CoChat
Sign Up
Cited by 648Personality Traits and PsychologyMental Health via WritingAuthorship Attribution and Profiling

Personality in 100,000 Words: A large-scale analysis of personality and word use among bloggers

Open access
Yarkoni, Tal·2010·Journal of Research in Personality
DOI 10.1016/j.jrp.2010.04.001PMC PMC2885844
Open this paper in CoChat — free account

Abstract

Previous studies have found systematic associations between personality and individual differences in word use. Such studies have typically focused on broad associations between major personality domains and aggregate word categories, potentially masking more specific associations. Here I report the results of a large-scale analysis of personality and word use in a large sample of blogs (N=694). The size of the dataset enabled pervasive correlations with personality to be identified for a broad range of lexical variables, including both aggregate word categories and individual English words. The results replicated category-level findings from previous offline studies, identified numerous novel associations at both a categorical and single-word level, and underscored the value of complementary approaches to the study of personality and word use.

AI snapshot

Written from the paper’s full text.

TL;DR

This study analyzed personality and word use in 694 blogs, finding pervasive associations between Big Five traits/facets and both LIWC categories and individual words. It replicated prior offline findings while showing that broad category-level analyses can mask or contradict narrower facet- and word-level associations.

Key findings

  • Of 330 Big Five–LIWC category correlations, 145 (43.9%) were significant at p < .05 and 49 (14.8%) at p < .001; the study replicated 15 of 30 correlations from Pennebaker and King (1999) and 15 of 24 from Hirsh and Peterson (2009).
  • Facet-level analyses found 28.8% of 1,980 facet–LIWC correlations significant at p < .05 and 152 (7.7%) at p < .001, with substantial heterogeneity: Extraversion and Openness showed 42 and 56 heterogeneous categories, respectively.
  • Openness had far more word-level correlates than other traits: it correlated significantly (p < .001) with 393 words, whereas Neuroticism, Extraversion, and Conscientiousness each correlated with fewer than 30 words.
  • Word-level analyses revealed opposite-direction associations within categories, e.g., Agreeableness correlated positively with affection words like “loves,” “love,” and “hug” (all ρ ≥ .2, p < .001) but negatively with “porn,” “gay,” and “fuck” (all ρ < −.21, p < .001).

Methods

  • Naturalistic sample of 694 blogs (524 female; mean age = 36.2 years, range 18–78, SD = 11.7), recruited via Google Blog Search and email; blogs averaged 115,423 words and spanned a mean of 23.9 months.
  • Personality was measured with the 50-item IPIP NEO-FFI for Big Five traits (N = 576) and the 300-item IPIP NEO-PI-R for 30 facets (N = 431); language was analyzed with LIWC 2001 across 66 categories and with normalized counts of the 5,608 most frequent words.
  • Analyses used Spearman rank correlations; word-level analyses included only blogs with ≥50,000 words (N = 406), used p < .001 thresholding, and applied split-half reliability and False Discovery Rate checks.

Limitations

  • The paper states likely selection bias: only a small proportion of bloggers publicly display email addresses and only a fraction agreed to participate, potentially restricting personality variance and ruling out direct comparison with the general population.
  • The paper states that many correlations were modest (largest LIWC–Big Five correlation = .23; mean absolute significant correlation = .14) and that all language variables were simple word counts, ignoring context, irony, images, and higher-order semantics.
  • Evident from the text: the sample was three-quarters female, included only Google Blogger blogs, relied on self-report personality, and was correlational, so causal and generalizability claims are constrained.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.jrp.2010.04.001

Cites

44 of 44 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 45 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Artistic Interests and Emotionality failed to show the robust positive associations with Articles and Prepositions demonstrated by the other facets and total Openness scores.

    “Conversely, Artistic Interests and Emotionality failed to show the robust positive associations with Articles and Prepositions demonstrated by the other facets and total Openness scores.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.
  • Findingown data

    Cooperation correlated with 'unusual' (ρ = .3).

    “Unanticipated associations included correlations between Self-Consciousness and ‘sizes’ (ρ = .27), Intellect and ‘against’ (ρ = .37), Trust and ‘summer’ (ρ = .31), and Cooperation and ‘unusual’ (ρ = .3), to name a few (all of these examples survived even an extremely conservative Bonferroni correction for 5,000 comparison—i.e., p < .00001).”
    Quote in the paperNot a substantial claim
  • Findingown data

    The analysis revealed significant heterogeneity for 44% of trait/category combinations.

