High-dimensionality personality structure in the natural language: Further analyses of classic sets of English-language trait-adjectives.

Abstract

Contemporary structural models of personality, like the Big Five, are rooted in natural-language lexicons in which socially important individual-differences concepts are sedimented. But key studies of these lexicons have been narrow in scope and mainly sought confirmatory evidence for one model of interest, rather than the maximum number of meaningful dimensions obtainable from the lexicon. Here, building on established methods for determining the appropriate number of factors, and comparing various methods of data-treatment and factor-rotation, analyses allowed higher-dimensionality structures to emerge from the same data. Factor-number-determination methods always recommended well more than 5 or 6 factors. In data using the largest 1,710-term set, we compared 18 candidate-structures derived from different method-combinations, identifying 15-, 21-, and 28-factor structures as most robust and promising. Among these, in a larger sample using a smaller 540-term set, the factors related to the 21-dimensional structure were most advantageously robust. Robustness-comparisons in an even larger sample with a 449-term subset of the original 1,710 converged on a similar number of factors. Though such a high-dimensionality model is slightly less robust than the Big Five across method-variations, we were able to confirm its clearly superior predictive capacity. And as comprehensiveness would imply, one can readily identify a low-dimensionality structure like the Big Five from within this higher-dimensionality structure, but one cannot generate this structure from the Big Five. A high-dimensionality structure can function as an improved scientific framework for cataloging trait variables and dimensions, including those that fall outside popular classifications involving only 5 or 6 factors.

Personality structure in east and west Africa: Lexical studies of personality in Maa and Supyire-Senufo.

“…The first goal in each of the two lexical studies was to define the optimal emic (indigenous) structure of character and personality terms. An initial comprehensive list of person PERSONALITY STRUCTURE IN EAST AND WEST AFRICA 15 descriptors in each language was refined to a usable number with frequency-of-use ratings from native speakers; the shortened list was administered as a survey to local native speakers who were asked to describe a well-known other; and analyses assessed the optimal emic structure using an approach defined by Saucier and Iurino (2019) to maximize model elaboration. A more conventional lexical analysis approach was used, however, to create factors for comparison with marker scales representing etic Big Two, Pan-Cultural Three, Big Five and Big Six models, to assess the degree of replication of these models in cultural contexts highly different from those typically studied.…”

Comparing predictive validity in a community sample: High–dimensionality and traditional domain–and–facet structures of personality variation

“…The data-sets examined by Saucier and Iurino (2019) generally excluded many person-descriptor categories, these being exclusions similar to those in classifications by Allport and Odbert (1936) and Norman (1967), as well as Angleitner, Ostendorf, & John, 1990). According to these now fairly typical exclusions made in lexicon-based studies on personality attributes, evaluative terms (e.g., Likeable, Evil, Weird), temporary states (e.g., Joyful, Afraid, Tired), social roles and effects (e.g., Wealthy, Fascinating, Intimidating), and physical and appearance descriptors (e.g., Attractive, Slender, Short) were excluded.…”

“…Aspects of the remaining 75% of the variance (much of which is not variation attributable to error) are drawn into the model only peripherally and inadvertently in measurement of subcomponents. Extracting a dozen or more additional factors appears to allow for an additional 10-15% of the overall variance in the adjective lexicon to be represented (Saucier & Iurino, 2019), which should widen predictive capacity.…”

“…In other words, they cannot be facets or subcomponents of the Big Five. More recent work indicates that when one utilizes parallel analysis to guide answers to the "how many factors" question, and ceases to rely entirely on orthogonal varimax rotation methods, a relatively robust structure of some 20 factors can be recovered in multiple data-sets with North American participants (Saucier & Iurino, 2019); this structure draws in part on many of the peripheral sources of variation identified earlier by Saucier and Goldberg.…”

An atlas of personality, emotion and behaviour

“…[6]. Taxonomies may be said to be parsimonious if they precisely describe a broad range of phenomena with the minimum number of independent variables [7]. The characteristics of a parsimonious model of personality are well understood [3,8,9], yet no 'periodic table' or grand theory of the whole person currently exists [3,8].…”