Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Lexical Personality Taxonomies and Trait Discovery

Abstract. Lexical personality taxonomy investigates the person-descriptive vocabulary available to linguistic communities and the statistical organization of judgments made using that vocabulary. This critical narrative research monograph reconstructs the history of the lexical hypothesis from nineteenth-century character measurement through Allport and Odbert, Cattell, the replication studies underpinning broad trait dimensions, and contemporary indigenous and multilingual psycholexical research. It examines the sampling of descriptor universes, semantic classification, parts of speech, participant and rater selection, factor extraction, multidimensional interpretation, and construct validation. Primary studies of English, Dutch, German, Croatian, Filipino, and other linguistic traditions demonstrate that factor solutions depend on variable-selection rules and that a common descriptive structure becomes more or less reproducible under different comparison criteria. The synthesis distinguishes descriptive taxonomies from causal trait theories and translated instrument results from indigenous lexical discovery. It develops an evidence-informed agenda for preregistered cross-linguistic sampling, semantic validation, open factor comparisons, computational corpus assistance, and ethical interpretation. It is a critical narrative synthesis of existing literature, without new participant data or an asserted meta-analysis.

Keywords: lexical hypothesis; personality lexicon; psycholexical taxonomy; trait discovery; personality descriptors; Allport and Odbert; factor analysis; Big Five; lexical six-factor structure; indigenous personality concepts; cross-linguistic replication; semantic annotation; construct validity; variable selection; cultural psychology; person perception; corpus linguistics; Artificial Era

The Scientific Object of Lexical Trait Discovery

Lexical personality taxonomy is the systematic identification and organization of recurrent person-descriptive meanings encoded in ordinary language. Its scientific object comprises the vocabulary used to describe relatively enduring differences between people, the rules by which terms are admitted to a research corpus, and the empirical covariation recovered when persons are rated with those terms. A taxonomy is a structured account of a descriptive domain; discovery concerns the inferential pathway from words to candidate trait dimensions. The distinction matters because a descriptor is a linguistic observation about social classification, whereas a trait dimension is a psychometric construction requiring independent validation Saucier & Goldberg (2001).

The fundamental question is how far culturally available descriptions of persons can identify reproducible patterns of individual differences. A lexical investigation begins with a language-specific sampling frame, proceeds through judgments about meaning and relevance, and ends with estimates of latent or component structure. Each stage creates uncertainty that must be carried into interpretation. A factor extracted from adjective ratings is evidence of patterned judgments about persons under specified measurement conditions. Convergent evidence from peers, instruments, populations, and external behavior is needed before the pattern receives a broader psychological interpretation Goldberg (1993) Cronbach & Meehl (1955).

This monograph treats lexical research as an evidence-generating research program rather than a predetermined inventory of five or six dispositions. Its specific focus is the construction and coverage of descriptor universes, history of taxonomy formation, methodological dependence of factor solutions, indigenous language comparisons, and reproducibility of trait discovery. Detailed exposition of the Big Five or HEXACO as finished personality models belongs to adjacent research objects; here their findings serve as comparative tests of discovery procedures. The synthesis draws on original psycholexical experiments and critical methods scholarship, without original human-participant data or a meta-analytic claim Ashton et al. (2004) De Raad et al. (2010).

The Lexical Hypothesis as a Research Heuristic

The lexical hypothesis proposes that socially consequential differences among persons tend to acquire names within the language of a community. Recurrent interpersonal observations, judgments, and expectations provide reasons to stabilize descriptive vocabulary. The resulting lexicon can therefore be sampled as a historical record of distinctions people have found useful. This premise justifies investigation; it does not establish that every frequent adjective denotes a distinct psychological mechanism or that a community possesses a complete vocabulary for every relevant disposition Saucier & Goldberg (2001) Goldberg (1993).

Three inferential moves organize the lexical program. Investigators first infer social salience from lexical availability, then infer clusters of shared descriptive meaning from ratings or semantic judgments, and finally consider whether those clusters track psychologically meaningful individual differences. Each move requires auxiliary assumptions about dictionaries, literacy, translation, social contexts, and rater access to behavior. Rare terms can represent important distinctions; common words may reflect moral evaluation rather than stable dispositions. Without those qualifications, lexical frequency and psychological importance can be conflated and the inferred domain becomes artificially narrow Saucier (1997) Kazmaly (2024).

A defensible lexical hypothesis is operational: define the population of terms, obtain an inclusive and auditable corpus, specify what counts as a disposition, and compare resulting structures when inclusion rules vary. This creates testable predictions about factor stability and cross-language recurrence. The hypothesis also supports competing, informative outcomes: a narrow adjective list may yield familiar factors, whereas nouns, evaluative descriptions, or verbs may expose additional dimensions. The scientific success of the approach lies in producing reproducible and interpretable observations, including findings that revise inherited trait models Saucier (2003) De Raad & Barelds (2008).

The Nineteenth-Century Problem of Character Description

Francis Galton's 1884 essay on the measurement of character anticipated the idea that language offered a large empirical register of recognizable personal qualities. His inquiry combined a methodological ambition to observe character with awareness that numerous verbal distinctions overlap. The historical contribution lies in treating apparently informal descriptions as data to be systematically enumerated, compared, and eventually measured. Its importance is conceptual and procedural: Galton did not deliver a modern factor taxonomy, a validated instrument, or an experiment demonstrating a definitive number of personality dimensions Galton (1884).

This antecedent arose amid a broader scientific effort to convert descriptions of persons into comparative measurements. Character words were readily available, but their meanings were embedded in ordinary social evaluation and unequal relations of observation. Terms concerning reliability, courage, sociability, and propriety could describe conduct in a particular institution as readily as general dispositions. The emerging methodological question was therefore not only which words existed, but whose speech practices supplied the words and whose behavior made a description plausible. Contemporary lexical research inherits these questions whenever a dictionary is treated as a neutral inventory Angleitner et al. (1990) Saucier & Goldberg (2001).

The route from historical character language to twentieth-century trait psychology required technical transformations. Dictionaries furnished lists rather than response matrices; psychological classification required decisions about synonymy, temporal stability, relevance to personality, and appropriate raters. Statistical methods then transformed many overlapping words into fewer dimensions, introducing further choices about correlations and rotations. Recognizing this sequence prevents retrospective projection of today's five- or six-factor labels onto earlier scholarship. The nineteenth-century precedent established a tractable discovery strategy, while the empirical content of the resulting factors remained an open question for subsequent investigators Cattell (1943) Goldberg (1990).

Allport and Odbert: Constructing the English Descriptor Universe

Gordon Allport and Henry Odbert's 1936 psycho-lexical study became a landmark because it attempted large-scale extraction and conceptual sorting of person-descriptive English words. Their list of 17,953 terms was organized into broad classes that separated relatively stable personal traits from temporary states, evaluative or character-related judgments, and other descriptive expressions. These distinctions anticipated a central difficulty of every lexical inventory: the psychological object depends on which meanings researchers treat as eligible personality descriptions. The original compilation was a classificatory resource, not the discovery of 17,953 independent traits Allport & Odbert (1936).