    “The analysis revealed significant heterogeneity for a large proportion of trait/category combinations (44%)5.”
    Quote in the paperNot a substantial claim
  • Findingown data

    Agreeableness showed a small but statistically significant positive correlation with use of Sexual words (ρ = .08, p < .05).

    “For example, Agreeableness showed a small but statistically significant positive correlation with use of Sexual words (ρ = .08, p < .05), and Extraversion, a trait often associated with increased incentive facilitation and agentic behavior (Depue & Collins, 1999), correlated negatively with several categories reflecting goal orientation and work-related achievement (Occupation, Job/work, and”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.
  • Findingown data

    Self-Consciousness correlated with 'sizes' (ρ = .27).

    “Unanticipated associations included correlations between Self-Consciousness and ‘sizes’ (ρ = .27), Intellect and ‘against’ (ρ = .37), Trust and ‘summer’ (ρ = .31), and Cooperation and ‘unusual’ (ρ = .3), to name a few (all of these examples survived even an extremely conservative Bonferroni correction for 5,000 comparison—i.e., p < .00001).”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. 9
  • May combine two claims. 35
  • The quote only partly states this claim. 1

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    The full sample contained 694 blogs (524 female; mean age = 36.2 years, range = 18-78, sd = 11.7).

    “In total, the full sample contained 694 blogs (524 female; mean age = 36.2 years, range = 18 -78, sd = 11.7), though the actual sample size was smaller for some analyses because not all participants provided personality data (see below).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Of the 330 different correlation coefficients between the Big Five traits and the 66 LIWC categories, 145 (43.9%) were statistically significant at p < .05, and 49 (14.8%) were statistically significant at p < .001.

    “Of the 330 different correlation coefficients between the Big Five traits and the 66 LIWC categories, 145 (43.9%) were statistically significant at p < .05, and 49 (14.8%) were statistically significant at p < .001.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    The results directly replicated previous findings at a rate substantially greater than chance; specifically, the present study successfully replicated 15 of 30 correlations between Big Five dimensions and LIWC categories reported by Pennebaker & King (1999), and 15 of 24 Big Five correlations reported by Hirsh & Peterson (2009).

    “Moreover, the results directly replicated previous findings at a rate substantially greater than chance; specifically, the present study successfully replicated 15 of 30 correlations between Big Five dimensions and LIWC categories reported by Pennebaker & King (1999), and 15 of 24 Big Five correlations reported by Hirsh & Peterson (2009).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Pennebaker & King (1999)

      Not checked in the free run.

    • cites Hirsh & Peterson (2009)

      Not checked in the free run.

    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    Openness correlated significantly with 393 individual words, whereas Neuroticism, Extraversion, and Conscientiousness each correlated with fewer than 30 words.

    “In particular, Openness correlated significantly (p < .001) with 393 words, whereas Neuroticism, Extraversion and Conscientiousness all correlated with fewer than 30 words.”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data

    Of the 1980 different correlation coefficients between the 30 facets and the 66 LIWC categories, (28.8%) were statistically significant at p < .05, and 152 (7.7%) were statistically significant at p < .001.

    “Of the 1980 different correlation coefficients between the 30 facets and the 66 LIWC categories, (28.8%) were statistically significant at p < .05, and 152 (7.7%) were statistically significant at p < .0014.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    Openness to Experience was negatively correlated with 37 of the 66 LIWC categories, and positively correlated with only 4 categories.

    “Most notably, Openness to Experience, which one might have expected to correlate positively with categories associated with emotional, intellectual, or sensory experience, was negatively correlated with 37 of the 66 LIWC categories, and positively correlated with only 4 categories.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    These dissociations could not be explained by heterogeneity in the Openness facets themselves; intercorrelations between the six facets were all directionally positive, and were moderate or strong in most cases.

    “Importantly, these dissociations could not be explained by heterogeneity in the Openness facets themselves. Intercorrelations between the six facets were all directionally positive, and were moderate or strong in most cases.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    The Assertiveness and Activity Level facets showed generally weaker relations with the LIWC categories than the other facets, with the notable exception that Activity Level was the only Extraversion facet to correlate positively with categories related to goal-directed and achievement-seeking behavior (School, Job/work, and Achievement).

