex representation places traits in a geometric space:

\[
\mathbf{p}_i = r_i(\cos \theta_i, \sin \theta_i)
\]

Interpretation: A person or trait profile \(\mathbf{p}_i\) can be represented by radius \(r_i\) and angle \(\theta_i\) in a two-dimensional interpersonal space. This is useful when trait meaning depends on relational position, such as dominance and warmth.

A network approach can be represented with an adjacency matrix:

\[
\mathbf{W} = [w_{ij}]
\]

Interpretation: \(\mathbf{W}\) contains associations among trait-relevant nodes, symptoms, behaviors, or items. Rather than assuming one latent factor explains all covariance, the model examines patterns of direct relation among components.

These equations do not resolve the theoretical debate, but they make it explicit. Structural models differ in how many dimensions they posit, how they partition covariance, how many nested levels they regard as meaningful, and whether they treat structure as latent, hierarchical, geometric, or networked.

Back to top ↑

R: comparing five- and six-factor structures

The R example below shows how a researcher might compare five-factor and six-factor solutions using a synthetic personality item pool. It also illustrates how to inspect loadings, compare rough model fit, compute factor scores, and examine whether a six-factor solution adds interpretable structure.

# Beyond the Big Five: HEXACO, Hierarchies, and Alternative Structural Models
# R workflow for comparing five- and six-factor structures

# Install packages if needed:
# install.packages(c("readr", "dplyr", "psych", "GPArotation", "broom"))

library(readr)
library(dplyr)
library(psych)
library(GPArotation)
library(broom)

# -------------------------------------------------------------------
# Load item-level personality data
# -------------------------------------------------------------------

# Expected structure:
# Each row is a respondent.
# item1:item72 represent a broad personality item pool.
# hexaco_honesty_proxy, big_five_agreeableness_proxy, and
# outcome_integrity are optional scored variables for comparison.

trait_data <- read_csv("alternative_structure_items.csv")

str(trait_data)
summary(trait_data)

# -------------------------------------------------------------------
# Select item pool
# -------------------------------------------------------------------

item_pool <- trait_data %>%
  select(item1:item72)

# -------------------------------------------------------------------
# Parallel analysis to inspect dimensionality
# -------------------------------------------------------------------

fa.parallel(
  item_pool,
  fa = "fa",
  n.iter = 100,
  main = "Parallel Analysis for Alternative Personality Structures"
)

# -------------------------------------------------------------------
# Fit five-factor and six-factor solutions
# -------------------------------------------------------------------

efa_5 <- fa(
  item_pool,
  nfactors = 5,
  rotate = "oblimin",
  fm = "ml"
)

efa_6 <- fa(
  item_pool,
  nfactors = 6,
  rotate = "oblimin",
  fm = "ml"
)

cat("Five-factor solution loadings:\n")
print(efa_5$loadings, cutoff = 0.30)

cat("\nSix-factor solution loadings:\n")
print(efa_6$loadings, cutoff = 0.30)

# -------------------------------------------------------------------
# Compare practical fit summaries
# -------------------------------------------------------------------

fit_comparison <- data.frame(
  model = c("five_factor", "six_factor"),
  nfactors = c(5, 6),
  rmsr = c(efa_5$rms, efa_6$rms),
  tli = c(efa_5$TLI, efa_6$TLI),
  rmsea = c(efa_5$RMSEA[1], efa_6$RMSEA[1]),
  bic = c(efa_5$BIC, efa_6$BIC)
)

print(fit_comparison)

# -------------------------------------------------------------------
# Estimate factor scores from both models
# -------------------------------------------------------------------

scores_5 <- factor.scores(
  item_pool,
  efa_5,
  method = "tenBerge"
)$scores

scores_6 <- factor.scores(
  item_pool,
  efa_6,
  method = "tenBerge"
)$scores

scores_5 <- as.data.frame(scores_5)
scores_6 <- as.data.frame(scores_6)

names(scores_5) <- paste0("factor5_", seq_len(ncol(scores_5)))
names(scores_6) <- paste0("factor6_", seq_len(ncol(scores_6)))

trait_data_scored <- bind_cols(
  trait_data,
  scores_5,
  scores_6
)