The value of this early research was breadth paired with an explicit attempt at theoretical discrimination. A word for a temporary mood may be used about a person who often experiences that mood; an evaluative label may summarize patterns of social behavior while also communicating condemnation. Thus even a carefully designed four-way scheme cannot remove ambiguity through definition alone. The design reveals why independent judges, linguistic examples, and sensitivity analyses are important when later researchers extract smaller sets from the original collection. Uncertainty at the classification stage becomes a source of uncertainty in every downstream factor solution Allport & Odbert (1936) Saucier (1997).

The study's historical position also merits careful interpretation. It sampled an English lexicon recorded in a particular social and historical setting, rather than the speech of all English users or the languages of the world. Dictionary coverage follows editorial conventions, social prestige, and patterns of literacy. A subsequent theory derived from the list inherits these constraints unless other linguistic communities and less formal registers are independently considered. Accordingly, the Allport–Odbert corpus remains an unusually influential starting point whose significance is as an inspectable sampling frame and stimulus for methodological innovation Angleitner et al. (1990) Saucier & Goldberg (2001).

Cattell and the Transformation of Words into Variables

Raymond Cattell's 1943 work transformed the psycho-lexical resource by grouping synonyms and opposites into 150 categories, supplementing those with ability and interest terms, and analyzing judgments about adults. The published description reports that one hundred adults were rated by people who knew them well, with correlations examined for clustered relationships. This design made the transition from a list of expressions to an organized matrix of measured interpersonal distinctions. It also exposed the inferential dependence of later structures on semantic aggregation before statistical reduction had begun Cattell (1943).

Aggregating words into categories is necessary for practical measurement but can silently determine the effective universe of traits. Synonyms may differ in formality, emotion, moral valence, and typical situations of application. Opposites can be asymmetric: describing someone as reserved may imply contexts or motives absent from the term outgoing. Combining such words because they appear adjacent in a dictionary or share an evaluative direction risks reducing psychologically relevant distinctions before the study begins. Cattell's contribution was a demanding empirical agenda, not a guarantee that one chosen compression of lexical data exhausts personality Cattell (1943) Kazmaly (2024).

Later psychometric research revisited both the scope and the statistical structure of early trait lists. A factor solution produced from selected variables can be internally coherent while differing markedly from solutions obtained from broader descriptors or another class of words. This is a general identification problem: item sampling, observer distributions, and mathematical extraction are jointly responsible for the observed pattern. The critical test is whether a hypothesized organization reappears after the investigators change plausible inclusion rules, recruit new raters, and replicate with independent lexical material Goldberg (1990) Saucier (1997).

From Peer Ratings to Recurrent Broad Dimensions

The mid-twentieth-century work associated with Ernest Tupes and Raymond Christal, originally circulated as a technical report in 1961 and published in journal form in 1992, analyzed relationships among ratings of selected trait terms across several samples. Their report describes recurrent broad dimensions named Surgency, Agreeableness, Dependability, Emotional Stability, and Culture. The study established the practical value of comparing solutions across differently constituted groups, acquaintanceship arrangements, and types of raters. Nevertheless, the stimulus terms had been selected as representatives of an already delimited domain, a consequential distinction from unrestricted word discovery Tupes & Christal (1992).

Warren Norman's 1963 peer-nomination investigation supplied another prominent instance of replicated higher-order organization. The reliability of a recognizable structure across raters or samples was an empirical achievement because it implied that the resulting dimensions were not wholly accidental products of one group of judges. Replication within a particular sampling framework, however, establishes stability conditional on the same framework. It cannot by itself determine whether the five-dimensional structure would also govern unfamiliar word classes, more inclusive language, or communities using different concepts of personhood Norman (1963) Saucier & Goldberg (2001).

This historical stage introduced a durable tension between parsimony and descriptive coverage. Broad factors provide economical summaries that work well for many purposes, but their apparent recurrence can be strengthened by using marker terms chosen for clear loading on those factors. Discovery requires a second question: what descriptions are excluded because they are semantically diffuse, socially evaluative, culturally specific, or difficult to rotate into simple structure? The next generation of lexicographic sampling and component analyses made that methodological question central to the field Goldberg (1990) Saucier (1997).

The Lexical Universe and the Sampling Frame

A psycholexical sampling frame is the explicitly defined universe of expressions from which a set of person descriptors is drawn. Dictionary headwords, spoken corpora, internet discourse, elicited descriptions, and curated adjective lists are alternative frames with different coverage properties. A comprehensive search within one dictionary is comprehensive only relative to that reference work. It can omit dialectal terminology, compounds, emerging expressions, and lexicalized social roles. The inference from factor analysis to a general personality taxonomy therefore depends on the degree to which the chosen frame covers the kinds of observations about persons that matter within the target population Angleitner et al. (1990) Saucier & Goldberg (2001).

Representativeness can be pursued through systematic extraction and a staged selection audit. Researchers should record source dictionaries, dates and editions, search criteria, excluded word classes, decisions about obsolete and regional terms, and the treatment of polysemy. Frequency thresholds improve respondent familiarity while privileging high-frequency social situations. Evaluative language may be overrepresented because people have strong reasons to praise or criticize others, whereas quieter dispositions receive fewer labels. A transparent item inventory allows later teams to reproduce the sampling process and to estimate the consequences of omitted domains instead of attributing all differences to population psychology Saucier (1997) De Raad & Barelds (2008).

The relation between a lexical universe and a trait measure is consequently directional. The lexicon generates candidate items, ratings estimate patterns, and further research tests whether a particular dimensional interpretation generalizes. Reversing the process by searching only for items that express an existing factor model turns the exercise into instrument construction or confirmation. Both enterprises have value, but their conclusions differ: a useful measure can be produced from selected markers without demonstrating coverage of the original universe of culturally available person descriptions Goldberg (1992) Saucier (2003).

Parts of Speech and the Form of Person Description

Adjectives have often dominated personality research because statements such as friendly or cautious are concise and compatible with rating scales. German lexicographic work demonstrated that a much wider inventory can be relevant, including person-type nouns and nouns for qualities. The Angleitner, Ostendorf, and John classification investigated thousands of terms and distinguished dispositions from roles, states, evaluations, and appearance attributes. Grammatical form matters because adjectives can denote gradable qualities while type nouns may package a social identity, stereotype, or narrative judgment into a single categorical expression Angleitner et al. (1990).

The descriptive contribution of verbs and multiword expressions can be equally important. A phrase such as keeps promises encodes an action pattern whose meaning varies with context, while an adjective such as trustworthy integrates expectation and evaluation. Dutch researchers deliberately sampled multiple grammatical categories and expressions, recovering dimensions beyond a conventional narrow adjective universe. This methodological expansion does not imply that every newly extracted factor is an enduring psychological mechanism. It establishes that vocabulary design influences the observable organization of interpersonal judgments De Raad & Barelds (2008) Saucier (2003).

