Written entirely by Claude, as a working record for agents more than for human readers: every number with its source file, and every claim since withdrawn. Some facts here may be out of date; where it disagrees with the post, trust the post. For the readable version, see the companion.

Big Five reference bibliography

Full citations for the lexical-tradition works behind the zoo's traits -- Goldberg's 100 unipolar markers as the primary set, the SAPA/TDA adjective bank as the secondary set, and the ancestry and alternatives around them.

This is the reference page: one entry per work, with a full citation, what it established, what this project uses it for, and a verification marker. The narrative is elsewhere -- History of the Big Five and Goldberg's factor labels and the 100 unipolar markers.

Two of these are load-bearing rather than background. Goldberg (1992) is the source of the 100 unipolar markers that are the zoo's primary traits: qwen35/traits_primary.json is a 100-entry list of {trait, factor, keyed} objects, 20 per factor, keyed plus or minus. Condon, Coughlin and Weston (2022) is the source of the secondary traits: qwen35/traits_secondary.json holds the 40 drawn adjectives (all labelled factor Lexicon, keyed +), and qwen35/traits_secondary_provenance.json records the exact provenance, including the source URL, a SHA-256 of the file pulled, and the filter counts. How many of the 40 reached the zoo is a question for the zoo pages, not this one. Everything else on this page is context for those two.

Verification markers: [verified] = citation details confirmed online on 2026-09-07; [not verified] = written from the project's own files or from general reference and not confirmed against a publisher record in this pass. No entry below was read in full.

The lexical tradition, in order

Allport & Odbert (1936) [not verified]

Allport, G. W., & Odbert, H. S. (1936). Trait-names: A psycho-lexical study. Psychological Monographs, 47(1), i-171.

The founding inventory: an extraction of every person-descriptive term in an unabridged English dictionary, sorted into categories. It is the reason the whole tradition is called lexical -- the hypothesis that the socially important individual differences are all encoded in the natural language. Everything below is a factor analysis of some descendant of this list, and the trait adjectives the zoo trains on are, at several removes, entries in it. Cited as an antecedent by the Condon et al. dataset.

Tupes & Christal (1961) [not verified]

Tupes, E. C., & Christal, R. E. (1961). Recurrent personality factors based on trait ratings (USAF ASD Technical Report No. 61-97). Lackland Air Force Base, TX: U.S. Air Force. Reprinted as Tupes, E. C., & Christal, R. E. (1992), Journal of Personality, 60(2), 225-251.

The first recovery of five recurrent factors across eight samples of trait ratings. Published as an Air Force technical report and largely invisible for two decades, which is why the reprint exists and why the model is often dated to Norman or to Goldberg rather than to here.

Norman (1963) [not verified]

Norman, W. T. (1963). Toward an adequate taxonomy of personality attributes: Replicated factor structure in peer nomination personality ratings. Journal of Abnormal and Social Psychology, 66(6), 574-583.

The replication that made the five stick, and the source of the phrase "Norman's Big Five". Peer-nomination ratings rather than self-report.

Goldberg (1990) [not verified]

Goldberg, L. R. (1990). An alternative "description of personality": The Big-Five factor structure. Journal of Personality and Social Psychology, 59(6), 1216-1229.

The systematic re-derivation of the five from comprehensive adjective sets, and the paper that fixed the modern names -- including Intellect rather than Openness for the fifth factor, which is the labelling this project's factor pages follow. The reason it matters here is that the zoo's factor labels are Goldberg's, not Costa and McCrae's, and the two traditions disagree about that fifth factor. See Goldberg's factor labels and the 100 unipolar markers.

Goldberg (1992) [verified] -- the zoo's primary trait set

Goldberg, L. R. (1992). The development of markers for the Big-Five factor structure. Psychological Assessment, 4(1), 26-42.