    “Third, the Assertiveness and Activity Level facets showed generally weaker relations with the LIWC categories than the other facets, with the notable exception that Activity Level was the only Extraversion facet to correlate positively with categories related to goal-directed and achievement-seeking behavior (School, Job/work, and Achievement).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingcitation

    People high on Openness tended to use more Articles (ρ = .2, p < .001) and Prepositions (ρ = .17, p < .001) than people low on Openness.

    “This pattern appeared to reflect a fundamental difference in language style rather than content (Chung & Pennebaker 2007); people high on Openness tended to use more Articles (ρ = .2, p < .001) and Prepositions (ρ = .17, p < .001) than people low on Openness, suggesting a potential tendency to favor high-frequency function words at the expense of the lower frequency content words that made up most”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Chung & Pennebaker 2007

      Not checked in the free run.

    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    Excitement-Seeking was the only Extraversion facet to correlate positively with the Negative Emotions, Anger, and Swearing categories or negatively with the Inclusive, Home, and Grooming categories.

    “Second, Excitement-Seeking was the only Extraversion facet to correlate positively with the Negative Emotions, Anger, and Swearing categories or negatively with the Inclusive, Home, and Grooming categories.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    For Neuroticism, most of the heterogeneity appeared to stem primarily from a single facet: Self-Consciousness differed from the other facets in that it showed no positive correlation with negative affect categories, and conversely, was the only Neuroticism facet to correlate negatively with categories related to interpersonal interaction.

    “For Neuroticism, most of the heterogeneity appeared to stem primarily from a single facet: Self-Consciousness differed from the other facets in that it showed no positive correlation with negative affect categories, and conversely, was the only Neuroticism facet to correlate negatively with categories related to interpersonal interaction (Table 3, Figure S1).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    The number of heterogeneous categories was relatively low for Conscientiousness (6 categories), Neuroticism (16), and Agreeableness (15).

    “The number of heterogeneous categories was relatively low for Conscientiousness (6 categories), Neuroticism (16), and Agreeableness (15).”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    The facets of Friendliness, Gregariousness, and Cheerfulness all showed consistent positive correlations with most categories related to positive affect, communality, and interpersonal interaction (e.g., Positive Emotions, Social Processes, 1st Person Plural references, Friends, Family, and Sexual words), whereas the other facets did not.

    “First, the facets of Friendliness, Gregariousness, and Cheerfulness all showed consistent positive correlations with most categories related to positive affect, communality, and interpersonal interaction (e.g., Positive Emotions, Social Processes, 1st Person Plural references, Friends, Family, and Sexual words), whereas the other facets did not.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    Agreeableness correlated positively with some words in the Sexual words category (e.g., 'loves', 'love', and 'hug'; all ρ's >= .2, p's < .001) but negatively with others (e.g., 'porn', 'gay', and 'fuck'; all ρ's < -.21, p < .001).

    “First, Agreeableness correlated positively with some words in the Sexual words category (e.g., ‘loves’, ‘love’, and ‘hug’; all ρ's >= .2, p's < .001) but negatively with others (e.g., ‘porn’, ‘gay’, and ‘fuck’; all ρ's < -.21, p < .001).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Artistic Interests and Emotionality correlated positively with use of the Position Emotion, Inclusive, and Physical States categories, whereas the other four facets generally showed negative correlations.

    “Artistic Interests and Emotionality correlated positively with use of the Position Emotion, Inclusive, and Physical States categories, whereas the other four facets generally showed negative correlations.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    Artistic Interests correlated .33 with Intellect, despite the fact that the two facets showed robust correlations in opposite directions with Positive Emotion word use (ρ's = .18 and -.24, respectively; p's < .0001).

    “For example, Artistic Interests correlated .33 with Intellect, despite the fact that the two facets showed robust correlations in opposite directions with Positive Emotion word use (ρ's = .18 and -.24, respectively; p's < .0001).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Findingown data

    In the present study, 44% of all correlations between the Big Five traits and LIWC categories were statistically significant, yet the mean absolute correlation was only .14.