Researchers should compare word classes empirically rather than treating grammatical categories as interchangeable. Independent item selections can be matched for familiarity, judged for dispositional relevance, and analyzed through comparable rating procedures. When different classes produce divergent factor structures, the differences can arise from syntax, target specificity, social evaluation, or genuinely distinct domains of person description. A discovery program must resolve these alternatives through additional validation instead of forcing all descriptions into an adjective-based structure selected in advance Saucier (2003) Kazmaly (2024).

Dispositions, States, Evaluations, and Social Roles

The distinction between enduring dispositions and transient states is useful but semantically porous. An expression may describe a temporary emotional episode on one occasion and a repeated personal tendency on another. Moral evaluations such as honest or selfish commonly refer to perceived patterns of behavior yet also express approval or blame. Social roles and reputation terms concern relationships with observers and institutions, not simply private attributes. A robust taxonomy must make these categories explicit and document disagreement among judges rather than projecting fixed psychological boundaries onto ordinary speech Allport & Odbert (1936) Angleitner et al. (1990).

Saucier's 1997 variable-selection experiment made this problem empirical. Fifteen judges classified five hundred familiar English person descriptors, and the research compared structure in self and acquaintance ratings. Restricting the variables to dispositions, or combining dispositions and states, supported a recognizable Big Five organization. Expanding the universe to additional person descriptors yielded distinct dimensions involving attractiveness and negative evaluation. The experiment therefore shows how categorical decisions about what counts as personality can change the factor count and content, even within one language Saucier (1997).

Interpretation requires further separation of three objects: a socially meaningful judgment, a regularity in how people describe others, and a dispositional explanation of observed action. A rating of attractiveness may produce stable covariance with other descriptions without measuring the same mechanism as persistence or curiosity. Likewise, negative-evaluation terms may capture shared social reputations as well as underlying tendencies. A broad lexical atlas can acknowledge these dimensions as legitimate phenomena of person perception while using external evidence to determine their status as psychological traits Cronbach & Meehl (1955) Saucier (1997).

Semantic Annotation, Polysemy, and Synonym Reduction

After lexical extraction, a research team must map expressions onto concepts without erasing distinctions that may carry psychological information. Human annotators commonly classify candidate items by familiarity, frequency of person-directed use, dispositional relevance, social evaluation, and semantic similarity. Agreement across annotators indicates the degree to which the operational definitions are usable, while disagreement identifies potentially informative ambiguities. Synonym reduction and opposite pairing lower respondent burden, but the outcome is conditional on who decides semantic equivalence and which contextual meanings the adjudication procedure retains Cattell (1943) Angleitner et al. (1990).

Lexical semantics supplies a further methodological warning. A familiar English adjective may encode an expected social setting, an affective response, and an evaluative stance at once. Kazmaly's 2024 corpus-assisted treatment of shy and outgoing demonstrates that folk personality terms can carry different semantic components even when used as apparent opposites in a questionnaire. Translation into another language can alter those components while preserving a superficial dictionary match. Treating these words as numerically symmetric indicators therefore requires empirical testing of interpretation rather than confidence in ordinary synonym lists Kazmaly (2024).

An auditable procedure can retain source examples, annotator rationales, alternate senses, and disputed category assignments alongside the final item matrix. Sensitivity analysis should then compare models fitted to unrestricted descriptors, narrow dispositions, and semantic clusters constructed through different admissible rules. This approach converts a seemingly preliminary editorial operation into a substantive scientific test. A taxonomy that survives alternative codings has stronger claim to lexical robustness than a solution that depends on one undocumented list of synonyms Saucier (1997) De Raad & Barelds (2008).

Self-Ratings, Acquaintance Ratings, and the Observation Problem

Lexical factor structures are usually inferred from people being rated by themselves or by observers. Self-reports have privileged access to experiences and intentions but are shaped by self-presentation, retrospective reconstruction, and the meaning respondents assign to each term. Acquaintance reports sample public conduct and reputation, yet depend on contact frequency, relationship type, and observer expectations. A single descriptive adjective can consequently express different information when supplied by a close friend, a co-worker, or the person being described Norman (1963) Saucier (1997).

The evidence base includes attempts to replicate dimensions across these sources. Goldberg's 1990 investigations compared comprehensive trait-term clusters with self- and peer-rating structures; Croatian and Dutch programs also considered multiple rating perspectives. Such convergence increases confidence that a factor captures more than one respondent's private theory. It still permits common social stereotypes or overlapping evaluative conventions to contribute to agreement. Independent behavioral indicators and longitudinal observations are useful because rating convergence alone does not identify the causal mechanism behind a descriptive pattern Goldberg (1990) Mlačić & Ostendorf (2005) De Raad & Barelds (2008).

Sampling the rated persons and the raters is as important as sampling the words. Restricting participants to university students can compress age, education, occupational context, and social diversity, potentially altering both covariance and item comprehension. Rater groups recruited from close relationships may have different knowledge of behavior than strangers assigned short observations. Researchers should describe target and judge demographics, relation length, recruitment, missingness, response styles, and the target sampling frame before interpreting lexical dimensions as population-general traits Tupes & Christal (1992) Gurven et al. (2013).

Factor Extraction, Rotation, and Dimensional Choice

Exploratory factor analysis and principal components analysis summarize covariation among rated descriptors, but they answer different mathematical questions. Principal components organize total observed variance into weighted composites, whereas common-factor approaches model covariance through latent variables under assumptions about uniqueness and measurement error. Either procedure can offer useful descriptive structure when its estimand is acknowledged. The data matrix, treatment of ordinal ratings, correlation type, and handling of missing values all influence extraction. Descriptive factors should not be interpreted automatically as independent causal modules of personality Cronbach & Meehl (1955) Goldberg (1990).

Rotation is used to make dimensions interpretable and can alter the allocation of adjectives among factors. Orthogonal rotations impose uncorrelated axes; oblique rotations allow correlated factors that may better represent social and psychological dependence. Decisions about the number of factors—whether based on eigenvalues, parallel analysis, cross-validation, interpretability, or theory—can yield distinct taxonomic resolutions. Ashton, Lee, and Goldberg explored one through seven dimensions within a single large English adjective dataset, documenting how new interpretive content becomes visible at different hierarchical levels Ashton, Lee, & Goldberg (2004).

A rigorous analytic report should disclose item preprocessing, correlation estimates, extraction and rotation procedures, factor retention rules, cross-loadings, and split-sample results. It should report alternative plausible solutions instead of presenting a favored number as if it were directly visible in the vocabulary. Comparative replication across item sets and samples is more informative than maximizing simple structure in one dataset. The objective is a tractable map of recurring descriptive distinctions whose stability can be evaluated under stated methodological transformations Saucier (1997) De Raad et al. (2010).