Over 1,000 college students completed adjective-anchored bipolar rating scales and unipolar adjective sets; the outcome was a set of 100 unipolar terms highly robust across diverse samples of self- and peer-description. These are the zoo's primary traits -- see The 100 Goldberg primary traits -- 20 per factor, carried in qwen35/traits_primary.json. Keying is not an even 10/10 split across the board: counting the keyed field in that file gives 10 positive and 10 negative for Agreeableness, Conscientiousness, Extraversion and Intellect, but 6 positive and 14 negative for Emotional Stability. The same asymmetry shows up downstream in qwen35/analysis/geometry_stage1.json#unwhitened_loo_keying (EmotionalStability.n_pos 6, n_neg 14) and is remarked on in qwen35/build_spider_page.py. Whether the imbalance comes from Goldberg's published marker list or from how it was transcribed into traits_primary.json was not checked; either way, any per-factor statistic that assumes balanced keying has to account for it. The list runs Extraverted, Talkative, Assertive, Verbal, Energetic, Bold, ... on the Extraversion-positive side, and Introverted, Shy, Quiet, ... on the negative.

Two things about the instrument to keep in mind when reading the zoo's judged evaluations. First, these markers are administered to a person as self-ratings, which is not what the project does with them. Second, the marker senses are the technical ones: qwen35/regen_goldberg_senses.py documents three markers that a teacher model rejected on the wrong sense, where Goldberg's instrument means verbally fluent, punctual-as-a-habit, and so on rather than the everyday reading.

Goldberg (1993) [not verified]

Goldberg, L. R. (1993). The structure of phenotypic personality traits. American Psychologist, 48(1), 26-34.

The consolidation and public statement of the case for the five-factor structure, addressed to psychology at large rather than to specialists. The standard citation when a paper needs one reference for "the Big Five".

Goldberg (1999) [not verified] -- IPIP

Goldberg, L. R. (1999). A broad-bandwidth, public domain, personality inventory measuring the lower-level facets of several five-factor models. In I. Mervielde, I. Deary, F. De Fruyt, & F. Ostendorf (Eds.), Personality Psychology in Europe (Vol. 7, pp. 7-28). Tilburg, The Netherlands: Tilburg University Press.

The founding paper of the International Personality Item Pool. Its significance for anything computational is that it put a full, validated, public-domain item bank into circulation with no licence to negotiate, which is why almost every LLM-personality paper that administers a questionnaire administers an IPIP one.

Goldberg (2006) [verified] -- bass-ackwards

Goldberg, L. R. (2006). Doing it all bass-ackwards: The development of hierarchical factor structures from the top down. Journal of Research in Personality, 40(4), 347-358.

The method paper for reading a factor hierarchy top-down: extract one component, then two, then three, and track how each level splits into the next, rather than picking a single k and defending it. Directly relevant to this project's dimensionality problem, where the question is not only how many factors the adapter cloud has but how the coarse structure decomposes into the finer -- see Weight-space geometry of the 134-adapter zoo and Factor-analysis methods reference.

The alternative traditions

Costa & McCrae -- the NEO inventories [not verified]

Costa, P. T., Jr., & McCrae, R. R. (1985). The NEO Personality Inventory manual. Odessa, FL: Psychological Assessment Resources.

Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO-PI-R) and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources.

The questionnaire tradition rather than the lexical one, and the source of the OCEAN naming and of the six facets per factor structure. This is the branch Persona Cartography builds its constitutions on: each of its ten OCEAN adapters is specified by six NEO-PI-R facets, each with three defining adjectives, crossed with two framings to give a twelve-item constitution. So the two source papers and this project sit on different branches of the same tree -- Persona Cartography on Costa and McCrae's five factors and their facets, this project on Goldberg's 100 adjective markers -- and the fifth factor is named differently in each. That mismatch is why the zoo's factor recovery is judged against Goldberg marker targets rather than against OCEAN.

Saucier & Goldberg (2001) -- cross-language [verified]

Saucier, G., & Goldberg, L. R. (2001). Lexical studies of indigenous personality factors: Premises, products, and prospects. Journal of Personality, 69(6), 847-879.

Reviews lexical studies in English and twelve other languages and asks which structures recur across them. The relevant point for this project is that the five are an empirical regularity of one family of lexicons under one family of methods, not a law -- which is the right prior for asking whether a set of weight deltas trained on English adjectives should be expected to reproduce them.

Related, and cited alongside it in the literature: Saucier, G., Hampson, S. E., & Goldberg, L. R. (2000). Cross-language studies of lexical personality factors. [not verified]

Ashton & Lee -- HEXACO [verified, with a date correction]

Ashton, M. C., & Lee, K. (2005). A defence of the lexical approach to the study of personality structure. European Journal of Personality, 19(1), 5-24.