    “In the present study, 44% of all correlations between the Big Five traits and LIWC categories were statistically significant, yet the mean absolute correlation was only .14.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    A post-hoc analysis showed Agreeableness correlated strongly in opposite directions with two four-word categories related to love and affection ('love', 'loves', 'loved', and 'loving'; ρ = .33) versus sexual behavior ('fuck', 'porn', 'gay', and 'rape'; ρ = -.4), and the two 4-word categories were themselves negatively correlated (ρ= -.15, p < .01).

    “A post-hoc analysis supported this supposition, as Agreeableness correlated strongly in opposite directions with two four-word categories related to love and affection (‘love’, ‘loves’, ‘loved’, and ‘loving’; ρ = .33) versus sexual behavior (‘fuck’, ‘porn’, ‘gay’, and ‘rape’; ρ = -.4), and the two 4-word categories were themselves negatively correlated (ρ= -.15, p < .01).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    83% of participants (N = 576) had scores for the Big Five factors, and 62% (N = 431) had additional facet-level scores.

    “Thus, 83% of participants (N = 576) had scores for the Big Five factors of personality (Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness), and 62% (N = 431) had additional facet-level scores2.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning

    Such writing samples limit the types of analyses researchers can conduct, as it is generally not possible to reliably estimate usage rates for individual words, but only for aggregate categories.

    “As discussed below, such writing samples limit the types of analyses researchers can conduct, as it is generally not possible to reliably estimate usage rates for individual words, but only for aggregate categories.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductreasoning

    Because most participants were recruited via email, the resulting sample was not truly random: bias could arise either because some people were more likely to publish their email address on their blog, or because some people were more likely to respond to the invitation than others.

    “Note that because most participants were recruited via email, the resulting sample was not truly random: bias could arise either because some people were more likely to publish their email address on their blog (a requisite for being contacted), or because some people were more likely to respond to the invitation than others.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisereasoning

    Previous studies typically gather data on a small number of occasions, which cannot establish whether associations remain stable over months or years.

    “Moreover, data are typically gathered from participants on a small number of occasions (often just one) spanning several hours or days; such datasets cannot be used to establish whether any identified associations between personality and language remain stable over much longer periods of time (i.e., months or years), or reflect transient influences (e.g., mood).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductown data

    Controlling for gender and age had negligible effects on the results; for virtually all analyses, > 90% of statistically significant correlations continued to show a significant correlation in the same direction.

    “The fact that females comprised three-quarters of the sample raised the possibility that results might be disproportionately driven by one gender; however, partial correlation analysis demonstrated that controlling for gender and age had negligible effects on the results. For virtually all analyses, > 90% of statistically significant correlations continued to show a significant correlation in the”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    It remains unclear to what extent the results of constrained studies generalize to less constrained real-world situations.

    “It remains unclear to what extent the results of such studies generalize to less constrained real-world situations where people's personalities can influence not only how they write or talk about specific topics, but also what topics they choose to write or talk about (cf. Pennebaker, Mehl, & Niederhoffer, 2003).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Pennebaker, Mehl, & Niederhoffer, 2003 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Studies have consistently identified theoretically predicted correlations between Extraversion and Neuroticism and usage of words related to positive and negative emotion categories.

    “studies have consistently identified theoretically predicted correlations between the dimensions of Extraversion and Neuroticism and usage of words related to a variety of positive and negative emotion categories (Hirsh & Peterson, 2009; Lee, Kim, Seo, & Chung, 2007; Pennebaker & King 1999).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Hirsh & Peterson, 2009

      Not checked in the free run.

    • cites Lee, Kim, Seo, & Chung, 2007

      Not checked in the free run.

    • cites Pennebaker & King 1999

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    A split-half reliability analysis confirmed that the writing sample supported reliable estimation even of word categories with a relatively low base rate: the mean split-half correlation for the 66 categories was .81 (range = .43 - .94), and only 2 categories (Anxiety and Inhibition) had correlations lower than .6.

    “A split-half reliability analysis (i.e., randomly dividing each participants posts into two halves, and then correlating each category's frequency across halves for all participants) confirmed this supposition: the mean split-half correlation for the 66 categories was .81 (range = .43 - .94), and only 2 categories (Anxiety and Inhibition) had correlations lower than .6.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductreasoning

    To minimize Type I error risk from multiple comparisons, interpretation was based primarily on aggregate patterns, and FDR correction (5%) was used.