Goldberg's Broad English Adjective Replications

Lewis Goldberg's 1990 research provided a particularly influential comparison of factor patterns across large and differently organized English adjective collections. The first study analyzed 1,431 adjectives assembled into 75 clusters and assessed the similarity of resulting structures across multiple analytic procedures. A second study used 479 common terms grouped into 133 synonym clusters with both self and peer ratings; a third developed a more compact set of candidate markers. Repetition of recognizably similar dimensions across these designs offered substantial evidence that broad English trait organization could be reproduced beyond one small questionnaire Goldberg (1990).

The achievement must be specified at the correct inferential level. Goldberg recovered a descriptive structure from English lexical stimuli and rating conventions; this was strong evidence for replicable dimensions under those conditions. It did not prove that every culture selects the same concepts, that the dimensions are biologically indivisible, or that no finer-grained traits contain additional information. The later development of Big Five markers served valuable measurement and comparability goals, but the very success of markers can make subsequent findings appear more inevitable than the original open-ended discovery problem warrants Goldberg (1992) Goldberg (1993).

The methodological significance of the Goldberg studies is the emphasis on replication across multiple variable sets, analytic techniques, and perspectives. Such triangulation is necessary because a statistically stable factor may be partly induced by preprocessing. Broad dimensions become stronger candidates for generalizable taxonomies when they are repeatedly recognizable after changes to lexical sampling and rater design. A scientific synthesis should therefore report the underlying sampling and retention procedures together with the resulting names, making recurrence observable rather than relying solely on familiar domain labels Goldberg (1990) Saucier & Goldberg (2001).

Variable-Selection Experiments and the Stability of Five Factors

Gerard Saucier's 1997 study directly investigated whether inclusion criteria shape lexical taxonomies. The source abstract describes five hundred familiar English person descriptors classified by fifteen judges and rated in a sample of seven hundred adults, alongside acquaintance ratings for 201 targets. Factor analysis of narrower dispositional variables yielded a familiar Big Five pattern, while broader selections introduced additional recurrent evaluative and appearance-related dimensions. These are especially important results because the predictor being manipulated was the investigator's definition of the target vocabulary itself Saucier (1997).

The implication is methodological rather than a simple rejection of five factors. Under some specifications, five broad dimensions offer a stable description of enduring attributes. Under other defensible specifications, the observational domain expands and new correlations organize additional kinds of person perception. The lexical question then changes from asking which factor count is correct to asking which universe of attributes is being modeled and what decisions produced it. Recurrent structures with three broad factors were relatively insensitive to certain item-selection changes in this study, illustrating the possible robustness of a coarser level of description Saucier (1997).

These findings motivate a nested rather than winner-takes-all interpretation of taxonomy depth. The same judgments may support several useful resolutions: a few very broad axes, intermediate factors, and specific descriptive clusters. Each resolution can be evaluated for replication, semantic coherence, and criterion relations. Factors appearing only when social evaluations are admitted deserve interpretation as evaluative organization until additional evidence supports narrower trait meaning. The practice of reporting several theoretically meaningful dimensional levels preserves informative variation that a single favored solution could conceal Saucier (2009) Ashton, Lee, & Goldberg (2004).

Person-Type Nouns and the Limits of Adjective Taxonomies

The dominance of adjectives raises a basic coverage question because people often describe one another with type nouns—such as a helper, rebel, or show-off—that encode roles, evaluation, and behavioral expectations together. Saucier's 2003 investigation examined 372 highly familiar English personality type nouns with ratings from 607 participants. The published findings supported strong one- and two-factor representations and reported an interpretable, replicable eight-factor structure, while the Big Five received comparatively weak support from this grammatical class Saucier (2003).

This study isolates an important source of apparent theoretical disagreement. An adjective scale and a noun scale can both be psychometrically coherent while organizing distinct semantic content. Type nouns may emphasize social category membership, gendered reputations, appearance, or moral standing because their linguistic function differs from that of gradable adjectives. A poor replication of a five-factor adjective model in nouns can therefore indicate incomplete lexical coverage, method differences, or genuinely distinct descriptive regularities. Evaluating these interpretations requires comparable samples and a transparent mapping of meanings across grammatical classes Saucier (2003) Angleitner et al. (1990).

Trait discovery improves when investigators distinguish the grammatical distribution of a concept from the existence of a psychological construct. A social concept may be prominent in narratives and person-type nouns yet uncommon as a standardized adjective; another may have numerous near-synonymous modifiers but little behavioral distinctiveness. Research designs can pair noun and adjective sets through independent semantic annotation and then test shared and unique factor variance. Such work extends the lexical hypothesis from word counting toward a broader investigation of how linguistic communities classify persons Saucier & Goldberg (2001) Kazmaly (2024).

Unrestricted Dutch Lexical Sampling

De Raad and Barelds' 2008 Dutch study offers a strong counterexample to the assumption that a complete taxonomy emerges automatically from a conventional adjective list. The researchers assembled 2,365 person-descriptive items drawn from several word classes and multiword expressions, then obtained self and other ratings from 1,466 participants. Their analyses included principal components of raw and ipsatized data as well as split-sample investigation of factor invariance. They reported an eight-factor organization containing familiar Big Five content together with dimensions named Virtue, Competence, and Hedonism De Raad & Barelds (2008).

The result is consequential because it links the expanded factor set to expanded item inclusion within one language. It should be interpreted as a reproducible structure of that specified Dutch rating universe, rather than as evidence that every population possesses exactly eight basic traits. The distinction between a broad descriptive taxonomic factor and a psychological mechanism remains open. Nevertheless, the study supplies a direct empirical reason to preserve evaluative, motivational, and other terms until their contribution has been examined rather than discarding them solely because they sit outside preexisting trait inventories De Raad & Barelds (2008) Saucier (1997).

An unrestricted design also creates additional methodological burdens. Heterogeneous expressions differ in specificity, familiarity, grammatical structure, and social desirability; correlation patterns may reflect those differences as well as target characteristics. Split-sample replication tests internal robustness but cannot substitute for independent corpus construction or transport to another linguistic ecology. The Dutch program illustrates how comprehensive lexical research should combine inclusive sampling with careful analysis of dimensional meaning, stability across analytical procedures, and external indicators of the attributes being described De Raad & Barelds (2008) Cronbach & Meehl (1955).

Six-Dimensional Lexical Solutions Across Languages

Michael Ashton and colleagues' 2004 comparative work examined standard psycholexical studies of adjectives in Dutch, French, German, Hungarian, Italian, Korean, and Polish. The authors documented similar six-factor interpretations across these seven languages, including an honesty–humility dimension alongside variants of other familiar personality domains. Crucially, the study compared independently derived lexical material rather than simply administering the same translated five-factor inventory to everyone. Its evidence bears directly on the discovery question because new lexical communities had the opportunity to contribute somewhat different salient descriptions Ashton et al. (2004).