# -------------------------------------------------------------------
# Example criterion comparison:
# Does the six-factor solution improve integrity-related prediction?
# -------------------------------------------------------------------

if ("outcome_integrity" %in% names(trait_data_scored)) {
  model_5 <- lm(
    outcome_integrity ~ factor5_1 + factor5_2 + factor5_3 +
      factor5_4 + factor5_5,
    data = trait_data_scored
  )

  model_6 <- lm(
    outcome_integrity ~ factor6_1 + factor6_2 + factor6_3 +
      factor6_4 + factor6_5 + factor6_6,
    data = trait_data_scored
  )

  prediction_comparison <- bind_rows( glance(model_5) %>%
      mutate(model = "five_factor_prediction"),
    glance(model_6) %>%
      mutate(model = "six_factor_prediction")
  ) %>%
    select(model, r.squared, adj.r.squared, AIC, BIC, sigma)

  print(prediction_comparison)

  write_csv(
    prediction_comparison,
    "alternative_structure_prediction_comparison_r.csv"
  )
}

# -------------------------------------------------------------------
# Save outputs
# -------------------------------------------------------------------

write_csv(
  fit_comparison,
  "alternative_structure_fit_comparison_r.csv"
)

write_csv(
  trait_data_scored,
  "alternative_structure_items_scored_r.csv"
)

This workflow is useful because it moves the debate out of slogans. Instead of asking in the abstract whether five or six factors are “right,” the analyst can inspect covariance patterns, compare extraction solutions, examine whether an additional factor is interpretable, and test whether it improves prediction for relevant criteria.

Back to top ↑

Python: exploring hierarchical and alternative trait models

The Python example below illustrates how to inspect broad structure, compare candidate dimensionalities, compute component scores, model facets nested within broader domains, and compare broad-domain versus expanded-model prediction. It is a practical bridge between conceptual debates and actual data analysis.

# Beyond the Big Five: HEXACO, Hierarchies, and Alternative Structural Models
# Python workflow for comparing alternative personality structures

# Install packages if needed:
# pip install pandas numpy scikit-learn statsmodels

from pathlib import Path

import numpy as np
import pandas as pd
from sklearn.decomposition import PCA, FactorAnalysis
from sklearn.preprocessing import StandardScaler
import statsmodels.api as sm

# -------------------------------------------------------------------
# Load item-level personality data
# -------------------------------------------------------------------

# Expected structure:
# item1:item72 represent a broad personality item pool.
# Optional scored columns:
# - broad_domain
# - facet_1
# - facet_2
# - outcome_integrity
# - outcome_social_functioning

data_path = Path("alternative_structure_items.csv")
df = pd.read_csv(data_path)

print(df.head())
print(df.info())
print(df.describe(include="all"))

# -------------------------------------------------------------------
# Select and standardize item pool
# -------------------------------------------------------------------

item_columns = [f"item{i}" for i in range(1, 73)]
item_df = df[item_columns].dropna().copy()

scaler = StandardScaler()
item_scaled = scaler.fit_transform(item_df)

# -------------------------------------------------------------------
# Inspect broad dimensionality with PCA
# -------------------------------------------------------------------

pca = PCA(n_components=10)
components = pca.fit_transform(item_scaled)

explained_variance = pd.DataFrame(
    {
        "component": range(1, 11),
        "explained_variance_ratio": pca.explained_variance_ratio_,
        "cumulative_explained_variance": np.cumsum(
            pca.explained_variance_ratio_
        ),
    }
)

print(explained_variance)

# -------------------------------------------------------------------
# Compare five- and six-factor approximations
# -------------------------------------------------------------------

fa_5 = FactorAnalysis(n_components=5, random_state=42)
fa_6 = FactorAnalysis(n_components=6, random_state=42)

scores_5 = fa_5.fit_transform(item_scaled)
scores_6 = fa_6.fit_transform(item_scaled)

score_5_df = pd.DataFrame(
    scores_5,
    columns=[f"factor5_{i}" for i in range(1, 6)],
    index=item_df.index,
)

score_6_df = pd.DataFrame(
    scores_6,
    columns=[f"factor6_{i}" for i in range(1, 7)],
    index=item_df.index,
)

df = df.join(score_5_df, how="left")
df = df.join(score_6_df, how="left")