    “To minimize this risk, interpretation of statistically significant findings was based primarily on the aggregate pattern of results with multiple categories or traits rather than on individual correlation coefficients. Additionally, the presence or absence of statistical significance is reported in key tables using a complementary False Discovery Rate criterion (FDR), which adaptively controls the”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Benjamini & Hochberg, 1995 source

      Not checked in the free run.

    • cites Benjamini & Yekutieli, 2001 source

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    Only the 5,608 words that occurred most frequently across all blogs were analyzed; only blogs containing 50,000 or more words were included in the word-level analyses (N = 406).

    “First, only the 5,608 words that occurred most frequently across all blogs were analyzed3; second, only blogs containing 50,000 or more words were included in the word-level analyses (N = 406).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Word-level results were thresholded at p < .001 in order to minimize the incidence of false positives.

    “Word-level results were thresholded at p < .001 in order to minimize the incidence of false positives.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Highly self-conscious people may be more attuned and concerned with their physical appearance.

    “suggesting that highly self-conscious people may be more attuned and concerned with their physical appearance”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    The pattern of Openness correlations reflects a fundamental difference in language style rather than content, with high Openness favoring high-frequency function words at the expense of lower frequency content words.

    “This pattern appeared to reflect a fundamental difference in language style rather than content (Chung & Pennebaker 2007); people high on Openness tended to use more Articles (ρ = .2, p < .001) and Prepositions (ρ = .17, p < .001) than people low on Openness, suggesting a potential tendency to favor high-frequency function words at the expense of the lower frequency content words that made up most”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Chung & Pennebaker 2007

      Not checked in the free run.

    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Interpretationown data

    Collectively, these findings are consistent with the notion that the affiliative and agentic aspects of Extraversion are relatively distinct and have only partially overlapping correlates.

    “Collectively, these findings are consistent with the notion that the affiliative and agentic aspects of Extraversion are relatively distinct and have only partially overlapping correlates (Depue & Morrone-Strupinsky, 2005; Watson & Clark, 1997).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Depue & Morrone-Strupinsky, 2005 source

      Not checked in the free run.

    • cites Watson & Clark, 1997

      Not checked in the free run.

    • Rests on an assumption: The blog content analyzed was authored by the participant blogger themselves.

  • May combine two claims.

    Study design and conductown data

    A split-half reliability analysis for words suggested that reliability was high to moderate (> .6) for the first 2,000 - 3,000 or so words, and somewhat lower thereafter; even for low-ranked words, split-half correlations generally remained above .4.

    “The analysis suggested that reliability was high to moderate (> .6) for the first 2,000 – 3,000 or so words, and somewhat lower thereafter. However, even for low-ranked words, split-half correlations generally remained above .4, a level considered acceptable for present purposes given that word-level analyses focused primarily on the aggregate pattern of associations with personality rather than”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    The positively correlated words in the Space category occurred relatively infrequently and predominantly expressed relational spatial concepts (e.g., 'among', 'between', 'further', and 'under').

    “The explanation for this seemingly paradoxical finding was that the positively correlated words occurred relatively infrequently and predominantly expressed relational spatial concepts (e.g., ‘among’, ‘between’, ‘further’, and ‘under’).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Individual differences in personality have previously been linked to differences in linguistic style in laboratory and experience-sampling settings.

    “Individual differences in personality have previously been linked to differences in linguistic style in laboratory and experience-sampling settings (Fast & Funder, 2008; Hirsh & Peterson, 2009; Mehl et al., 2006; Pennebaker & King 1999).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Fast & Funder, 2008

      Not checked in the free run.

    • cites Hirsh & Peterson, 2009

      Not checked in the free run.

    • cites Mehl et al., 2006 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    High-level associations between broad traits and aggregate word categories can mask, and in some cases even contradict, robust but relatively narrow associations.

    “The results demonstrate that high-level associations between broad traits and aggregate word categories can mask, and in some cases even contradict, robust but relatively narrow associations.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Hirsh and Peterson (2009) found fewer than 15% of tested effects to be statistically significant, yet obtained a much larger mean statistically significant r of .23.

    “In contrast, Hirsh and Peterson (2009) found fewer than 15% of tested effects to be statistically significant, yet obtained a much larger mean statistically significant r of .23.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Hirsh and Peterson (2009)

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Highly Open individuals use "big words" more often, effectively resulting in strong positive correlations for many individual words.