A complementary English investigation by Ashton, Lee, and Goldberg analyzed self-ratings on 1,710 trait adjectives from 310 participants. The authors compared structures at multiple extraction levels, reporting a recognizable five-factor solution and a six-factor configuration with a dimension resembling honesty–humility. A seven-factor extraction additionally brought religiosity into view. Those observations help clarify why the number of factors is inseparable from chosen resolution, variable population, and methodological conventions. They establish patterned relationships within a corpus, while leaving external behavioral interpretation for further validation Ashton, Lee, & Goldberg (2004).

Six-factor convergence is an important empirical challenge to any argument that English five-factor markers exhaust ordinary trait language. Yet the evidence also raises questions about item equivalence across cultures, the boundaries of moral vocabulary, and the degree of factor similarity required for a cross-language match. The same broad label can hide different configurations of words at a facet level. A convincing comparative taxonomy must therefore evaluate factor congruence, defining adjectives, and local semantic interpretations alongside the recurrence of a convenient numerical factor count Ashton et al. (2004) De Raad et al. (2010).

Inclusive Multilingual Evidence for a Broad Six

Saucier's 2009 synthesis analyzed broader person-descriptive variable selections from seven languages, including Chinese, English, Filipino, Greek, Hebrew, Spanish, and Turkish. The study reported six recurrent dimensions in these inclusive lexical materials and found that six-factor interpretations derived from wider selections captured content not fully represented by narrower conventional Big Five solutions. This is evidence about the consequences of breadth in the linguistic research domain, rather than merely an additional questionnaire test conducted on existing factors Saucier (2009).

The wider six-dimensional organization and the narrower six-factor structures described by Ashton and colleagues overlap in meaningful ways, although their precise semantic composition depends on the sampled terms. Distinct programs can converge in the importance of morally evaluative and socially regulated conduct without identifying identical indicator sets. The scientific response is to examine where independently derived dimensions map onto one another, where correspondence deteriorates, and which unique content makes an alternative taxonomy predict or explain external outcomes Saucier (2009) Ashton et al. (2004).

More inclusive vocabularies may improve descriptive completeness while increasing susceptibility to cultural reputational differences and rater-evaluation effects. Successful trait discovery therefore requires both a lexically faithful map and criterion-based comparison of useful interpretations. This broader program has no obligation to choose one factor count for every purpose. Its task is to determine under which extraction conditions, linguistic environments, and external criteria a given structure remains identifiable and scientifically informative Saucier & Goldberg (2001) Saucier (2009).

Indigenous Lexical Studies in Filipino and Croatian

An indigenous or emic lexical design derives the initial descriptor universe from the language and social categories being studied. Church, Reyes, Katigbak, and Grimm's 1997 Filipino research used an indigenous adjective list and investigated relations between locally derived components and the Big Five. The first study drew on self-ratings by Filipino high-school and university students of 861 Tagalog adjectives, permitting the investigators to observe descriptive structure before making external model comparisons. This sequence is a methodological safeguard against treating an imported English taxonomy as the complete set of possible indigenous distinctions Church et al. (1997).

Croatian research by Mlačić and Ostendorf similarly began with dictionary extraction, obtained semantic classifications from judges, and then administered selected dispositional adjectives to self and acquaintance raters. The reported work included 483 selected adjectives, self-ratings from 515 University of Zagreb students, and acquaintance ratings from 513 people. Its derived factors resembled familiar broad personality dimensions, illustrating that independently sourced lexicons can exhibit recognizable cross-language organization without presuming strict correspondence at every semantic level Mlačić & Ostendorf (2005).

Comparing the Filipino and Croatian examples highlights why indigenous extraction should precede transportability claims. Each study has its own lexicographic, demographic, and rating limitations, but both expose language-specific descriptors for examination. Local terms for obligations, interaction styles, and reputation should be interpreted by speakers familiar with their practical contexts. The scientific aim is to explain agreements and divergences among independently constituted lexical datasets, preserving locally meaningful content even when familiar broad axes can be recovered at a higher level Church et al. (1997) Saucier & Goldberg (2001).

Cross-Language Replication and the Three-Factor Dispute

De Raad and collaborators compared fourteen independently developed trait taxonomies from twelve languages and investigated solutions containing one through six factors. Their 2010 analysis concluded that only three dimensions met the study's criteria for full replication across the selected languages. This finding is central to lexical personality science because its inferential target differs from the recurrence of a five- or six-factor solution in selected linguistic families. Stronger universal criteria can yield fewer reliably shared dimensions when heterogeneous local item selections are compared De Raad et al. (2010).

The relationship between these results and earlier multilingual six-factor studies is a question about operationalization of replication. Studies may differ in the language set, sampling frames, common descriptor pool, rotation strategy, similarity indices, and thresholds for considering an axis equivalent. A dimension can be recognizable in several cultures yet fail a strict all-language replication requirement. Conversely, a very broad axis may survive because it captures substantial interpersonal evaluation shared across settings while leaving important local variation unexplained Ashton et al. (2004) De Raad et al. (2010).

A responsible synthesis therefore makes universal claims conditional on the design of cross-language comparisons. Three, five, and six dimensions may be differently resolved descriptions of partially overlapping covariance, rather than mutually exclusive inventories of discrete biological properties. Only detailed item-level alignment, sampling diversity, and independent replication can establish how much structure is stable across speech communities. The debate clarifies the distinction between descriptive recurrence, factor congruence, and stronger assertions about global universality of psychological mechanisms Saucier et al. (2014a) De Raad et al. (2010).

Very Broad Dimensions and the Big Two Convergence

Saucier and colleagues' 2014 cross-language investigation compared personality attributes derived from nine languages that included Asian, European, and African linguistic traditions. The researchers reported convergence for two very broad dimensions described as Social Self-Regulation and Dynamism. This result provides an example of a common denominator that can persist even when narrower factor organizations differ. Such broad dimensions describe major distinctions in social appraisal, agency, and interpersonal coordination, though their breadth necessarily leaves much specific personality vocabulary unrepresented Saucier et al. (2014a).

The Big Two hypothesis is best understood as a test of taxonomic resolution. Coarse solutions may optimize recurrence across disparate languages because they aggregate meanings that split apart at finer levels. A shared high-level factor can coexist with culture-specific midlevel categories and nuanced local trait vocabularies. The empirical issue is which information survives hierarchical aggregation and which predictive distinctions are lost. Replication alone does not answer the appropriate level for every research purpose; scientific utility must be assessed against the intended outcome and context Saucier et al. (2014a) Saucier (2009).

Related work by Saucier, Thalmayer, and Bel-Bahar examined relative ubiquity of human-attribute concepts across twelve mutually isolated languages. Its emphasis on concept prevalence broadens comparison beyond mechanically aligning the same English item labels. Studies of this kind identify which kinds of person descriptions reliably appear in diverse language systems and which may be disproportionately represented in the vocabulary of influential research traditions. Combined with factor analysis, concept-level comparison can distinguish sparse translation matches from genuinely convergent descriptive organizations Saucier et al. (2014b).