Ashton, M. C., & Lee, K. (2007). Empirical, theoretical, and practical advantages of the HEXACO model of personality structure. Personality and Social Psychology Review, 11(2), 150-166.

Correction. This work was requested as "Ashton & Lee 2004 HEXACO". The lexical-approach defence is 2005, not 2004, in European Journal of Personality 19, 5-24; the HEXACO advantages paper is 2007. The 2004 HEXACO citation that circulates is usually Lee, K., & Ashton, M. C. (2004), Psychometric properties of the HEXACO Personality Inventory, Multivariate Behavioral Research, 39(2), 329-358 [not verified] -- note the reversed author order. Whichever is meant, the substance is the same: cross-language lexical work recovers a sixth factor, Honesty-Humility, that the Big Five absorbs into Agreeableness. It is the standing empirical argument that five is a result rather than a truth, and therefore the reference point for anything the zoo's geometry finds at k other than 5.

The secondary trait set

Condon, Coughlin & Weston (2022) [verified]

Condon, D. M., Coughlin, J., & Weston, S. J. (2022). Personality trait descriptors: 2,818 trait descriptive adjectives characterized by familiarity, frequency of use, and prior use in psycholexical research. Journal of Open Psychology Data, 10(1), 1-9. Online database at https://pie-lab.github.io/tda/.

The direct source of the zoo's secondary traits -- 40 adjectives in qwen35/traits_secondary.json. From qwen35/traits_secondary_provenance.json, which is the authoritative record:

• Source file https://raw.githubusercontent.com/pie-lab/tda/master/masterkey.tab, SHA-256 904533eb01b5650bbcec93104737e7050fcac9438591dea4d664f0fdf24a0190, cached at qwen35/tda_masterkey.tab.
• Source description as recorded in that file: "Condon, D. M., Coughlin, J., & Weston, S. J. (2022) -- 2,818 trait-descriptive adjectives, master key of the pie-lab/tda item bank; each adjective appears once per response form (A/B)".
• Funnel: 5,636 rows in the master key, 2,818 unique adjectives, 2,408 after the familiarity filter, 2,303 after excluding anything overlapping the primary set.
• The familiarity floor is derived rather than chosen: wordfreq.word_frequency at or above 1.197e-08, the 5th percentile of that frequency over the 100 primary-set words. 410 adjectives were dropped by it. Five of the 100 primary words themselves fall below the threshold (Untalkative, Unenvious, Unexcitable, Imperceptive, Uninquisitive), which is the honest cost of deriving the floor from the primary set.
• Primary-set exclusion is by exact lowercase match or Porter stem match; 105 adjectives were excluded that way.
• This is draw 3; see The 34 Lexicon traits and how they were drawn and The two discarded secondary draws. Draws 1 and 2 are retained as qwen35/traits_secondary_DISCARDED_draw1.json and ..._draw2.json. Three filters were removed in draw 3, and the provenance file says why: the part-of-speech and dispositional filters were redundant because the source is adjective-only and trait-descriptive by construction, and the WordNet-gloss sieve "was miscalibrated (deleted 'virtuous', kept 'lustful'); no semantic filter replaces it".

The dataset's own lineage is worth stating because it makes the secondary traits continuous with the primary ones: it contains all 1,710 adjectives previously studied by Goldberg (1982) and draws on Allport & Odbert (1936) and Norman (1967).

Verification note

Citation details for Goldberg (1992), Goldberg (2006), Saucier & Goldberg (2001), Ashton & Lee (2005/2007) and Condon et al. (2022) were confirmed by web search on 2026-09-07 and are marked [verified]; the searches returned volume and page numbers consistent with the citations above. The remaining entries are marked [not verified]: they are standard citations written out here from the project's own files and general reference, and were not confirmed against a publisher record in this pass. No work on this page was read in full. The two that matter most -- Goldberg (1992) and Condon et al. (2022) -- are used here only through the derived artefacts qwen35/traits_primary.json and qwen35/traits_secondary_provenance.json, whose contents are quoted directly and which are the citable source for anything the blog post says about the trait lists.

Neighbours

Other literature notes: Factor-analysis methods reference, LoRA and merging reference, Open Character Training (Maiya et al., 2025), Persona Cartography (Baines et al., 2026), and the 2026-09-09 reading note on SliderSpace and Gradient Atoms, Paper reading, 2026-09-09 - SliderSpace and Gradient Atoms.