    “it is reasonable to suppose that highly Open individuals use “big words” more often, effectively resulting in strong positive correlations for many individual words (but conversely, producing negative correlations for LIWC categories that tend to be made up of a relatively small number of high-frequency words).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    The primary consequence of selection bias would be a restricted distribution of personality scores among self-selected participants, which would generally tend to deflate effect sizes and statistical power, leading to results that actually underestimate the magnitude of true population effects.

    “However, it is important to note that the primary consequence of selection bias would be a restricted distribution of personality scores among self-selected participants, which would generally tend to deflate effect sizes and statistical power, leading to results that actually underestimate the magnitude of true population effects.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Bloggers who participated had systematically different personalities from those who did not.

  • May combine two claims.

    Interpretationcitation

    The results converge with other recent findings suggesting that people do not present themselves in an idealized and overly positive way online, and maintain online identities that reflect the way they genuinely see themselves and are seen by others.

    “The results converge with other recent findings suggesting that, contrary to popular wisdom, people do not present themselves in an idealized and overly positive way online (Turkle, 1997), and maintain online identities that reflect the way they genuinely see themselves and are seen by others (Back et al., 2010; Vazire & Gosling, 2004).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Turkle, 1997

      Not checked in the free run.

    • cites Back et al., 2010 source

      Not checked in the free run.

    • cites Vazire & Gosling, 2004

      Not checked in the free run.

  • May combine two claims.

    Study design and conductnone given

    All language variables in the present study were ultimately derived from simple counts of word use, and no contextual factors or higher-order semantic variables were taken into consideration.

    “Finally, although the present study explored language use at the level of both aggregate categories and individual words, all language variables were ultimately derived from simple counts of word use, and no contextual factors or higher-order semantic variables were taken into consideration.”
    Quote in the paperSubstantial claim
41 of 41

  • The paper relies on this without support.

    Bloggers were unaware that their writing would be analyzed in relation to personality at the time of writing.

    “Because bloggers were free to write about any topic of their choosing and were (at the time of writing) unaware that their writing would be analyzed in relation to personality, the data provided an naturalistic window into the influence of personality on language use that could not be influenced by demand characteristics.”

  • The paper relies on this without support.

    The blog content analyzed was authored by the participant blogger themselves.

    “The contents of participants' blogs were subsequently downloaded and parsed using a set of custom scripts written in the Ruby programming language.”

  • The authors state this as an assumption.

    Bloggers who participated had systematically different personalities from those who did not.

    “It is reasonable to suppose that bloggers who participated had systematically different personalities from those who did not (e.g., they might be more Agreeable or Open).”
3 of 3

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Conflicting counts of words analyzed

    Minor

    The methods say 5,608 individual words were analyzed, but the results say the analysis used 5,068 words.

1 of 1

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Yarkoni, T.
Affiliations
University of Colorado Boulder (US)
Corresponding author
Yarkoni, Tal
ORCID
Yarkoni, Tal
Venue
Journal of Research in Personality
Publisher
Elsevier BV
Published
2010-06
Volume
44
Issue
3
Pages
363-373
ISSN
0092-6566 (ISSN-L)
DOI
10.1016/j.jrp.2010.04.001
PMC
PMC2885844
OpenAlex
W1973871257
Cited by
637
References
41
Citations by year
2026: 54 · 2025: 40 · 2024: 41 · 2023: 39 · 2022: 57 · 2021: 73 · 2020: 56 · 2019: 62 · 2018: 37 · 2017: 44 · 2016: 32 · 2015: 29 · 2014: 25 · 2013: 22 · 2012: 15
FWCI
16.82
Type
journal article
Language
en
Open access
Licence
Locations
  • Journal of Research in Personality (publishedVersion)
  • PubMed (publishedVersion)
  • PubMed Central (acceptedVersion)
Funders
  • National Institutes of Health
  • National Institute of Nursing Research (F32NR012081)
Topics
  • Social Sciences › Psychology › Clinical Psychology › Personality Traits and Psychology
  • Social Sciences › Psychology › Social Psychology › Mental Health via Writing
  • Physical Sciences › Computer Science › Artificial Intelligence › Authorship Attribution and Profiling
Keywords
personality, word usage, bloggers, English vocabulary, language and personality, replication

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review