# -------------------------------------------------------------------
# Compare criterion prediction when an integrity outcome exists
# -------------------------------------------------------------------

def fit_ols(frame, outcome, predictors, model_name):
    model_df = frame[[outcome] + predictors].dropna()
    X = sm.add_constant(model_df[predictors])
    y = model_df[outcome]
    result = sm.OLS(y, X).fit()

    return result, {
        "model": model_name,
        "outcome": outcome,
        "n": int(result.nobs),
        "r_squared": result.rsquared,
        "adj_r_squared": result.rsquared_adj,
        "aic": result.aic,
        "bic": result.bic,
    }

comparison_rows = []
coefficient_tables = []

if "outcome_integrity" in df.columns:
    result_5, summary_5 = fit_ols(
        df,
        "outcome_integrity",
        [f"factor5_{i}" for i in range(1, 6)],
        "five_factor_integrity_model",
    )

    result_6, summary_6 = fit_ols(
        df,
        "outcome_integrity",
        [f"factor6_{i}" for i in range(1, 7)],
        "six_factor_integrity_model",
    )

    comparison_rows.extend([summary_5, summary_6])

    for result, model_name in [
        (result_5, "five_factor_integrity_model"),
        (result_6, "six_factor_integrity_model"),
    ]:
        coefficient_tables.append(
            pd.DataFrame(
                {
                    "model": model_name,
                    "term": result.params.index,
                    "estimate": result.params.values,
                    "standard_error": result.bse.values,
                    "p_value": result.pvalues.values,
                }
            )
        )

# -------------------------------------------------------------------
# Example hierarchical model:
# broad domain predicted by two facets
# -------------------------------------------------------------------

if {"broad_domain", "facet_1", "facet_2"}.issubset(df.columns):
    hierarchy_result, hierarchy_summary = fit_ols(
        df,
        "broad_domain",
        ["facet_1", "facet_2"],
        "facet_to_domain_model",
    )

    comparison_rows.append(hierarchy_summary)

    coefficient_tables.append(
        pd.DataFrame(
            {
                "model": "facet_to_domain_model",
                "term": hierarchy_result.params.index,
                "estimate": hierarchy_result.params.values,
                "standard_error": hierarchy_result.bse.values,
                "p_value": hierarchy_result.pvalues.values,
            }
        )
    )

# -------------------------------------------------------------------
# Save outputs
# -------------------------------------------------------------------

explained_variance.to_csv(
    "alternative_structure_pca_explained_variance_python.csv",
    index=False,
)

df.to_csv(
    "alternative_structure_items_scored_python.csv",
    index=False,
)

if comparison_rows:
    pd.DataFrame(comparison_rows).to_csv(
        "alternative_structure_model_comparison_python.csv",
        index=False,
    )

if coefficient_tables:
    pd.concat(coefficient_tables, ignore_index=True).to_csv(
        "alternative_structure_model_coefficients_python.csv",
        index=False,
    )

print("Alternative structure workflow complete.")

This kind of workflow is especially useful for structural comparison because it reveals how much explanatory work is being done by broad dimensions, how much is left for narrower subcomponents, and whether expanded models such as HEXACO improve prediction for the outcomes they are supposed to clarify.

Back to top ↑

GitHub repository

The companion GitHub repository provides reproducible research scaffolding for this article, including synthetic alternative-structure item data, documentation, validation materials, and multi-language workflows for comparing five-factor and six-factor structures, HEXACO-style expansion, hierarchical trait scoring, facet-level interpretation, bandwidth-fidelity tradeoffs, and structural model comparison.

Complete Code Repository

Access the full companion repository for this article, including reproducible analysis materials and multi-language code workflows for HEXACO, alternative personality structures, hierarchical trait models, factor comparison, facet scoring, and structural model interpretation.

View the Full GitHub Repository

Back to top ↑

Responsible interpretation

Alternative structural models require responsible interpretation because structural refinement can create the illusion of moral, diagnostic, or predictive certainty. HEXACO, hierarchy, facets, circumplexes, and networks can all improve personality science, but they can also be misused if their scores are treated as verdicts on a person rather than as measured indicators within a specific model.

The first principle is non-reduction. A person cannot be reduced to a Big Five profile, a HEXACO score, a Honesty-Humility estimate, a facet pattern, a circumplex location, a network node, or a factor loading. These tools describe patterns of response, behavior, or judgment under particular measurement conditions. They do not exhaust identity, culture, life history, moral character, trauma, disability, spirituality, relational context, institutional position, or future possibility.

The second principle is structural humility. A model is a map. It organizes personality space, but it does not become personality itself. Different maps may be useful for different purposes. A five-factor model may be adequate for broad description. HEXACO may be better for integrity-related questions. Facets may improve narrow prediction. Circumplex models may clarify interpersonal patterns. Network models may clarify interdependence. The model should fit the question.

