Nature Human Behaviour最新文献

To discern speech or appreciate music, the human auditory system detects how pitch changes over time (pitch motion). Here, using psychophysics, computational modelling, functional neuroimaging and analysis of recorded speech, we ask whether humans can detect pitch motion using computations analogous to those used by the visual system. We adapted stimuli from studies of vision to create novel auditory correlated noise stimuli that elicited robust pitch motion percepts. In psychophysical experiments, we discovered that humans can judge pitch direction from spectrotemporal intensity correlations. Robust sensitivity to negative spectrotemporal correlations is a direct analogue of illusory 'reverse-phi' motion in vision, constituting a new auditory illusion. Functional MRI measurements in auditory cortex supported the hypothesis that human auditory processing may employ pitch direction opponency. Linking lab findings to real-world perception, we analysed recordings of English and Mandarin speech and found that pitch direction was signalled by both positive and negative spectrotemporal correlations, suggesting that sensitivity to both types confers ecological benefits. This work reveals how motion detection algorithms sensitive to local correlations are deployed by the central nervous system across disparate modalities (vision and audition) and dimensions (space and frequency).

Animal domestication and development of pastoralism in southwest Asia revolutionized human subsistence strategies. Various centres of ruminant domestication and diffusion routes of agropastoralism have been identified. The area between the northern and central Zagros Mountains on the Iranian Plateau is a cradle for goat domestication and eastward spread of agropastoralism. However, the early exploitation of ruminant milk by pastoral communities in the Zagros remains insufficiently studied. Here we show residues of caprine dairy products that were detected from the analysis of lipid residues in pottery vessels and protein residues in human dental calculus. These results, combined with the faunal spectra and radiocarbon analyses directly on the dairy residues, show that sheep and goat dairy products were widely exploited in the Zagros from the seventh millennium BC. This pattern parallels the contemporaneous exploitation of cattle milk in Anatolia. Neolithic communities in both regions reveal similarly complex dynamics of early ruminant milk use, marking the emergence of independent yet synchronous trajectories in the diffusion of agropastoral lifeways.

Language barriers and translation costs are persistent obstacles to communication and have particularly pronounced economic impacts in technical domains. Here we provide causal evidence on the effects of language barriers on the speed and extent of knowledge diffusion by exploiting a change in US patent policy that resulted in earlier disclosure of English-language technical knowledge from Japan. Using a targeted sample of 2,770 citations from US-based inventors to Japanese inventions, we find that language barriers accounted for almost half the diffusion lag of Japan-originating knowledge to US-based inventors, relative to Japan-based inventors. This acceleration is significant only for firms with limited ability to translate (small research and development scale, or little involvement in the Japanese market) and is more pronounced for the diffusion of high-quality inventions, suggesting difficulties in quality-targeted translation. Thus, early publication of patent applications provides a substantial public good for cumulative innovation through accelerated access to translated foreign patents.

There are increasing calls for economic assistance in the form of social safety nets (SSNs) to be designed and implemented to promote women's economic inclusion and agency, contributing to closing gender disparities globally. Here we investigate the extent to which SSNs affect women's economic achievements and agency through a systematic review and meta-analysis of randomized controlled trials implemented in low- and middle-income countries. We searched six databases utilizing search strings in English, French and Spanish through December 2024. Studies were assessed for risk of bias using an adapted version of the Joanna Briggs Institute critical appraisal tool. Our sample includes 1,307 effect sizes from 93 studies, representing 218,828 women across 45 low- and middle-income countries. Using robust variance estimation meta-analysis, we show significant overall pooled effects (Hedges' g = 0.107, P < 0.001, 95% confidence interval (CI) 0.085-0.129), driven by increases in economic achievements (productive work, savings, assets and expenditures) and agency (voice, autonomy and decision-making). We find significant treatment effects for unconditional cash transfers (Hedges' g = 0.128, P < 0.001, 95% CI 0.097 to 0.159), social care services (Hedges' g = 0.122, P < 0.001, 95% CI 0.071 to 0.174), asset transfers (Hedges' g = 0.115, P < 0.001, 95% CI 0.071 to 0.160) and public work programmes (Hedges' g = 0.127, P = 0.031, 95% CI 0.015 to 0.239). We find comparatively smaller effects for conditional cash transfers (Hedges' g = 0.059, P = 0.019, 95% CI 0.011 to 0.108) and found no evidence of effects for in-kind transfers. SSNs can empower women economically and socially; however, limitations and evidence gaps remain, including the need for further rigorous testing of design and operational components, the role of contextual factors and cost-benefit analysis with a gender lens.

A long-standing challenge for psychology and neuroscience is to understand the transformations by which past experiences shape future behaviour. Reward-guided learning is typically modelled using simple reinforcement learning (RL) algorithms. In RL, a handful of incrementally updated internal variables both summarize past rewards and drive future choice. Here we describe work that questions the assumptions of many RL models. We adopt a hybrid modelling approach that integrates artificial neural networks into interpretable cognitive architectures, estimating a maximally general form for each algorithmic component and systematically evaluating its necessity and sufficiency. Applying this method to a large dataset of human reward-learning behaviour, we show that successful models require independent and flexible memory variables that can track rich representations of the past. Using a modelling approach that combines predictive accuracy and interpretability, these results call into question an entire class of popular RL models based on incremental updating of scalar reward predictions.