Intelligence最新文献

This study examined how general cognitive ability, g, is formed from preschool to late childhood and how it interacts with specific mental processes. A large sample (N = 401), about equally drawn from each of the age years 4 through 12 were examined with a large array of attention control, working memory, relational integration, Raven-like matrices, and awareness of perceptual and inferential origins of representations. Confirmatory Factor Analysis examined if g is a reflective construct causally affecting these processes or a formative construct gradually emerging from mastering these processes, and how it varies throughout this age period. We found that g is a reflective construct gearing on a core of relational integration and mental awareness, which changes in cycles: it is primarily based on attention control and perceptual awareness from 4 to 6, inferential awareness and working memory from 7 to 9, and inhibition, inferential awareness, and complex inductive reasoning from 10 to 12 years. The implications of the study for the century-old dispute about the nature and development of human general intelligence and modern theories of intelligence and cognitive development are discussed.

Generational IQ test score changes in the general population (i.e., the Flynn effect, typically reported as increases of 2–4 IQ points per decade) have recently been observed to behave inconsistently. It has been speculated that these inconsistencies may be attributable to the well-established negative relation of test score gains with psychometric g. Here, we provide the first direct empirical investigation of cross-temporal changes in the positive manifold of intelligence. In this cohort-comparison study, we examined performance changes in two population-representative Germanophone samples (N = 1267) across six measurement-invariant intelligence subscales from 2005 to 2024. Our analyses revealed substantial declines in single-factor analysis-based g assessments (ΔR² range: ‐.037 to -.066) from 2005 to 2024. Despite this decrease in the positive manifold strength, we observed meaningful test score increases in all domains (d range: 0.18 to 1.24), with the largest gains in the lower tail of the intelligence distribution (i.e., conforming to Rodgers', 1998, idea of narrowing ability distributions). Our findings provide direct evidence for a decreasing strength of the positive manifold of intelligence as a noticeable driver of the accumulating evidence for negative Flynn effects, which may be a consequence of increasing ability differentiation in the general population.

We studied trajectories of school achievement in England to determine sex differences in performance and changes in these differences throughout students' development. Using a sample of 5795 children from England born in 2000–2001, this secondary data analysis examined sex differences across a range of school subjects, including differences at the upper and lower tails of the distribution of performance grades. We expected trajectories to differ by subject and to find support for greater male variability in each subject. We found a small male advantage in mathematics at age 11 but no sex differences at ages 7 and 16. Girls achieved higher language grades at each age, but this advantage was notably wider at age 16. Unlike other educational data, there were no sex differences in science achievement at ages 7 and 11 and a small female advantage in science, biology, and chemistry at age 16. Boys' school grades were more variable than girls' in English, reading, and writing at each age. Boys' STEM grades were not consistently more variable than girls' STEM grades. Sex differences were larger at the lower tail in English and the upper tail in mathematics and more balanced in science after age 7. Trajectories of sex differences are age- and subject-specific. By age 16, fewer boys achieved the upper grades, and more boys achieved the lower grades in mathematics and language than at age 11, and we found a female advantage in most school subjects. Implications for practice and directions for future research are discussed.

Large language models (LLMs) are advanced artificial intelligence (AI) systems that can perform a variety of tasks commonly found in human intelligence tests, such as defining words, performing calculations, and engaging in verbal reasoning. There are also substantial individual differences in LLM capacities. Given the consistent observation of a positive manifold and general intelligence factor in human samples, along with group-level factors (e.g., crystallised intelligence), we hypothesized that LLM test scores may also exhibit positive inter-correlations, which could potentially give rise to an artificial general ability (AGA) factor and one or more group-level factors. Based on a sample of 591 LLMs and scores from 12 tests aligned with fluid reasoning (Gf), domain-specific knowledge (Gkn), reading/writing (Grw), and quantitative knowledge (Gq), we found strong empirical evidence for a positive manifold and a general factor of ability. Additionally, we identified a combined Gkn/Grw group-level factor. Finally, the number of LLM parameters correlated positively with both general factor of ability and Gkn/Grw factor scores, although the effects showed diminishing returns. We interpreted our results to suggest that LLMs, like human cognitive abilities, may share a common underlying efficiency in processing information and solving problems, though whether LLMs manifest primarily achievement/expertise rather than intelligence remains to be determined. Finally, while models with greater numbers of parameters exhibit greater general cognitive-like abilities, akin to the connection between greater neuronal density and human general intelligence, other characteristics must also be involved.

