Samuel D. Gosling

The built environment serves as the setting for most daily human experiences. Geography studies have long established that places significantly influence human perception, behaviors, health outcome...

Does our speech—“what” we say and “how” we say it—reveal how we subjectively feel in daily life? Speech and emotion are often thought to be linked, but most supporting evidence comes from controlled laboratory settings and focuses on expressed, enacted, or externally labeled emotions, leaving open the question of whether naturalistic speech reflects subjective emotional experience in daily life. To address this question, we extracted both modern foundation-model embeddings and established speech-analysis variables from brief, prompted smartphone recordings, including Linguistic Inquiry and Word Count features and prosodic descriptors, and used them as inputs to supervised machine learning models predicting concurrent self-reported emotional states (934 participants; 12,285 observations). Cross-validated models evaluated on unseen participants captured information about self-reported emotional states (contentment: median Spearman ρ = 0.37; sadness: ρ = 0.24; arousal: ρ = 0.35), with spoken content represented by foundation-model text embeddings carrying the strongest emotional signal. Interpretability analyses provided further insights into the linguistic characteristics of everyday emotional language. These findings provide large-scale evidence that brief speech samples contain information about subjective emotional states in daily life.

Person-centered psychology has primarily used exploratory discovery algorithms, such as latentprofile analysis and cluster analysis, to identify population subgroups (e.g., “Resilient”). Thesemethods face several well-documented limitations: identified profiles vary across datasets(sample dependence), the number of profiles relies on ambiguous stopping rules, and forcingcontinuous personality configurations into discrete categories can conflate the measurementgranularity of Likert lattices with genuine structure. In response, we introduce fixed-framearchetypal analysis (FFAA), which represents each individual as a continuous compositionalmixture relative to fixed theoretical extrema (the absolute corners of the Big Five hypercube)using closed-form multilinear interpolation. The method contributes at two levels. For theindividual respondent, FFAA is a lossless reparameterization of trait scores into a fixed, sample-independent coordinate system, providing (1) fixed profile definitions, (2) high-resolutionmeasurement of configural differences, and (3) computational scalability. Across respondents,because compositions are nonlinear functions of the traits, averaging them encodes thecovariation among traits that marginal averages do not retain. We illustrate the individual-levelproperties through simulation studies and establish a normative baseline from over 10 millionrespondents; at the aggregate level, these data characterize the maturity principle across the adultlifespan not as discrete stages but as a continuous geometric migration that is broadly similaracross regions. We propose that moving from “profile discovery” to “compositionalmeasurement” offers a mathematically grounded approach to person-centered psychology,though establishing predictive validity and extending FFAA to other personality measures remainimportant next steps.

Satisfaction with life (SWL) is central for mental well-being, making it a key focus for practitioners and researchers alike. Individual differences in SWL relate to self-reported patterns of everyday physical activity during daytime and overnight. However, these associations rarely replicate in objective physical activity measures, leaving real-world SWL manifestation unclear. Here, we use an interpretable machine learning approach to examine how physical activity patterns (across four weeks) captured via smartphones predict SWL in a diverse dataset (N = 2,272) spanning three countries, years, seasons, and sample demographics. The cross-validated results demonstrate that extremes (minima, maxima) and variation (standard deviation, skewness) in physical activity were particularly informative in robust out-of-sample predictions of SWL (r_Md = .18, r_IQR = [.14, .21]), and that personality traits were associated with the extent of activity predictiveness. Overall, these findings highlight the importance of more nuanced approaches to studying associations between SWL and everyday physical activity.

This study examines how age moderates gender differences in Five-Factor Model personality traits across cultures. Prior research suggests that women generally score higher than men on these traits, with a larger gender gap observed in personality traits in more economically developed, egalitarian, and individualistic countries. Gender differences also vary across age, yet the timing of these changes and the extent to which they are culturally universal remain unclear. Using multilevel modeling in a large cross-sectional online sample (N = 3,038,567; ages 10-65 years, 68 countries), we found that age significantly moderated gender differences in Five-Factor Model personality traits, with substantial cross-national variation. The gender gap in Neuroticism, Extraversion, and Openness was wider among older adults in countries with higher levels of individualistic and self-expressive values, higher Human Development Education Index scores, and later mean ages at first marriage and childbearing for women. These findings suggest that broader educational, cultural, and family-related contexts may shape the expression of personality traits across the lifespan. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

This study examines how age moderates gender differences in Five-Factor Model (FFM) personality traits across cultures. Prior research suggests that women generally score higher than men on these traits, with a larger gender gap observed in personality traits in more economically developed, egalitarian, and individualistic countries. Gender differences also vary across age, yet the timing of these changes and the extent to which they are culturally universal remain unclear. Using multilevel modelling in a large cross-sectional online sample (N = 3,038,567; ages 10–65 years, 68 countries), we found that age significantly moderated gender differences in FFM personality traits, with substantial cross-national variation. The gender gap in Neuroticism, Extraversion, and Openness was wider among older adults in countries with higher levels of individualistic and self-expressive values, higher Human Development Education Index scores, and later mean ages at first marriage and childbearing for women. These findings suggest that broader educational, cultural, and family-related contexts may shape the expression of personality traits across the lifespan.