Translation, Semantic Transport, and Linguistic Equivalence

A translated item can preserve a familiar English label while altering its pragmatic force in the receiving language. Words such as shy, outgoing, conscientious, or humble may differ in implied situations, social desirability, and expectations about behavior. Kazmaly's 2024 linguistic analysis used corpus evidence and semantic explication to demonstrate the contextual complexity of two prominent extraversion-related terms. This work makes a concrete methodological contribution: before comparing scores across languages, investigators should establish what respondents are actually being asked to affirm Kazmaly (2024).

Semantic transport requires a sequence of checks. Bilingual translation and back-translation are useful safeguards, but they should be supplemented by local-language elicitation, cognitive interviews, independent usage examples, and evidence about item functioning. Structural similarities from translated inventories test whether the imported measurement architecture can be recovered under a new linguistic setting. Independently derived lexical studies ask an additional question: which attributes a community selects for its own person descriptions before foreign terms are introduced Church et al. (1997) Angleitner et al. (1990).

The two approaches are complementary when their estimands remain clear. Imported measures can facilitate direct numerical comparison and the study of measurement invariance. Indigenous taxonomies provide evidence about local domain coverage and the possibility that an existing instrument omits meaningful attributes. Agreement between the approaches strengthens cross-cultural inference; disagreement can identify translation difficulty, social context effects, or a genuinely different local lexical map. Reporting these alternatives improves interpretation and avoids treating one measurement tradition as universally exhaustive De Raad et al. (2010) Saucier & Goldberg (2001).

The Tsimane Evidence and Limits of Model Transport

Gurven and collaborators tested the Big Five Inventory among Tsimane forager–horticulturalists in the Bolivian Amazon. The original research describes a translated forty-four-item inventory administered to 632 adults, together with additional informant material, and reported that the conventional five-factor structure was not recovered in the same way as in many industrialized samples. This result is important for assessing the transport of a widely used questionnaire into a population differing in educational, cultural, and ecological characteristics from typical validation samples Gurven et al. (2013).

The Tsimane study should be integrated carefully into lexical taxonomy research because it used a translated instrument rather than exhaustive induction from the Tsimane language. Its findings challenge simple assertions about universal recoverability of a particular item structure, while leaving open what factors an indigenous comprehensive lexical project would discover. Differences in language comprehension, social organization, sample characteristics, and item relevance can contribute to poor structural fit. These possibilities do not erase the empirical result; they specify what it can and cannot decide Gurven et al. (2013) McCrae et al. (2005).

Other cross-cultural work using observer ratings has reported recognizable broad trait structures across numerous populations, including the fifty-culture project led by McCrae and Terracciano. These findings concern different instruments, samples, and comparison strategies. Their coexistence with the Tsimane evidence underscores why a universal statement must specify language coverage, population, elicitation method, and replication criterion. The strongest scientific conclusion is a structured program of comparative tests that separates measurement transportability from the independent existence and organization of local descriptors McCrae et al. (2005) De Raad et al. (2010).

From Descriptive Structure to Construct and Criterion Validity

Discovering a factor is a beginning of construct validation. A high-loading adjective cluster may reveal a stable judgment pattern without specifying why its members covary. Cronbach and Meehl's construct-validity framework requires evidence from a network of theoretical expectations, relations with other measures, and external phenomena. In lexical studies, relevant tests may involve peer agreement, behavioral observation, longitudinal consistency, criterion prediction, discriminant relations, and interpretive examination of the words themselves. No single factor-analysis result can substitute for these independent evidential connections Cronbach & Meehl (1955).

Criterion validity must also respect the nature of the criterion. Social acceptance, occupational performance, moral reputation, and observed cooperation are partly constituted by the same cultural expectations encoded in person descriptors. Correlation with an outcome may therefore reflect shared appraisal rather than a distinct latent causal tendency. Inclusive lexical models may predict some outcomes more effectively because they explicitly incorporate evaluation, while narrower disposition models can be preferable for questions about behavioral regularity. Comparing incremental prediction with suitable holdout samples can clarify this distinction Saucier (2009) Goldberg (1993).

Trait discovery becomes scientifically useful when it establishes where a descriptive taxonomy generalizes and where it changes. Independent rater sources, preregistered structural comparisons, multiple language samples, and outcomes obtained without the same word-based judgments all strengthen the evidential network. A factor's name should summarize the indicator content, not serve as a speculative explanation of it. This approach supports dialogue with personality development, culture, social cognition, and psychometrics while preserving the distinct inferential task of lexical taxonomic research Cronbach & Meehl (1955) De Raad & Barelds (2008).

Reproducibility, Reliability, and Analytic Identification

Reproducible lexical discovery requires more than obtaining a similar factor label from two analyses. A rigorous comparison examines whether independent teams can reconstruct the descriptor sampling frame, apply documented classification decisions, and recover semantically similar loadings in new participant samples. Internal replication by splitting a dataset is valuable, but participant-level replication retains the original language, instrument design, and recruitment context. External replication additionally changes investigators, speech communities, and measurement circumstances, testing the generalizability of the research procedure itself De Raad & Barelds (2008) De Raad et al. (2010).

Reliability and structural identification are related without being equivalent. A scale can have highly consistent responses because its words are near synonyms, while the larger domain from which it was selected remains incompletely characterized. Conversely, broader vocabularies may have lower apparent internal coherence precisely because they sample a more heterogeneous conceptual universe. Investigators should distinguish item-level stability, agreement between raters, factor congruence across samples, and predictive performance against independent criteria. Each quantity addresses a different aspect of the scientific claim made for a taxonomy Cronbach & Meehl (1955) Goldberg (1992).

A contemporary replication protocol should archive the original descriptor lexicon, semantic coding guide, item exclusions, analysis scripts, factor retention decisions, and anonymized metadata compatible with consent and privacy rules. Preregistered primary comparisons can be paired with explicitly exploratory alternative solutions. Robustness reports should show what happens when evaluative words, rare terms, multiword expressions, and different grammatical classes are included separately. Such transparency makes disagreement productive: competing results become explainable in terms of assumptions and study design, rather than reduced to labels of success or failure Saucier (1997) Saucier & Goldberg (2001).

Corpus Linguistics and Computational Trait Discovery

Large digital text corpora permit researchers to expand the discovery stage beyond traditional dictionaries. Candidate person descriptors can be retrieved through patterns of predication, dependency parsing, co-occurrence, and contextual semantic similarity. Computational methods may expose regional language, rapidly changing slang, and underrepresented multiword expressions. Their outputs nevertheless require validation because a language model can retrieve words about physical appearance, social categories, reputations, emotional episodes, and moral condemnation that satisfy surface linguistic patterns without constituting enduring dispositions Kazmaly (2024) Angleitner et al. (1990).