Planning is a core component of executive functioning that has been hypothesized to be central to many activities in daily life and occupational settings. Despite its practical and theoretical relevance, there is a lack of psychometric tests, whose item parameters can be predicted by item design features, that have been shown to be linked to cognitive processes involved in planning (=radicals). In the present article the automatic min-max approach was used to construct k = 140 (study I: N = 1573) and k = 17 (study II: N = N = 548 Austrian and N = 572 Italian adolescents) scheduling problems measuring planning. The psychometric quality of the items was evaluated in three studies. The results indicated, that the 1PL Rasch model and the Linear Logistic Test model fitted the data reasonably well, and that the item- and basic parameter estimates can be assumed to be invariant across relevant socio-demographic (study I and II). The radicals jointly explained 89.30% of the variance in the item parameter estimates, and all of them contributed significantly to the prediction of the item parameters. Furthermore, planning – as measured by the scheduling problems and the Tower of London (TOL-F) – was moderately correlated with G_c, G_q, and G_v, and highly correlated with G_f (study III: N = 249). By contrast, G_f was highly correlated with planning ability and the other three second stratum factors. Thus, G_f and planning ability differ in their structural relation to other second stratum factors, which provides evidence that planning ability cannot be regarded to be synonymous to G_f. The article discusses the theoretical and practical implications of these findings.

Finke, Scheiner, Giurfa, and Avarguès-Weber (2023) published correlational data on the performance of honeybees (Apis mellifera) in three learning tasks (associative, reversal, and negative patterning, capturing the domains of operant conditioning, executive-functioning-like ability, and inhibition plus configural processing, respectively) evaluated under both visual and olfactory stimulus conditions. They speculate that general cognitive ability (GCA) may be weakly causing all-positive correlations between performance in these different learning modalities, but do not formally test this possibility. A factor-analytic model applied to Finke et al. (2023) data revealed the presence of two perfectly congruent GCA factors (one for each stimulus condition). Both exhibited all-positive loadings, with the visual factor accounting for 46.8% of the performance variance and the olfactory factor accounting for 52.3%. Diagnostic statistics confirmed that in both stimulus conditions, the correlation matrices were adequate for factor analysis. These findings support extant hypotheses that GCA influences covariation between cognitive measures in honeybees, and constitute the first formal potential demonstration of GCA in an invertebrate. It is argued that GCA might be ubiquitous with respect to metazoans possessing organized nervous systems, perhaps because it convergently evolved multiple times in independent phylogenies, this being a key prediction of Christopher Chabris' Law of General Intelligence. Indeed, GCA has now been identified in insect, avian, mammal, and fish taxa. Some “primordial” aspects of GCA may even by basal to metazoans, and experiments employing Caenorhabditis elegans are suggested that could potentially shed light on such aspects. The findings are also strikingly inconsistent with evolutionary and comparative psychological theories positing a “modules first” understanding of cognitive evolution, such as one recent proposal that smaller brains cannot accommodate structures that give rise to GCA. Other theoretical implications of these findings are discussed.

Numerous studies have explored the complex relationship between culture, cognitive competence, and economic performance globally. However, findings from these investigations vary significantly and occasionally contradict each other. This study delves into this connection by analyzing variables or dimensions from three distinct models of national culture concerning the results of the Programme for International Student Assessment (PISA) and the economic strength and growth of 86 nations between 2000 and 2018. Religious affiliations emerge as a significant statistical explanatory factor, accounting for a substantial portion of the variance in overall PISA performance, ranging from 24% to 40%. Remarkably, an underutilized cultural model, the Axiological Cube, surpasses others, exhibiting explanatory power ranging from 45% to 63%. Path analyses rooted in both Human Capital Theory and Cognitive Capitalism Theory reveal that cultural variables exert their influence on economic growth mainly indirectly through their impact on student competence. Cultural variables exhibit robust predictive capacity for overall student competences, as indicated by PISA mean scores. However, they prove inadequate in explaining certain cognitive competence patterns, such as disparities in inequality and subject-specific variations, like mathematics and reading. This study also highlights uncertainties surrounding the effects of Confucianism and East Asian religions, prompting further discussion and investigation.

Σ words: 200.

The developmental cascade model, elaborated by Fry and Hale (2000) emphasizes the role of age-related increases in processing speed and working memory for the development of fluid intelligence. Given the intimate relationships between intra-subject variability and the aforementioned constructs, the present study set out to determine the role of intra-subject variability within the pathways outlined in the developmental cascade model, postulating a fundamental role of intra-subject variability for the development of processing speed, working memory and fluid intelligence. To that end, N = 403 participants aged 8–18 years took a testing battery including choice reaction time tasks to measure processing speed and intra-subject variability as well as span, operation span and coordination tasks to measure working memory within the empirical framework of Oberauer et al. (2003). Cattell's Culture Fair Test (CFT-20 R) was used to measure fluid intelligence. Our results confirm the well-known close relationships between processing speed, working memory, and fluid intelligence, and show that intra-subject variability is also closely related to these constructs. The results of the present study suggest the extension of the developmental cascade model by the inclusion of intra-subject variability as a fundamental construct.