The third principle is measurement discipline. A scale must be validated for the construct, population, language, and intended use. Reliability, factor structure, criterion validity, measurement invariance, and fairness should be examined at the level of interpretation. A model supported in one context should not be assumed to transfer automatically to another.

The fourth principle is moral caution. Honesty-Humility is important, but it is not a moral sentence. A score cannot prove honesty, dishonesty, corruption, exploitation, virtue, or character. It can support probabilistic research and careful interpretation, but it should not be used as a shortcut for judging people.

The fifth principle is proportional use. Alternative-structure workflows are suitable for professional education, research prototyping, psychometric demonstration, consulting support, organizational learning, coaching reflection, and reproducible workflow development. They are not standalone systems for hiring, promotion, termination, clinical diagnosis, educational placement, legal evaluation, insurance decisions, surveillance, relationship matching, or individual prediction. Any consequential use involving real people would require validated instruments, qualified interpretation, privacy safeguards, documented intended use, informed consent where appropriate, fairness and measurement-invariance analysis, and appropriate ethical and legal oversight.

Going beyond the Big Five should deepen personality science. It should not become a more elaborate language for unsupported labeling, exclusion, or gatekeeping.

Back to top ↑

Conclusion

Going beyond the Big Five does not mean abandoning the achievements of broad trait structure. It means recognizing that personality architecture remains an open scientific problem. HEXACO argues that six factors, especially with Honesty-Humility, better capture certain morally consequential forms of individual difference. Hierarchical models argue that the level of analysis matters and that broad domains must often be decomposed into narrower, more predictive units. Circumplex and network-based alternatives remind the field that structure can be represented in more than one way.

The deepest lesson is that personality structure is neither arbitrary nor final. Good models compress complexity without pretending to eliminate it. The Big Five remains powerful because it compresses large domains of personality variation into a stable and communicable map. HEXACO is powerful because it makes morally consequential content more visible. Hierarchies are powerful because they organize scale. Circumplexes are powerful because they show relational position. Networks are powerful because they reveal interconnection.

The best structural frameworks are therefore not those that silence debate, but those that make personality differences more intelligible, more measurable, and more theoretically fruitful. Going beyond the Big Five is valuable not because the Big Five failed, but because success in personality science always creates new questions about what the current map still leaves out.

A mature personality science should be comparative, layered, culturally attentive, and humble about its models. It should use broad domains where breadth is needed, HEXACO where honesty and exploitation matter, facets where specificity matters, circumplexes where interpersonal geometry matters, and networks where direct interdependence matters. No single structure is the whole person. But better structures can help us understand persons more carefully.

Back to top ↑

Further reading

  • Ashton, M.C. and Lee, K. (2020) Individual Differences and Personality, 4th edn. London: Academic Press.
  • Ashton, M.C., Lee, K. and de Vries, R.E. (2014) ‘The HEXACO Honesty-Humility, Agreeableness, and Emotionality factors: A review of research and theory’, Personality and Social Psychology Review.
  • DeYoung, C.G., Quilty, L.C. and Peterson, J.B. (2007) ‘Between facets and domains: 10 aspects of the Big Five’, Journal of Personality and Social Psychology, 93(5), pp. 880–896.
  • McCrae, R.R. (2009) ‘The Five-Factor Model of personality traits: Consensus and controversy’, in Corr, P.J. and Matthews, G. (eds.) The Cambridge Handbook of Personality Psychology. Cambridge: Cambridge University Press.
  • Roberts, B.W., Yoon, H.J., Magee, C.A., Soto, C.J., Wright, A.G.C. and Briley, D.A. (2022) ‘Personality psychology’, Annual Review of Psychology, 73, pp. 489–516.
  • Saucier, G. and Srivastava, S. (2015) ‘What makes a good structural model of personality? Evaluating alternatives to the Five-Factor Model’, in emerging review traditions on personality taxonomies.
  • Soto, C.J. and John, O.P. (2017) ‘The next Big Five inventory (BFI-2): Developing and assessing a hierarchical model with 15 facets to enhance bandwidth, fidelity, and predictive power’, Journal of Personality and Social Psychology, 113(1), pp. 117–143.

Back to top ↑

References

Back to top ↑

Scroll to Top