Distributional embeddings organize word use through statistical relationships in text. Their proximity can suggest synonyms, contextual neighborhoods, and candidate clusters for human annotation, but it is not itself a factor analysis of individual differences among persons. Corpus co-occurrence records what speakers say near particular words; a personality trait rating records what respondents attribute to particular targets. These are distinct data-generating processes. Conflating them would turn a map of linguistic associations into an unsupported psychological measurement claim. A sound computational pipeline uses semantic modeling to propose coverage and psychometric designs to test descriptive covariance Cronbach & Meehl (1955) Saucier (2003).

The Artificial Era extends the relevance of this distinction. Automated systems can assist multilingual corpus search, translation audit, and annotation, provided researchers evaluate training-domain bias, language coverage, semantic drift, and sensitivity to prompts. Descriptions of artificial agents as helpful, cautious, honest, or sociable are observable human judgments about their outputs and interactions; their factor structure need not be identical to the organization of human dispositions. Cross-system comparisons require separately defined behaviors, contexts, rating procedures, and validity criteria before any claims of analogous psychological organization are justified Kazmaly (2024) Saucier & Goldberg (2001).

Ethical and Cultural Limits of Person-Descriptor Sampling

Lexical research inherits social inequalities embedded in the sources from which it extracts terms. Dictionaries and public corpora are shaped by literacy, publication access, editorial judgment, and the relative visibility of groups. The same person descriptor can be used to assign stigma to a minority, celebrate conformity, or enforce status expectations. If researchers interpret every frequent negative label as a property of its target population, social stereotypes can be reified as personal essences. The scientific alternative is to examine whose language is sampled, who performs the rating, and what independent observations support each interpretation Allport & Odbert (1936) Saucier et al. (2014b).

Culture also shapes the practical meaning of apparently universal distinctions. Terms for obedience, assertiveness, interpersonal restraint, or diligence can refer to different situational demands under different economic and social arrangements. These differences may enter questionnaire responses without indicating fundamentally distinct human capacities. Indigenous vocabulary collection, collaboration with fluent speakers, and transparent decision making about translation can improve descriptive validity. Participation and consent procedures should account for participants' control over sensitive recordings, labels, and locally meaningful knowledge Church et al. (1997) Gurven et al. (2013).

Ethical integrity is part of empirical validity because stigmatizing or poorly understood items can produce patterned responses that investigators mistake for latent personality structure. Modern research should avoid releasing identifiable person ratings, protect sensitive demographic data, and document procedures for community consultation where appropriate. Scores derived from lexical factors should not be presented as clinical diagnoses or deterministic classifications of individuals. Taxonomies are probabilistic summaries of research observations, and their practical interpretation remains bounded by instrument validation, context, and purpose Cronbach & Meehl (1955) De Raad & Barelds (2008).

Integrating Competing Taxonomies Without Erasing Their Evidence

The empirical literature supports a layered picture of lexical personality structure. English adjective studies repeatedly recovered broad five-factor solutions; independent lexical studies across several languages produced informative six-dimensional organizations; some inclusive dictionaries yielded additional evaluative or motivational factors; and strict cross-language comparisons found fewer universally shared broad dimensions. These findings are about different descriptor universes, populations, and analytic criteria. The appropriate synthesis is to identify the common dimensions and document the conditions under which additional distinctions emerge Goldberg (1990) Ashton et al. (2004) De Raad et al. (2010).

A useful organizing framework separates four levels: the vocabulary available in a linguistic community, the rated semantic variables assembled by researchers, the covariance structure of person descriptions, and the external psychological constructs proposed to explain those descriptions. The first two levels are cultural and methodological; the third is statistical; the fourth requires independent behavioral and theoretical evidence. Disagreements about factor number often arise when claims at one level are used as direct proofs about another. Maintaining these distinctions allows a genuinely cumulative comparison of lexical taxonomies Saucier (1997) Cronbach & Meehl (1955).

Model pluralism has precise scientific content when each model is linked to a target and loss function. A two-factor model may capture very broad cross-cultural commonality, a five- or six-factor model may efficiently summarize recognizable behavioral descriptions, and a richer lexical atlas may preserve evaluative distinctions relevant to social judgment. These models can be assessed for coverage, replication, semantic coherence, predictive validity, and transportability. Such comparisons explain why different dimensional resolutions can be useful, while still permitting decisive empirical rejection of models that fail specified tests Saucier et al. (2014a) Saucier (2009).

A Research Agenda for Independent Lexical Discovery

An ambitious future design would begin with parallel indigenous lexicon collection in languages selected for structural, geographical, and sociocultural diversity. Participating teams would independently elicit ordinary descriptions, search representative corpora, and classify candidate terms by dispositional meaning, evaluative content, grammatical class, and context. Shared documentation would enable comparison of inclusion rules while preserving local concepts. Sampling plans should incorporate speakers with different ages, educational backgrounds, and social positions, reducing dependence on the university populations common in earlier studies Church et al. (1997) Mlačić & Ostendorf (2005) Saucier et al. (2014b).

Rating studies would then employ independent targets and multiple informant types, with preregistered comparisons of extraction methods and dimensional levels. Researchers could compare broad and narrow solutions, evaluate factor congruence after semantic alignment, and measure holdout predictive utility against behavioral criteria not formed from the same adjectives. A separate module would test translation effects by contrasting imported scales with indigenous terms. Where possible, longitudinal assessments would examine whether taxonomic positions persist beyond a particular observation period, without assuming stability from single-session ratings De Raad et al. (2010) Gurven et al. (2013).

Computational methods can scale lexical discovery while keeping human judgment at the center of meaning validation. Corpus retrieval systems should be evaluated for recall and precision across regional registers, and proposed semantic clusters should be audited with native speakers. Open analysis code, documented exclusions, preregistration, and careful data governance make the results inspectable. The central future objective is a map that distinguishes widespread descriptive regularities from language-specific concepts and tests what each contributes to independent understanding of personality, conduct, and social evaluation Kazmaly (2024) Saucier & Goldberg (2001).

Conclusion

Lexical personality taxonomies arise from a defensible scientific insight: ordinary language preserves many of the distinctions human communities repeatedly use when describing persons. From Galton's early observations and Allport and Odbert's descriptive inventory through Cattell's statistical reduction and later extensive factor studies, the research program has continually translated lexical material into candidate psychological structure. Its enduring contribution is a reproducible route for discovering dimensions that were not necessarily specified by a theory in advance, provided the route from words to variables is explicitly documented Galton (1884) Allport & Odbert (1936) Goldberg (1990).

The accumulated evidence demonstrates that discovery outcomes depend on the chosen descriptor universe, grammatical classes, semantic exclusions, rating sources, populations, and extraction procedures. Five-factor patterns recur under important conditions; six-factor and more inclusive solutions reveal additional content; and cross-language studies support different degrees of recurrence depending on the level and criterion of comparison. Each result specifies a part of the larger empirical problem. A scientific account must preserve both successful replication and systematic divergence, because those divergences identify where the descriptive domain changes Ashton et al. (2004) Saucier (1997) De Raad et al. (2010).