Age, cohort, and period effects are three ways to explain personality trait change over time. While past research mostly focused on age differences, showing relatively consistent patterns, evidence for cohort differences is more mixed, and period differences have hardly been examined. However, age, period, and cohort are exactly collinear (age = period-cohort), such that estimates are likely confounded and always hinge on so-called identification assumptions. Identification assumptions shape substantive conclusions, and inappropriate or inconsistent strategies may explain past discrepant findings. To address this age-period-cohort identification problem in personality change, we leveraged four large-scale (N_total > 2 Mio) repeated cross-sectional data sets from 2003 to 2022. Our aims were to demonstrate how identification assumptions common in personality studies impact estimates for age, cohort, and period and to use weaker, substantively informed assumptions to narrow down the range of plausible solutions. Results showed that common identification strategies of constraining one temporal effect to zero can dramatically affect conclusions-less for age-graded, but more for generational differences. Using weaker assumptions, our results indicated that all three factors-age, cohort, and period-likely contribute to trait differences over time. Assuming age-graded change in a certain direction revealed cohort-related decreases in extraversion, openness, and neuroticism and increases in agreeableness, alongside period-related increases in extraversion, openness, and conscientiousness. This suggests that several previously assumed cohort differences may actually be driven by period effects, overlooked due to strong identification assumptions. Overall, highlighting the importance of appropriate identification strategies, our results offer unique insights into factors driving trait differences over time. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

Does our speech – what we say and how we say it – reveal information about our emotional experiences? Scholars have long made assumptions about such links but most evidence is derived from emotion expressions in controlled laboratory work, with only limited studies of emotional experience in naturalistic settings. Using both modern foundation models and traditional speech analysis methods (LIWC and prosodic descriptors), we tested whether brief, prompted speech samples recorded via smartphones could predict concurrent, self-reported emotional states in daily life (N = 934 participants; 12,285 observations). Cross-validated models showed that speech samples predict emotional states (contentment: r_md = .37; sadness: r_md = .23; arousal: r_md = .34), with spoken content captured by foundation-model text embeddings revealing the strongest emotional signal. The findings offer insights into the linguistic characteristics of naturalistic emotional speech, with implications for scalable emotion inference via speech in daily life.

Personality traits (e.g., the Big Five) shape human behavior, decision‐making, and life outcomes. Evidence from various countries suggests that these traits are not randomly distributed but follow systematic regional patterns, fueling interest in their geographical variation. We analyze N = 22,225 geo‐coded personality assessments, aggregated at regional levels. Using Hot Spot analysis (Getis‐Ord Gi), we identify spatial clusters of personality traits and test their associations with health and well‐being indicators at the regional level. Even though Sweden has a reputation for equality, our analysis reveals notable spatial differences in personality composition. The data reveal overlapping geographical clustering in Extraversion, Conscientiousness, and Emotional Stability (i.e., low Neuroticism), traits that are established predictors of better psychological functioning. These traits co‐occur at relatively higher levels in southern Sweden, particularly in the Scania region, whereas they co‐occur at relatively lower levels in parts of central Sweden. These patterns are descriptively supported by municipal‐level correlations with health and well‐being variables, although the associations are not uniform across all measures. In view of these results, we reflect on how these clusters may connect to long‐standing regional histories, including industrial legacies. Such interpretations remain speculative, and we outline a future research agenda to further investigate the causes and consequences of regional personality variation, as well as its possible societal implications.

Socializing, moving, working, and leisure form the foundation of human experience. We examined whether these foundational, ostensibly nonpolitical behaviors are nevertheless bifurcated along political fault lines, revealing "lifestyle polarization." Study 1 quantified the association between political identity and 61 social, movement, work, and leisure behaviors collected from smartphone sensors and logs (i.e., GPS, microphone, calls, texs, unlocks, activity recognition) and ecological momentary assessments (i.e., querying activity level, activity type, interaction partners, locations) at multiple temporal levels (i.e., daily, mornings, afternoon, evenings, nights, weekends, weekdays) in a sample of up to 1,229 students on a college campus. We found that liberals and conservatives behave differently in everyday life; the behavioral differences were small but robust, not accounted for by other plausible factors (e.g., demographics), and most pronounced in the leisure domain. Study 2 showed that the behavioral differences between liberals and conservatives were not accurately discerned by other students, who overestimated the extent of lifestyle polarization present on their campus. Together, these studies suggest that political identity has penetrated some of the most foundational aspects of everyday life, but not to the degree that people think. We discuss how communities may feel divided not only because of deep ideological disagreements across partisan lines but also because such disagreements are accompanied by distinct lifestyles-both real and (mis)perceived-that may prevent liberals and conservatives from engaging in cross-partisan contact and developing mutual understanding. (PsycInfo Database Record (c) 2025 APA, all rights reserved).