The most productive interpretation treats lexical structures as empirically constructed maps of human person description that can contribute to broader personality theory when supported by independent validity evidence. Future work should combine indigenous sampling, explicit semantic analysis, representative raters, robust statistics, independent behavioral criteria, and cross-language replication. Digital corpora and artificial semantic tools expand the scope of such inquiry while making linguistic and cultural provenance increasingly important. Lexical trait discovery advances psychological knowledge through the quality of its sampling, inference, and validation rather than through allegiance to one fixed number of dimensions Cronbach & Meehl (1955) Saucier et al. (2014b) Kazmaly (2024).

References

  • Allport, G. W., & Odbert, H. S. (1936). Trait-names: A psycho-lexical study. Psychological Monographs, 47(1, Whole No. 211), i–171. https://doi.org/10.1037/h0093360
  • Angleitner, A., Ostendorf, F., & John, O. P. (1990). Towards a taxonomy of personality descriptors in German: A psycho-lexical study. European Journal of Personality, 4(2), 89–118. https://doi.org/10.1002/per.2410040204
  • Ashton, M. C., Lee, K., Perugini, M., Szarota, P., de Vries, R. E., Di Blas, L., Boies, K., & De Raad, B. (2004). A six-factor structure of personality-descriptive adjectives: Solutions from psycholexical studies in seven languages. Journal of Personality and Social Psychology, 86(2), 356–366. https://doi.org/10.1037/0022-3514.86.2.356
  • Ashton, M. C., Lee, K., & Goldberg, L. R. (2004). A hierarchical analysis of 1,710 English personality-descriptive adjectives. Journal of Personality and Social Psychology, 87(5), 707–721. https://doi.org/10.1037/0022-3514.87.5.707
  • Cattell, R. B. (1943). The description of personality: Basic traits resolved into clusters. The Journal of Abnormal and Social Psychology, 38(4), 476–506. https://doi.org/10.1037/h0054116
  • Church, A. T., Reyes, J. A. S., Katigbak, M. S., & Grimm, S. D. (1997). Filipino personality structure and the Big Five model: A lexical approach. Journal of Personality, 65(3), 477–528. https://doi.org/10.1111/j.1467-6494.1997.tb00325.x
  • Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52(4), 281–302. https://doi.org/10.1037/h0040957
  • De Raad, B., & Barelds, D. P. H. (2008). A new taxonomy of Dutch personality traits based on a comprehensive and unrestricted list of descriptors. Journal of Personality and Social Psychology, 94(2), 347–364. https://doi.org/10.1037/0022-3514.94.2.347
  • De Raad, B., Barelds, D. P. H., Levert, E., Ostendorf, F., Mlačić, B., Di Blas, L., Hřebíčková, M., Szirmák, Z., Szarota, P., Perugini, M., Church, A. T., & Katigbak, M. S. (2010). Only three factors of personality description are fully replicable across languages: A comparison of 14 trait taxonomies. Journal of Personality and Social Psychology, 98(1), 160–173. https://doi.org/10.1037/a0017184
  • Galton, F. (1884). Measurement of character. The Fortnightly Review, 36, 179–185. https://galton.org/all.html
  • Goldberg, L. R. (1990). An alternative “description of personality”: The Big-Five factor structure. Journal of Personality and Social Psychology, 59(6), 1216–1229. https://doi.org/10.1037/0022-3514.59.6.1216
  • Goldberg, L. R. (1992). The development of markers for the Big-Five factor structure. Psychological Assessment, 4(1), 26–42. https://doi.org/10.1037/1040-3590.4.1.26
  • Goldberg, L. R. (1993). The structure of phenotypic personality traits. American Psychologist, 48(1), 26–34. https://doi.org/10.1037/0003-066X.48.1.26
  • Gurven, M., von Rueden, C., Massenkoff, M., Kaplan, H., & Lero Vie, M. (2013). How universal is the Big Five? Testing the five-factor model of personality variation among forager–farmers in the Bolivian Amazon. Journal of Personality and Social Psychology, 104(2), 354–370. https://doi.org/10.1037/a0030841
  • Kazmaly, A. (2024). The lexical semantics of personality words: On ‘shy’ and ‘outgoing.’ Lingua, 302, Article 103695. https://doi.org/10.1016/j.lingua.2024.103695
  • McCrae, R. R., Terracciano, A., & Personality Profiles of Cultures Project. (2005). Universal features of personality traits from the observer's perspective: Data from 50 cultures. Journal of Personality and Social Psychology, 88(3), 547–561. https://doi.org/10.1037/0022-3514.88.3.547
  • Mlačić, B., & Ostendorf, F. (2005). Taxonomy and structure of Croatian personality-descriptive adjectives. European Journal of Personality, 19(2), 117–152. https://doi.org/10.1002/per.539
  • Norman, W. T. (1963). Toward an adequate taxonomy of personality attributes: Replicated factor structure in peer nomination personality ratings. The Journal of Abnormal and Social Psychology, 66(6), 574–583. https://doi.org/10.1037/h0040291
  • Saucier, G. (1997). Effects of variable selection on the factor structure of person descriptors. Journal of Personality and Social Psychology, 73(6), 1296–1312. https://doi.org/10.1037/0022-3514.73.6.1296
  • Saucier, G. (2003). Factor structure of English-language personality type-nouns. Journal of Personality and Social Psychology, 85(4), 695–708. https://doi.org/10.1037/0022-3514.85.4.695
  • Saucier, G. (2009). Recurrent personality dimensions in inclusive lexical studies: Indications for a Big Six structure. Journal of Personality, 77(5), 1577–1614. https://doi.org/10.1111/j.1467-6494.2009.00593.x
  • Saucier, G., Thalmayer, A. G., Payne, D. L., Carlson, R., Sanogo, L., Ole-Kotikash, L., Church, A. T., Katigbak, M. S., Somer, O., Szarota, P., Szirmák, Z., & Zhou, X. (2014). A basic bivariate structure of personality attributes evident across nine languages. Journal of Personality, 82(1), 1–14. https://doi.org/10.1111/jopy.12028
  • Saucier, G., Thalmayer, A. G., & Bel-Bahar, T. S. (2014). Human attribute concepts: Relative ubiquity across twelve mutually isolated languages. Journal of Personality and Social Psychology, 107(1), 199–216. https://doi.org/10.1037/a0036492
  • Saucier, G., & Goldberg, L. R. (2001). Lexical studies of indigenous personality factors: Premises, products, and prospects. Journal of Personality, 69(6), 847–879. https://doi.org/10.1111/1467-6494.696167
  • Tupes, E. C., & Christal, R. E. (1992). Recurrent personality factors based on trait ratings. Journal of Personality, 60(2), 225–251. (Original technical report circulated 1961). https://doi.org/10.1111/j.1467-6494.1992.tb00973.x